diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/lm_head/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/lm_head/_0.pt
index b26634d31627d1a0044c3f3651ff144ed5eb8f4c..257f78e580e29f2b211849ea7bc8ce9dac122cc9 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/lm_head/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/lm_head/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/lm_head/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/lm_head/_s.pt
index 45c403718a90bc915d3f3e1aa42781945fe8407a..3026bd3e072de29ff79eb46c76f95508661620c0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/lm_head/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/lm_head/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.fc1/_s.pt
index f01b89f382c4d480e4455cf0d9ea9f8bbae693ed..fc6af853cc9ae73c3ebad72874cc7f3b7d95cd35 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.fc2/_s.pt
index 9ea884d126b233671dc7c6731313caebf190a53a..c78213a74d932355041c07244edf2acebd997407 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.k_proj/_s.pt
index cb44cc6a554eaff29b07bff8f28c762601ed4e01..d2d33c5100b957979bae7fc1fa4ca7f379ddffe8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.out_proj/_s.pt
index 19e1c7301118c5ea8e27b6344e2e02e3de5dde3b..17b23ac2ef9bab5f620f8bcda610096f0f1e2319 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.q_proj/_s.pt
index e8c66349d5747f514bd59ceb8485d5ab3d6596a8..051821bc7911a905292936d85ae19ffe210df6d0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.v_proj/_s.pt
index 4a021359433005d9d34e491baaaffe9c4e161b19..1f3ae696a58020eede842bd9ca33bf7f8c29638a 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.0.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.fc1/_s.pt
index f7749af032f67cebf1c1170fc90e256f31f2df59..d6f2a92496febfa0aaeb2fe0e602e42fc642ebac 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.fc2/_s.pt
index 263fa2efc0c28219aa673ade31b472e94f768fa2..fa0aa75e79ea5bfae5a4dbe76db0fc8b74de8481 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.k_proj/_s.pt
index 115e6c548d5ebcf8b86bd1bfa8b9f191142f76fa..28eddfa09f8d8fba94f6376952c2d748401fe08c 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.out_proj/_s.pt
index 41447065ac478ace0f31ac528d3913ba0a29f86f..246a42e32fa50aac0f1c1ab11ad303d6eb862a9e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.q_proj/_s.pt
index 70931376d39482a20976068730fa792b993148d1..f2d801dd1f7f513d6acc145b69fc9c51e9de98cb 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.v_proj/_s.pt
index 4782a6ac8fd23b6d63be8cf97a751b5e0777aa54..70ad134e07da47bef8eeada701780fb609177def 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.1.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.fc1/_s.pt
index 877a7cc09395a324ba6257dd7c09b59bbd831110..8d6dcbe6f1458037098c0c12aa84cfe5ed288811 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.fc2/_s.pt
index 0df44ee17f0afeaf274f06373fdbab3b25fd5618..c966ca71ada720003e0b8ed01ea65483e187b2c6 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.k_proj/_s.pt
index da67c686250da18fce691f9e3b07854c8ab5e643..b36dce8cee41610f23e4b6299e15d04d94fc7148 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.out_proj/_s.pt
index 428fa664ba59be4db72342fffbaa08bc41b522cc..c59869e5c0b3e9f0ead4cb96b31d10c53fe4b99f 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.q_proj/_s.pt
index 87332fa56953fee75158de7fc8d7d1b17bb7b3a3..01b55a33bfae081a601ca66e76ada08b07aed013 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.v_proj/_s.pt
index 59c7f562d758a02c06c77646b1b4fcb8a27872d4..d7911ef5b1228e657826907c36b376553bc802db 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.10.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.fc1/_s.pt
index 60f2c812813dffc829da4a5419373db62e6fff82..ebe6310b610c5218bf400aba23400736ff72b949 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.fc2/_s.pt
index 382f0cb9d7c128d96ef9edf486a037a1a2431a13..aa205596af29c69e2e89032390ac6d98cd19d937 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.k_proj/_s.pt
index edaa5e99db016fff1aa8299a2785770919fad827..04768280b3e3d57914228d31803753fd03e09edb 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.out_proj/_s.pt
index 3a3a0f5d66e464eb91cf9be31e0b6ce358b0fb30..6384d140079fbf1b592e224c33ca4ab458816b88 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.q_proj/_s.pt
index 702fea2128916dfa137fba3bf6b3268fe222d771..7d68edbdada295e84923dca6ddb79205aacac99f 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.v_proj/_s.pt
index 96f3ab6cd43819dfc0385933cad19a17fb2722bb..b03744824cfb67d1c158125b8f9cf3ec30202df1 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.11.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.fc1/_s.pt
index f04587216df54ff04615c1107ba3a7df120c229f..9ae083ab38802989275e67b3a9b99f808888fe24 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.fc2/_s.pt
index 10b8925c78ddbc72c73d3a4bc3384ceb2d0d15e4..7317c3bf02a70eba78e491c569717db3ae4ce8bd 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.k_proj/_s.pt
index ab4da2f8471e075f712656daf576815d709a5c54..c695b7184dc882958e5c48e9205ea50b7cd3bc2b 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.out_proj/_s.pt
index 46fec2c5159d2781b0ad658af0fe0deacf2427b0..a9f14b5952eff80720d1c4b7de438bee6c725bf0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.q_proj/_s.pt
index ac28e60c180b8872bee7494faab0d8208baa3907..2b2bf839a18fe3ad7c40594d1b4203cdec5504bc 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.v_proj/_s.pt
index 45dac7c4f7e91bcddc44bffa09896f9537165aad..31d9eb53e8060637d441bbff16faecd0622ff115 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.12.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.fc1/_s.pt
index 72d235e8cdf388ae809cc2d2f01faf7297112111..cf1bfb8aa01324502a696b0bc85e40ad9c96ff0c 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.fc2/_s.pt
index d9986ff375a25fa4c78b4cee7b53c736a5e65771..2d95b1379d37c64de758b011454322ea32ba9463 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.k_proj/_s.pt
index 01bf1594dfc108b8c6ddc7dcfd8ba0757a55fb79..c88ce566d4b9b57c482fdf975e0e0331f5876d21 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.out_proj/_s.pt
index 04233b2293a91b648a2737c7560143140d4352c5..93b82c1f79a1945e1a6a7da820d62bd92e7169b0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.q_proj/_s.pt
index e60de2f3537452b0fc1b1460c74f2316ab784e6f..4996ffca40a861827decd20be5c6c10c2a7a42ac 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.v_proj/_s.pt
index 1b065992ae884b0089a3280b819168cfd6bdb0c3..ca63b0914b0a26510e9b6476ca614b4d035ac5d2 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.13.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.fc1/_s.pt
index 7c2a8d3d807b6027f2239c3926b34932b507f5a8..1d64c9bd7557b027282eb86209cf4b26e1b838a2 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.fc2/_s.pt
index c1a175b98615d36a43eac35de94af73c50c6c290..7b0b00feb1c292741f803f441118f4ba22af965f 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.k_proj/_s.pt
index 79c753d76bd32e3f74a1d12326f7fccfc084b4f4..41289228467e6f79bde74f231c0717c4778c226a 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.out_proj/_s.pt
index 4e44c9c48044ba38b5d2ec6c51472ab1419729c6..f25c7edd1fa48b53977d7710fe7d3cf56ee1d3a2 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.q_proj/_s.pt
index 094773c04314dc749a08615c86969fd97276835a..c33a2355ccf3d7f3dd64cad58743c8619eb75f5a 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.v_proj/_s.pt
index a8a4f0a1d489b7fc1d7cbde993d462578ee197a8..09eb37a987292d6fce2d7ca01af84c4fb6f8a104 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.14.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.fc1/_s.pt
index 6ba52bab369842268426a2f0c57c72073a0c6364..4e283dd98c0f2382a463355a5693d9acb1d4d810 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.fc2/_s.pt
index cf42be0e151268cfec34205003d8f9c9b34f7e37..ae0857d6b5d9b7da826608bee5876f229a9b38c4 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.k_proj/_s.pt
index 541e601664b5c63248fa3733f81f6a7732b96b7a..51d3a4689f20cace4b90ad8eaca2ec289dcc12af 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.out_proj/_s.pt
index 1790bf2d0bf76cab514a054b7899e69c232d5ece..cabf0e237bb896dc7f4694e20de04414104d5382 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.q_proj/_s.pt
index 7b6c95f5dad1e28b303c325c2875b1323a3e11cd..51e43b144e26179eb4675c3f0521c8f851960241 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.v_proj/_s.pt
index d7f6aa63ff59a52e6a8ec91f933255d6d2411a64..3c201b8e7747a042e88647f25c9f269706174780 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.15.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.fc1/_s.pt
index e234020910f661fb216f300c7673e7e981ccb316..70a16466acea37c44c8415f03b425ac58485a40e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.fc2/_s.pt
index c3c420777d0c90a689dff65bc9b20dbde6ddfaf4..f62eb448907ca0612463e79e4d625dac335a3950 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.k_proj/_s.pt
index badc7636c8915847b416bac8c08da86eafab6281..48ef287664a447e4c22d7570e3e4a35d2ba78619 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.out_proj/_s.pt
index 06b8c3b6a4d3c6bbf4d99888b6e3c6a95df3180c..62b05e0f3a15831265add0ed0662819b95c980cd 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.q_proj/_s.pt
index 18b30ebc2a4241633984325fea96bcb9ea4dd0b1..9077490da914d2a59774104a9f1dacdd44b8c418 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.v_proj/_s.pt
index 83927b44b386e733735fd1a5780e9250436307a6..642a3565d381a699bf2fbc0b1b41d842cc683b34 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.16.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.fc1/_s.pt
index 657a6a1203c30340a6efdf7df6dd4e1ebc330340..6c5da92cc7bba18cb78a73e51ed2491728822478 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.fc2/_s.pt
index 579980b231a377b50d913f9ef81b59752eaf74d8..d27f051e0a94dccce1e2c373541b04779994626e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.k_proj/_s.pt
index 7081a34e25c173cbe790ca599c4e33250c439a5f..ccf17c9a3f2ceb76a09b6ec6aa18303255602cd8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.out_proj/_s.pt
index 1162a22cac77386f43411fb1e8966573878d4060..3baf0f990fef9485fe2d5f84ab704b2802c2b01a 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.q_proj/_s.pt
index edd2c4842950cacf00d1c36c3d261e3479240862..931a51a3c6e394c43cfc710ba40214eaed0924fa 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.v_proj/_s.pt
index a9fa050a32cfdc427800314e5abbe3e993ec4296..698af00e557189a89076e416d521ccd72cfc1ec2 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.17.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.fc1/_s.pt
index 3f1b280f6f77f44bfa7f1d8b4b15c6e453d796c9..46e1e573f27aba18ae88c885b4b6583401dc4707 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.fc2/_s.pt
index 095e8d64f9c42ffe9e1ebc7ae89b58110bacd39f..0c3a58640676e4eab6ace83bc238342529d316ab 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.k_proj/_s.pt
index a7cf241b456386d499494ba0c7df708722750860..167416126b05c0471b9f81f8be95bb15bc321354 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.out_proj/_s.pt
index 5366c17d0f7cc9a39c23d99302cbcca08f96a9ec..c2ba8f662a408c009f4ea1c8aad22d0385d8a2a0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.q_proj/_s.pt
index 5a09179b8e68529886aef7f0ee14301524a6342e..15a6bb8932005abbfce85ebf1e5fbf87f00c248a 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.v_proj/_s.pt
index 34cc9248cd3b68eee10257a8f34ffe94318de9eb..5132d792b6077543e3b9645267b8bbe652fdabac 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.18.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.fc1/_s.pt
index fe1c3a32b4403c6110788920121c23c67e114cd2..6a79e16c394c9eba60de895be87c610c6d0abaf2 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.fc2/_s.pt
index 38a1a801b6cde9d9df7a5c84784c0eabf725f9a9..f7a3308828f6e687c6adf18c80e3f8c208762e09 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.k_proj/_s.pt
index 6948bd42d6cd8a70293060015c8ce8e9842d855e..fa0496e3ae404fad9d9bf6e53f28c6270b02aead 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.out_proj/_s.pt
index 811b08a3cde7683935fe85876156cbe6f34f18b4..22b6e629f909c0b3172726382f73beac6b653af6 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.q_proj/_s.pt
index cb5eb0cb9f08987be870732f2a44ac75a36aa22d..4ba96f70c09505a219dfe7a997366b1754903bd4 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.v_proj/_s.pt
index 3ea1b8700438aaeb408e6638efc53c2ffe236853..dc0f003707c5f1ac0db3858760b537f13d3bc8c4 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.19.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.fc1/_s.pt
index 9241d584355b42976743b793a22b6d301dc12279..d3ed434ab2d7bcd13e99ea77e529d4256bb568ce 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.fc2/_s.pt
index 945a4798a35aef72a7973b3ba12ee5ea8346623a..5c8118a7b100f780831998025e7359541662e36e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.k_proj/_s.pt
index 2f769afd17e6df0898c6ff4379ac5d7fdc5bfbe8..07af073dd186884ea6aba15cdb6427851f4203b0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.out_proj/_s.pt
index 0a04bd9ad7934651cea6c707a56e732c866eb2f7..5157a8bde8ff773dfee0a39f8a594d58b3f3c94b 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.q_proj/_s.pt
index c72bc6bc084ebcd00ec3a650c24acc03082dbb43..a9ac33b3720fbcca76d7c0f404db546b1dee93a0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.v_proj/_s.pt
index 024d28216af9d5d3cfa459081a161f13b184782e..9a248df5a8e1d7dd3e247652418f2ed5d8797a82 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.2.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.fc1/_s.pt
index ddaddde8ffe8eb4376c466ddf4ad1e59396a4e62..ce061da05ec5a4ac894f19fb0523932e8dac70d6 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.fc2/_s.pt
index 6217d13de62ebe1261e21299b070537a4c9de636..a92943d5e7142d2579a32d8a5465c9ac59da573e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.k_proj/_s.pt
index a84de15fadefb03f45f038b949c6a283d84eaf67..2a94afe4fd8ca17c988130ef6d2dd95abd6aa8b3 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.out_proj/_s.pt
index 0c5715800069cd7aa60f21fb95148bb4e2b07a21..f8a9f602dd74f8986174e8b80c47fbfba7910bf0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.q_proj/_s.pt
index f24507cf70be53635561fea53f97aaa36f273bfc..b6569521c1d515b0a07ef5ae0634b20036e3b416 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.v_proj/_s.pt
index 1a99ee429bc975ffbf6c116c26dbb8bc66cadcda..541ac50a14288e87330ffd34a5c90cac2b6c457c 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.20.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.fc1/_s.pt
index d5b77ffe972adaef9412241d82bdec6573faecde..000adbea3bd0abd7ae6e824a168ca95acf20f290 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.fc2/_s.pt
index 6ad47b78bf8cb9d826a83deff076d9a7aca12f48..bb6ba5e869186d6161b71ec5c0262119fe5bd6b3 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.k_proj/_s.pt
index 43449aa42f7f233a5236ef010740681077365754..d21473e14e2ccbb3f3757d0dc34406fce699ee5a 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.out_proj/_s.pt
index fed33c7d87ef9f33b0aecf4bd29d3eea5af8dc5d..45e29d995bc285cd74ceba7c65b1cc8ebf01e2a6 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.q_proj/_s.pt
index e8d4f0ea47027e7f3261f4a22bddeeb1d69d94d8..136e2e97fbe43fba1439426bcd886597b6e06430 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.v_proj/_s.pt
index c25237e2017c5d36e9e6692dacc79b8977b7b6c3..d59ab96930121543554c8c5012133be2970356be 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.21.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.fc1/_s.pt
index cc22602b60b7230bbca6519d4744fdf6e586c95b..d9826615e4436c39b9f86d3b8b09ede009ffde4d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.fc2/_s.pt
index 30783a58e6875543ca1b9a0ed3a1a8d0228c9512..98618f97ac0328323cb4074c0ebd2ef91bf4bb30 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.k_proj/_s.pt
index d8b55a1d2d90518fd3947c400ee7357a409d3f40..1eb9e9746636203a1a03c4294c9a8819149c42bb 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.out_proj/_s.pt
index 13f102d2b9e47141654da8500e3722e1f4fa1c76..6cd409fc6f19cedbcc6619b1c3798e874dc89b1a 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.q_proj/_s.pt
index 312e831759c52e239a4b01d44dfb84621ee4cb43..3622d448fbaa27f1c99d7dc069b673c3f16e33c3 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.v_proj/_s.pt
index b719b2db38c256e080a203841403d6918b4c5543..455e04cb73a2f9e82a15fe39bc31a046943ab74b 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.22.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.fc1/_s.pt
index 6ba2d2466621f034efe62dabd96ceae217253b0f..b4d861903ff7fc97428980997d115223194d1dec 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.fc2/_s.pt
index 333afad4e6f22657a2db3ee0fffe3bfb16cb3c1d..cf80a7f733a398d053706c43c97536b4818f4f65 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.k_proj/_s.pt
index 5c095cd0d33721d31973da7ad479460ed263c5c2..f1207bed30d056e4f3f7c1487688ad47b430d59a 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.out_proj/_s.pt
index bfecd31afd74d076f46bc270be8cf733e95809c9..2a075ddbe9aa471507fe835428cdf5f70f4cb71b 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.q_proj/_s.pt
index f94ee4d15810be16f303d1aa866e50027da1f479..e44fe946039212adcf36a5c5316c7d85de41f477 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.v_proj/_s.pt
index c444e031cbd8c3b603bee5d7fdd7ff8184a076c6..24c9d886d8b1be1bc5b6b696e6e35aca9ffa64fa 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.23.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.fc1/_s.pt
index 896b345723207ee64286662c7b806bd297fce059..b843d3484661f0373b8657e165f882bbc2fb7173 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.fc2/_s.pt
index b929a6d6c89cbbed2b028acc24bb8f15e123c422..f8ba8102e842fd9f5c8d4425dcbf0273202f6230 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.k_proj/_s.pt
index 078fafee3bf07cc65c0cbffbad3f6faa9af92d01..0c431290b35188c81e3c1b48fad85ead3461448c 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.out_proj/_s.pt
index aebd74a29e3eb411de2d2699d0e10045be012dd5..67fd3510bff3928a2d70cb5172fd8e3a2f411dde 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.q_proj/_s.pt
index fa35a9e044ed7a6c3067474aa757245d3b1fee55..9bcc225d68ee15c7b99c458916a9e108918d6e54 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.v_proj/_s.pt
index 7382dd384af6748c716173261d649b4a5ad5e609..800697f995a418a08742fe7b42d965d3dbbcee9e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.3.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.fc1/_s.pt
index 4832969536158ac69823dc854646160b42acebe9..953ac73483215ed02c628257dcbef4a52abb234f 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.fc2/_s.pt
index a828870be19864fc4177fd5e1a1b8f2efe72414c..e63f3d01cc899f5a736f8c46be0ef88cd5ea0894 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.k_proj/_s.pt
index 8f269f66c66a9f37de1c903d7b202cc81e2bd32c..83e850b203efb744ef7201f3dc85fef3ef49ac33 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.out_proj/_s.pt
index d7a0b1ed5215675f4682ef404c1e318736d313ca..cf65bbe2e390f9a265996f6424b9dbd2c7ab80e3 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.q_proj/_s.pt
index 9210a4df922fc9061d37e7b10a29f83444c7e087..7a121c73b9001fc4ec4c3ecba7ff984b404ef0e3 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.v_proj/_s.pt
index ef8eb15c49ba00f1deb46a939e2f53961b32889a..947575b382498697805be121ff56233710fd5a3e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.4.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.fc1/_s.pt
index 324d98b918f219d9c2d6acd144c3acbdadbe1818..a1edb80b727f649fa49515f351cd220b63dc4e9d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.fc2/_s.pt
index c3a9fc61eb30873b6333ba836174cc3d280e0a1b..c3b39144bd09497218e648f3a1ddcb06b33d3af1 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.k_proj/_s.pt
index 6a2b3e3831a6f2cbad26bdb4fe3312725cd6666b..4d94695c2739bfe7be27d73a96dc9550b16cd135 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.out_proj/_s.pt
index 3bcb0b1804dd42bf1e60207046eac8612898426b..29537b5e759866cab56712d485ce9539e49b08c2 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.q_proj/_s.pt
index ce580953abb14a5e594b173f9ee858010aecff38..19867f0e892c463a6ec915c05caa2f8a4ece9dc3 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.v_proj/_s.pt
index 7dfe25251316281454b706c9fcb89bc416c88d03..5ed28f4e12cca0d853cc3a0980d15271e34cddb7 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.5.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.fc1/_s.pt
index 72aa864df88cb3df8123feef8e0f61cfab624208..5023dfebc557e30359d45b0467c97d3188db2153 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.fc2/_s.pt
index a2d5a70e438bcc31e094e5423b2fba6d02dd7700..80e0ecbfefdc851e4ecffe926e1b0ff58b2f92da 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.k_proj/_s.pt
index f1278bb34a8b1fd24c203eca65caff88b6807539..c5da970674d43c554d924a3398953a883c380043 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.out_proj/_s.pt
index 88728d1b054b5c45ef29f81bdb649aaaf2d31e1a..d71185a55dd7dbe76f662b2b4451e633c8d461e8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.q_proj/_s.pt
index ab5098f4fb5b8f8f6d46411a7f5b887e3b1c5c0e..db21b365cdd84ed9ff78dc0e5cd2fc6e7f22a4c9 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.v_proj/_s.pt
index 8c513c4f082a933d43bcd14ce3c239fa1796f7dd..6cfd7ad5509498701a07202357cbfe5b3d27c30b 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.6.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.fc1/_s.pt
index 5b8d8b94f79e58a0f1352f75eae06ac0198a41f3..67957da68b944b1449c028e41a82612a435a483d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.fc2/_s.pt
index 8e9e85ed80b3b91a350ee87fec3bffe40b7acc2a..04094c350897f46a0683af291687b66331034c35 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.k_proj/_s.pt
index 0f75b45a190afd55ab4e2c8ba0c6b5953beff99d..f2168feeaaf3ff6e64d32a29e4ccd409d6a1bcd1 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.out_proj/_s.pt
index b21ad408c8ff1c688b0c073d7b4d4cd2dbba6397..3edde7de7935cc2f309b4838c88172116a80a742 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.q_proj/_s.pt
index 52ae03b1af757aaade571e266d2fc2626a8c766b..eb7ea19728e3b5d6786a5d54a45c17b8bb3533f1 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.v_proj/_s.pt
index 881e5deaaa0dfdb35f17c91b4289b25a1b660a3e..6f3378c1bdfc458554c8e902d4f4b5e265e0f237 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.7.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.fc1/_s.pt
index 2144fbe43ee3019b38881d5a78ba27b28a961018..c7ad96e5d467869962442c65aea27c0936ca2f8b 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.fc2/_s.pt
index e312b13a63db6abc1bc7e714fff49fae4dc2e19b..8a938a1c28315a5d687a2c127800a3a32e528bdd 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.k_proj/_s.pt
index aa997bfef05ffc04fb582c96d203644c90d29f32..57cd0538406669f64a5c5127d618d12259cc85eb 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.out_proj/_s.pt
index 50b924e92c40a2f6b9e647b7f8c4f187780b11f3..933f1193002fe39f86280e5ef6cc44f9dfd15ed8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.q_proj/_s.pt
index f567a17533291593c7233b7c18bfdaff68ecb89f..5edb51c4656314411c6339e615fa0ba843cb1d1d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.v_proj/_s.pt
index 63e096575723c39356ee1dddaee35d5f9dc5867f..45178be59e04d0b92b3f22e757c4f303559eadca 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.8.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.fc1/_0.pt
index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.fc1/_s.pt
index 407456684ccfe9ecc34442970340a9eeac462a77..8c8027e8fdfc56dd7fb28b9d242e8d5f76f08ade 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.fc2/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.fc2/_s.pt
index 6ffc15f0fda39e6ce9760fd9a22ee01a99f1f68e..d05d260de8362474e88b12f9ef65dfb94496666a 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.k_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.k_proj/_s.pt
index 04473acb4944c5867fb989df37fe50d08c8babd0..b1a166e1f2d680a5cb67cdadcc098af5f2d469b3 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.out_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.out_proj/_s.pt
index 36e7549deeab7203b2fe1a504f53903562f563f4..7fcc479c52d1af8dcdd5d9908a3c5f74cf8cac76 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.q_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.q_proj/_s.pt
index 831cece5a0485be8d99b70038e57b3af8a445b7f..79e45137ff97021849686ab8192240e16069a567 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.v_proj/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.v_proj/_s.pt
index 7bd5ffa19cfe862ede3bceb2c838c5636a0cba24..2de254f0f2b6e3ae00f66459372c0f3caa0fbfb4 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.layers.9.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.project_in/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.project_in/_0.pt
index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.project_in/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.project_in/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.project_in/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.project_in/_s.pt
index b9cf16136cb65408e93bcc467090107d707922d7..fe355ad5f36abad07cd57298a55c80d8caf9a140 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.project_in/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.project_in/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.project_out/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.project_out/_0.pt
index 7dafd7dbb653cbd2daae7b5c32e1db4207127f70..6f3ba354ec876f170cb0943973ce2085a27e0750 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.project_out/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.project_out/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.project_out/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.project_out/_s.pt
index 7d435ce321a2c3e069d82a1e75f1acae96b446c1..065e51d4a779878dbcf2c69ef521546450eab7ba 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.project_out/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/init/model.decoder.project_out/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/lm_head/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/lm_head/_0.pt
index 94c24cfbdf8057747ee28dc87bc29b1c055a556a..b0e457ac4104d2409afec35b4ba77d2ac253a59b 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/lm_head/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/lm_head/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/lm_head/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/lm_head/_s.pt
index 87b3ca638216d927bb95bab955bba14e517a76eb..3637e5a19d0de81cd70a1d58b011867393d4b87d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/lm_head/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/lm_head/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.fc1/_0.pt
index 81101eff3ba103ab285012c7c458fee4ba7ecfe6..5f66d775634cbb719011546fe895597f11aaa4e6 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.fc1/_s.pt
index f9db30fc1709ced7ace245e99f0901ec87565eb8..db019812f90220587f44d2b3b108b3bc685460e0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.fc2/_0.pt
index d2571c5b56a0ce622ef4b13f4efd8c51b9781595..b71d27792e33e50d1960bd139a15556a3ae148c6 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.fc2/_s.pt
index 3191ff99447fedeee4b54682252e3362c16d52e3..a3f4c81e45401b5fb9a3f206e95efd0610c3bd12 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.k_proj/_0.pt
index 5538e0c13e549a474c52e176ad9df5c23a4a5036..a4acf8fdd9ba8f3b077275d9dcc025429281003a 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.k_proj/_s.pt
index 749b4ad5aa8ec3aa21fba582c87a25192f136d54..e2eb4ad7cfdaf0addf6063317696aca888fa0fc2 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.out_proj/_0.pt
index d737086b899b3e3e6fb0e8c0abb388b5f2777c6b..0612b9c3cf729c385fbc8b70678f7de666aaff63 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.out_proj/_s.pt
index 3fbe74f8bc2a3614593091baf8856dd087a5956f..482a539bd64a9b26ec6715c57733f778aeccbdbe 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.q_proj/_0.pt
index c8d983e639169208b8d4ad0a1affa44d139aaf54..4430586b0ed68e049bf89978c38eff6d0bca45de 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.q_proj/_s.pt
index 847d28aad68d19acc4f41c2da8938cf9d45ce521..9b090acd72437733fe69199c06576281f89eeb56 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.v_proj/_0.pt
index 6f7a3abf33220467fa2386cdda404176854d5c4b..1b56098aa5da4d978bea064e574fefb28e7898e9 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.v_proj/_s.pt
index d838ae16282c7bc7d2db46214dcdcc518f68c17f..18cc836cf89e510f0f4f09d12a8400a345aab32c 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.0.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.fc1/_0.pt
index 83f550b6e6347d83fd43f035c1a1fd590c920ca9..a6f37e8d6517df8819689882602f65f200564b66 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.fc1/_s.pt
index f5d4977b83805551251e690d2288eafdb8af1068..0f85d6590eddb4353695a122f7360cb792b758f3 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.fc2/_0.pt
index b531684e33d4403853f83587f8c728bab2c44dbc..4a8707975a9045d926f8eff4a30be40f15807924 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.fc2/_s.pt
index 65ad181c21aeabdc9a21da35dbf86bf52efdf096..c452704ef64690b346232c5f71f8de8ba12f7611 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.k_proj/_0.pt
index 9f02193c6eec54813b7fec7e0edfe465345f1746..8d2325da67a538354d1727e19a78bbcd09302e0c 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.k_proj/_s.pt
index 99411a26069e2d811a3fa1fc7256a7d9d3aedd48..e67ec470d94bf14d7b1af1a3150d8233217f688e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.out_proj/_0.pt
index b5f7f3061b8ba16152cab53c01870ed6022788ec..ebbf65bf818e4871ff23454150459f7aefb39061 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.out_proj/_s.pt
index 1bde0b046ddc47c47565f351c933b53fc6b74fb2..6093965b38cc8d80ad500165014c6d9ed18d6ca8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.q_proj/_0.pt
index f6a26ab04639f962d841b551ba52f5c04e99e8cd..3d42286133a601838f787d82cbff4882b265d6b0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.q_proj/_s.pt
index b840750cbcad3ee2aafda01b958e1e92dad7d71d..cf8a6f90fe33e16e56cf0c4ce53a6a8f5c57ca59 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.v_proj/_0.pt
index 6587f41d339fba7eef618088eba8a5ad55a0ad91..e4d4d1864bc693400fd464e0af4ac67e915c6270 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.v_proj/_s.pt
index 5847c26d876266a99043dffcb52a7538e91e65ec..165da07033d591c94cdbcb44a8dd6d7dfd5b9fd7 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.1.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.fc1/_0.pt
index faf59eda0f8a08f56e23cd394ec7a70d839f1fc5..894af98eeeb47e309b2c29e375a02687304c228d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.fc1/_s.pt
index 0a489947ac46dcc752543d46dde19b80d04cd667..ef71ce90b160412a37bf43997dcfa35e428dbc42 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.fc2/_0.pt
index 842dc9b0ead57226549a141865d89e28b198e275..c1c8e0a9ef5ae177a8547cf76dad8fcbe21729d5 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.fc2/_s.pt
index 4e24c5785169d3535f0545b165c2317ece1fce63..7a21375f41380b792485a17b9003f13bfe304f0c 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.k_proj/_0.pt
index 2c1e0f322ce7cc14d692a59ce5e59a3b7e2853a4..21df7f4c59961123529a49388a8547267060bfc5 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.k_proj/_s.pt
index 7e5b4c43903b78f044696d10c882c7c8e7a4ac9d..eca0c001264e7db3ad73763fad42d7c280fad984 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.out_proj/_0.pt
index 7c63cfaa82bd97053d735dc0f0f30c82b5fc3eb6..5435a38135651db75b0e9f8c55b6c68e7f195be9 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.out_proj/_s.pt
index 7d9071388e9a7e3ee7442fad95edba38f3925e5d..9512ee202b2a1c73a537eb37700658d2c47218fc 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.q_proj/_0.pt
index a531caf41b7c6632663c1e4034c508a92a228f6c..41a5d0a7664c8ac499d8a35ed58d76723aa340b9 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.q_proj/_s.pt
index 592b459f92a115e88f0911a19b38b304099101d3..0d882d7e84ca9f5545403febfe3e5e7885ea217b 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.v_proj/_0.pt
index f82cbc8e76345b0e48f2b748e6231220fbcc851f..e6d293e2027c57973b22b8c7c5569a0b4f729246 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.v_proj/_s.pt
index 4d1e74c0ec1106cd707befbdfb2e388dc14f8244..3bc63452154da6b6888904ad7294feb8d5b5f57f 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.10.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.fc1/_0.pt
index f5cdbb19ac115d2799ed337d1565a61d2d5de831..10ba56a5a556df923a6153d96a86e72b92543d54 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.fc1/_s.pt
index d5f089373b9808c2d41efbbe4b0d6f4f28333bf1..10dbccb8970bd4dc7ae93408a85402e72e6ca5ea 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.fc2/_0.pt
index a054810d89fbc6b34c12a8fdecc0a4e426efa002..365a30fa5d3cca189a70448c3d8d09ceb402f3ff 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.fc2/_s.pt
index a9e441b698e3fbdf55bf75014247090a957f87b1..b1f929346912226a69c676e8c6a3850db09aa53f 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.k_proj/_0.pt
index 1f0c389d29a09bc5f0678436e964eb5161bef07a..63a162edc2c82769aa0060b624a691f70d42b62a 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.k_proj/_s.pt
index 7716cdcbd9b9d302495c9f70d09817fb66bdbee8..dcf6be74a404bfa825f6d66e6ba9b67d6b1eb804 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.out_proj/_0.pt
index 4dad60396f5ae1206873c1a22226a0fbd545d561..0ba9acda0edb12084a21687035b53cf3c98d21b9 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.out_proj/_s.pt
index 4ba8872083de7fa91ff9e8f275740aca67f88b92..ab8bacd4bb9c021a7885ca72212447780ad21d43 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.q_proj/_0.pt
index 5e42d5fa6c6418bc7b5bc31555695969efa8e38a..4915293f6bc42860482c5f1131f5aea16e132460 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.q_proj/_s.pt
index b17fa02a3daf11813549bbf9bef18368825f0416..6728767139681f199075b2d2e5beaf805e7df6f6 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.v_proj/_0.pt
index 0375f9fc854d7b966739034c04ded195957c8380..e7a7b90167b623dcc52a3c806e38092b8ac9ed4d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.v_proj/_s.pt
index 507cff132a72c83f9f6b22103e086b69aa7fccc9..a32461f30833498cd258da2a8f7ac130b10a2676 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.11.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.fc1/_0.pt
index 05d5362b6b8d30e75437d63c8295d4e02b9e7ca3..18411263c20f2315018217122f2a07c55198dadd 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.fc1/_s.pt
index 8f9a6161821299d7ece5e6f791f84e8cebe1cc6c..4d9f2cfa390c82269720c04152d36615bc7e0e23 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.fc2/_0.pt
index b14124f060b388a6aedcbcb86718091d6bf09d2c..433176e5103af234028a835f58973390f9c8d497 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.fc2/_s.pt
index 1bd1a91cc7f94852212b2ab9aef4e1f8b09e8b32..d71deef59e9aca1756087969152e56c968d5a795 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.k_proj/_0.pt
index b24167466387cfdb45e93978050d0776ca0ff8e9..e93095d57c3fc486780b9e69ada0c823ae97f5ef 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.k_proj/_s.pt
index b1b06151b712f1a1b3dd57eb5761721b1e41281e..af6ba5891136a17e2c2472cacb809385d9769e53 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.out_proj/_0.pt
index ac405c7b69044052ec01b3490a36d7458e2ad5b2..8d40e317bf5c18534a3a380b6e58af4bb1a7d9e6 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.out_proj/_s.pt
index 886c25175b2e12b736fe434ed8f6feabd38cccae..dc89d6dd4306af35f33381e047191bc9c0c411e0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.q_proj/_0.pt
index 4d0545ea30db7f1edceb6e2873edb5d45f6b94a0..7a0a574cd714df827f39d9e5f227e061794bb880 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.q_proj/_s.pt
index da91b6eed220d5f320efafc57e57e490d1935467..e35bb7d6ce55becdeab77d4b1c0f595e88ef9b37 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.v_proj/_0.pt
index 092689208877efdbdbb09178550d756d26b4af41..3d6f2a95b53297bb97f97e9f0abf4c11ab0a1941 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.v_proj/_s.pt
index 04de1d731b45e1f05106df2179dd7b67f2c95ffe..6b8322b51fdc6cb537007785fe30a8b3f6cae913 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.12.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.fc1/_0.pt
index f5b4cec7969f150bbd4f62324b3cde15a9f897b3..759ee07b5db0412ccb5408468014ce254b3042e8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.fc1/_s.pt
index 9e275b91b3cf7a25f1722ced9c37b5ce594f7e86..fa3ee232446aadf416463310f2c49822559cbe3d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.fc2/_0.pt
index b6069eb549f5d86517f36952a4774896d5605f1c..3b1559f3b19246643021dec41cb350c2018944aa 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.fc2/_s.pt
index 7bc109858ddbe54b4962d7bedf70bf2bad53d0db..1fbaeef6c44f863879cf156edb59f570ec0830c2 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.k_proj/_0.pt
index ff01a5627bece6a39326c434d7d2c5b79ef683fb..8a3200a305233c0238011b5cebdf8dfe07ebe6bc 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.k_proj/_s.pt
index 9a8511d9c888819eb61c1e14654897e67da963d8..a388a30556e9bf8349e67a8aaca21264b85e2fa1 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.out_proj/_0.pt
index 4ce7aa43d2e6c8348e0da0f7842d504618b3fd57..33e0f2618aea496e05ba15d96996690586220bd2 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.out_proj/_s.pt
index 421264c28ab5ffec24d4c77a2b77535878a62edb..8cc64db170e04677b43b5b80e08298fc38705d1d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.q_proj/_0.pt
index 7c002491a4b5102dfab6ea846d1c36b2f52695bd..e6c4cff74632acaf69fb08f82a49e7a341ef88d0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.q_proj/_s.pt
index 28be3653e10f30854a3e3a8f7ed6666dd9e2194b..57e0396b227cc9508238667dbb026af04bcf66e0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.v_proj/_0.pt
index 5afa487f0aec4020f1ed6caef13192c2ebcb7f96..154c77d220d56bf869cf79bd8370cee9edc37836 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.v_proj/_s.pt
index 13ab2603876647a6e7d41c22637ab98e0769be03..a79484214d1731ee07725b68b4186eae953b28df 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.13.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.fc1/_0.pt
index 8100e2208d86c49bb33b36975fc7925fc414a775..57968c57817cea217d83ac060bbacc7791d3cfba 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.fc1/_s.pt
index ed22d45eb8403ff12adf193dc99ecddbd7f1c02d..c95cd67355df87727bb7918fa3185b906c21cbdf 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.fc2/_0.pt
index 90fb76f19bbf211b25dd15607c83ab77877fb07e..54b7238d4b70dd0b3e9efe763b5a7f90b2ae6ad3 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.fc2/_s.pt
index 403def9a47f087906b9f2b520d768df7bf02452a..2ff190d5314b80796daa36d3ea8c668449078ab3 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.k_proj/_0.pt
index a71fe0b0e8dee73387e8856c919c78b8ffd07658..1b828153db594d68fb9dbf781d5f4e17e37e81cd 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.k_proj/_s.pt
index f29dba08632a5653d177002fe28ae83af3d9e50f..86af8b839258943630075f8ff99e1dcddb12d486 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.out_proj/_0.pt
index ea041a38313f8f520fd83ddcbc74fa05d961a8d4..78f03a4526f4722b4645be4e308456c93e48ae6d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.out_proj/_s.pt
index 522709adae858aef1e3c870a8b4077ccfcdc61ba..a3abaf643ef2b5df1d1e9e840d927c5b20760434 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.q_proj/_0.pt
index 6fa67742caf28ee5715aba19bf0d7ddb963ebafc..9eb72c81cb3a82a0b7797588ed02065e0515a4cd 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.q_proj/_s.pt
index 6b2b5e0844540a574e3fe9d939234facd962a1dd..3902e8d180d21581c9a29a266cc94a03316ab355 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.v_proj/_0.pt
index 45d604eefb874bdbbba73698e144657a16b7b6ec..0e4deb74fbb170b21f594295091376f4cb751cbb 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.v_proj/_s.pt
index f8b9b9a2424571b45009ca8148033f336b914641..47596465f785895ac43423c712c61c8f629f961b 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.14.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.fc1/_0.pt
index d09e3eb42f61766264e0b5757f532fe166b56c93..95bbab07ba474b0d517dfa7241eb7d33cd148aba 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.fc1/_s.pt
index 1c9fa0eedf48d410a254da98a1034381be0e3847..e1ebf1c9cb614034a34298f7d5e1c69e0083b258 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.fc2/_0.pt
index c9d34e8bb6589e9498af08b9b5abe5777bf26218..2af4b0c919b622d5af5da7d23c92071c209076ef 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.fc2/_s.pt
index 469a309257cb2242c91311857df919c282ec27ed..b5701cd7325e5bfda738ae9e04dc5342fff55324 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.k_proj/_0.pt
index 6680550b91cd9496e92e7fb4a84d622417c7607c..c492026f229e57f5bff6f7f1e865f862a0ca7b2e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.k_proj/_s.pt
index 107b512f3ab338d72c09931b8da7a5583c6c4b7e..f2f8b1852107b07a1fdb5a8ba08d6e45f7b679bb 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.out_proj/_0.pt
index cc9e6b5f1753e1b3dcc0887bbade2837786c8992..9b056fc1da8a180fc5cf2b1bc83f32a828b7d563 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.out_proj/_s.pt
index fd48459126cc2c86719d2622097cf970eb61935f..d6de71c33988280a9a93db2d33bbfc84102ed99e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.q_proj/_0.pt
index bfa91eb5082c5da0e5888b900c9df7575fa055b6..4b1418489040242e4972278b539dade92ae132b6 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.q_proj/_s.pt
index cd785887124ac7579c74b3aed13db52d52b4260d..23c0f1c7515d0d94f26f44c5488be7ef601fabb8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.v_proj/_0.pt
index 9707db9fd77dd9cc921c3b3beb9d7150b5baec31..7d5f0dd7fdb43222140434bec188703778dfc666 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.v_proj/_s.pt
index a2df1a1763ee47be1056ae7297127ac7e2f825c8..b36cd9ca7c051a5f1dec488071df03fa92b93013 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.15.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.fc1/_0.pt
index 41687935609d2ee2603d7d97ecab4f61b2bf841a..6c77a0648e3abf08f5a849ac77fd71863049675d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.fc1/_s.pt
index 24c6237a8857992cef6936885f25df55f798af77..3a7f3c6996d2a1472b719a398af787f737aa66f5 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.fc2/_0.pt
index 5f9870e26c3b846e983b77b6a96e1d0d3d55257c..3b7525751599a45b55c59a2f9bb3f0194a735f06 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.fc2/_s.pt
index b9fba271fc466351fc6f04a8cc76cc302188ee00..495f1b58002d605a730fcde397bd58f3a152d3d7 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.k_proj/_0.pt
index a4b84075d3e3e1ec956a7c601b3ca0441b4084f5..3474218d6fd69a33d7210ffe3c23986f3c055343 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.k_proj/_s.pt
index 3e3de8d8e28b27381f03c6cbca77ed5d8f6e29db..f0d589eda3af6803b2f4c746932cc3ac6c1a1d24 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.out_proj/_0.pt
index 7f4425bbf1e50883333bf1379b38eeca1175541f..a485bed0ac6b3121f4f856df7a9cc8d9004c38b7 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.out_proj/_s.pt
index 24ae2ef634ee1e7d54852258a67c2ee289268a11..7204f3dcccfd7198bbb344fede38b1afa15a20ae 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.q_proj/_0.pt
index e6fab70e0c042369562962bfea8372a9037f0481..821553b456e3470b4cdfa39e9cde6e3dedd9a7d5 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.q_proj/_s.pt
index ac3cfb84595b9a848caac00e821ce76c8d713257..507933b74ffae1fbea9d807060625a7ba101ff36 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.v_proj/_0.pt
index ffe699d0fd35557e2de3f1036cb34a9083bf2975..8f03c19f76baf87906dac20c7adeae551faaaea9 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.v_proj/_s.pt
index aed115ba5b4aadf7cc65fe8f5789072780813ee7..9834a6368fca5e2234af22267ae847bfbccdb28f 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.16.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.fc1/_0.pt
index c8d0e373972dbde0ff6ced231e68124333dfb2c7..270c8db47275e0e4459fca74b0fbd448009da0f2 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.fc1/_s.pt
index e2544f2df3e44e41f7e6c368852fa5a4b1a89511..3fdbb17e1fea10de2b58f26f358bb0a14fa89b1e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.fc2/_0.pt
index 01ef686fefd4f88bf9d9b65f7d12f70a626d45b0..2c5fdfb169ba7ac06eb7657767b1a3d3760ada07 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.fc2/_s.pt
index a9cc863cb699a87e6f737a6b947d3a91676c62c3..e99eebdd5b2c263c7b0a23b7045e3478c7706cf8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.k_proj/_0.pt
index 1624aedc5e06bb329886dd17344287ad4c9c429e..66d2782f8eda3ec101b83490667f377b91c70653 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.k_proj/_s.pt
index c2d5480a9eac48cfd409df8cc2bb938b986986e6..99d39b4365be83d2165261383460da96715cb578 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.out_proj/_0.pt
index 5ab65b5ee1d2cf8bd25360e2ed1d910a5bda51d3..a6c2fae8a1059c0f43a519f055f4bc1a4c7aa93e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.out_proj/_s.pt
index cf3648e5290b1f2f5d722d369f911b1d089ea06a..b2557642b0a54569aadd733860bccc70f31ebbde 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.q_proj/_0.pt
index 5f601d53d4b03622de1610931a3f8b70c9dfd40c..013ae48a86af52d9341c636aca0f747c81917011 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.q_proj/_s.pt
index f4b6a2cc9d858fd659032820b52fbaa08c3d1e4b..601142909420a16d483eb2b516f478f35c8ee8bc 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.v_proj/_0.pt
index 81473d6f2d930a2408d22d69512a794c1169f95a..22ef0e08e5d14427a6a9143ad14db9f07de3e1a1 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.v_proj/_s.pt
index e0681ee471f41d623063c5224367dc2344bb9760..63a932eee6034332d0ce998f2f842fc569691f31 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.17.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.fc1/_0.pt
index d37b83ce878bc2d48e897611f71c7cb64e6c02e0..b8471c355bcaee049dc46b2e98a0cfd4df05876f 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.fc1/_s.pt
index 746f64cd2053a69a53560a8c8318c8168d2c5954..ebe52dcb0adf6953618f3cc6f867a918dab386c8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.fc2/_0.pt
index 2adb54cce50561acdf65985faa89af8114cd3b8a..d8143edbd8023753d7c8db7b01bfdaf1ecf82b52 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.fc2/_s.pt
index 51fb8dcbda549b90d6352d6dbc48ff2833557ce3..52938f9504539c94912fb886c81de0f02441f523 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.k_proj/_0.pt
index 14e16eb0c15c5a3d4376b4d22ebcebaa83431f9b..5618a42874b23678bc0741f56ab28bcbfa6a9a14 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.k_proj/_s.pt
index d8a6d51765d1fd0ee11c5fd785485ce5cf2074c5..59530c3fc170ef3a835435daad0ad48714b975fd 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.out_proj/_0.pt
index 2124342b3cdde77b4d34451a28679bdfe79770eb..90a0e3a19b512191dd082364486cde7d497336d2 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.out_proj/_s.pt
index 954211388ac473c3b9f17f37a6ce785943b02e82..1e951c82cba589ef55a7c203f7afdd0bf23439f8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.q_proj/_0.pt
index 55ce4ae83dafe531407f258b53cd71b062c0a1e9..20b7c7c22874db5d36cf4dca1ec9fb87c11a53fb 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.q_proj/_s.pt
index 35f58659735cf5e3046564547f4cd40598adf71c..f7c9b1c6e7961f42afbf4ced3922b59a8ff9be72 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.v_proj/_0.pt
index 80d0c281bc3df94d9a5ba2f3ca3b31d7d37b67ec..73bc14c8ea51e56352fbc90760f18d643dc216a5 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.v_proj/_s.pt
index 006864ee204f03100da08418b917538cddf73d8e..8e1fa2a3e0710c4c08ba7f47087a1bbb38e4d79e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.18.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.fc1/_0.pt
index dcab8883ec9b1bf7012f4a31a720f6d22215c207..16940c159949ffd71f82a44cc3b2863010ca9105 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.fc1/_s.pt
index 9a407bfd78c57cfa2411524ec3f16c28acc26a1d..e3f157dd2b788743a88615b30bdb03d6ad0dea7e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.fc2/_0.pt
index b5bcc1cb7b6b99c8fbfd092ec8a5ae0aac90b1e4..5628911ea6b43b18870dcf7bf9409765a1ba4ac8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.fc2/_s.pt
index a6219baf1655d74725374dc31eab25fdab4f9765..7cee76738dd7e5fb28699b6e4284891e18c9aef8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.k_proj/_0.pt
index e7dd3335a8e7cdd86cc750918a12592f86cb8783..29e33527e49d941674bc115951f91cdf26b4ffa8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.k_proj/_s.pt
index 35a8a2c7031b6ece43e77ddafce1bcf2959e7498..b097ad74857fcba61dc00122a3a30931001353e8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.out_proj/_0.pt
index 421fadf7941ba1aecd423ac8bcf171260186283a..f2cc8c7b4847cb7e4d090f59279ab945c3df82fc 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.out_proj/_s.pt
index b6424f817caad3310517122aa38f5d91f33cc372..54171d696da27f5a3bc1b85ad5bec83d5b6e446b 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.q_proj/_0.pt
index aab4394653ac91bbf858a596179730c347d31706..802a4eaaf88976e9366593f3b833c8e96ac8c723 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.q_proj/_s.pt
index 7014471f41b4befd9ba0871a0e03b2f16d04c6cf..0babafb8e019a3bb9961cdd57ed998b5a772f4b9 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.v_proj/_0.pt
index afc3b3b9613c7384a8b1be0d197ac768f7d29b96..d62cf82d208db3c4d663702801c5b7ec7b4a449d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.v_proj/_s.pt
index 70adfff12cfad716b31b586d27c578bc2d216d37..dfe73e67338c954dc9d71182b49bb99a6088218e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.19.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.fc1/_0.pt
index d7fdc74efa0ca6152a4506dd232fc1462f3439d7..cfc2f4e21ae8a2dc89a6aa39018c4b3b697f754d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.fc1/_s.pt
index 36d6e3d41888d9eb5d66f81aa083f0d653c3f7ea..33a19e0e4a4b8a692adb47b727a321911c47643d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.fc2/_0.pt
index 4e24185c948e0a510d8ef36aa0ed6ea33e00ee77..e0336cc97552c937e86278f5afb1e4a49575755b 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.fc2/_s.pt
index 6f834df6926f8237293cb4927fb42b5757b5b43b..67ee404f0e3af9ae1b74cc1ab587a83704c45a28 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.k_proj/_0.pt
index ffadc1a803b63ea31c24f544cbbe987838f14997..0a7688062d6214b5601d199259cfe6bbd49892d8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.k_proj/_s.pt
index 39deb10d5abee9db3965fe16e47743a64b28195f..734763bc6297b782e6525dfce0f8ed9dbecaed29 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.out_proj/_0.pt
index 56174bae86bb97c73f493defba30ac18238ef477..9fa23aa9d9bea822e397c0e818498b8f80b6d5a2 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.out_proj/_s.pt
index 3a637f4ae9bd87b052f4fe7d01c4bee6d86917eb..5820b55e1e71b23683a2650e179c4b8feea8c32d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.q_proj/_0.pt
index 8cf34fd8964a4ff6639e9f55c9e1eb99798ecd4d..89101cd4d4f0b97f451cfaa673abebbf234d37d1 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.q_proj/_s.pt
index 66e98618cdfc98d1387ae0fa88d0a9e7cfecd2c7..e9414db2e8975c3f40eb25ab425bf87014a91cb0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.v_proj/_0.pt
index 274de50efc71c2fc2739f55d1af92d4e02fa7c67..1e670020f4850e53802c30af54dd21946586a1c0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.v_proj/_s.pt
index a418773942d0e7144b6ee8c333476e90159a7fbf..b9ac2b5cb233e45b20f739c82f36fad48cfea45f 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.2.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.fc1/_0.pt
index c120aff27cbaeae948d20423c688fc4051889df1..01d63bbe2dfb50d96a692143a7a9f626fa999752 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.fc1/_s.pt
index b013c4b5c573d135eaf32ab74e7b6c44013771e0..138f085a8a2ff30fdeae25350ca299d51ebfad0d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.fc2/_0.pt
index 15e45d3d794a79cd8aaf60da55a056f18a9accfb..7b5995fdce885555aecfae151c6f841f988b62db 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.fc2/_s.pt
index 168268bf81f215222b4e63fb93b5ccfe01b94e9b..0e0fd7d467eea3927b101cfb7bafc8d6bf4ce2b3 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.k_proj/_0.pt
index c6f16ff10426deaf916d0feb905bed3aa90ea6b8..518f494c82ecb6a9af4581bff0503f61b4338ef8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.k_proj/_s.pt
index d1187690adaf055f4404a7a652be91c3172309f9..4b8e7ea44e9ab85d318f1660e988bc679e8ceae7 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.out_proj/_0.pt
index f300f7658da342b26b1f3197cb0854d70ac68d3a..10b780b0322ec306f5c364fe48cdac444309e8a0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.out_proj/_s.pt
index 36bf2df82453a3fc2b65e4627831e938083eaa75..70f72ed76c0f3a0125e5ef539735cf507291e5e2 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.q_proj/_0.pt
index 8a4a8a2e720e5edad887ae64d727afba8fd508d3..abb151757be4e19bafe73b303c26d01f2e38463e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.q_proj/_s.pt
index d84e4ea83319354173789fb0981a2070b2dd65fd..4ae285e0728557bf262e881a62e4b0842cd15e64 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.v_proj/_0.pt
index f3c307c8d49f4a610f33c2c5f0f7e7746030e80e..40f37392bbbcd74da24d12fd83a52cf8dd898213 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.v_proj/_s.pt
index 0c5ed2857e22954268d43ceaeb9627f888cb8619..d3730b1d6606c1b8dd8c7437d976becbdd296636 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.20.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.fc1/_0.pt
index 01b35c4dfa05ee08f3fc737aa5b1f441c9fbe8fc..16338114fafabb0aaf855ea8bc7e8e3656963a3b 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.fc1/_s.pt
index 8a87317fa06a59f53ad10065503a20310bea8da4..e0bbacf263a138aafc7aa536eb8f2e4f18299a85 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.fc2/_0.pt
index 2fca48d00e812aa2899e1b03b18ca9659a4410ff..668bffe8ac9152f884c3b4195108912e9d4eaf70 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.fc2/_s.pt
index ed80efb8cf7dd1525f1d4db7ef15fa49da4ed85a..8847b8c42648b7a5bd65d42d40bb7d92f58c7a12 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.k_proj/_0.pt
index 19b257ed970347306fbb5f2b901b348ca54a68ee..75f18e102227ee22ac586f75b47fac7ea5719c03 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.k_proj/_s.pt
index 26975a248be53e46092f6f68254a89fa73661906..72ecc710428de2ecf1086d7e80f5477b560812f2 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.out_proj/_0.pt
index d399498bfd785b1baf80759590f3738947b41523..d94ad0fd2ca89092b466e1c00e8ffdc76055cf68 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.out_proj/_s.pt
index 79f439272c30326580355d7783028cce2defc44a..534a0c7c9958f5ff94250c2344719136fabf6162 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.q_proj/_0.pt
index ea1774164f8757b73e1870e4761d5584982756b0..76ada260fda74649404d41cc1d627aa704973b0c 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.q_proj/_s.pt
index ebed9639b502f5239a36a64fa51421725b52c751..df712e0fb0e86f1ba936ffc7a87cbb83f7fbc9ab 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.v_proj/_0.pt
index fedb3b8c59bab38d16b42b4ca1fe541cdc851a76..1121b3a53fae55d29235452ced4dd7295c704161 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.v_proj/_s.pt
index 3d813aed01897eb28e0915070380516efd542de9..60c6c13c1e6cef29b0b4ec1ca34b955491a64138 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.21.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.fc1/_0.pt
index b3dae288d1669f7522447cb84c297d71b7161b9d..0ac95294bb3ec931c0da782111bab33cb67f55dd 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.fc1/_s.pt
index 1de591e98e67f45033f8266ddfc3e5752bb93373..77f15defa26cc4b89ec74b7dda5053b6af2723a6 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.fc2/_0.pt
index 10b6a972014a14f3f95328ead3adef618bd9aa2b..35d466dc3cfa4e2c73ef1e05369ada0b4fa53f39 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.fc2/_s.pt
index 57049ff727c15c0c5ef23b47a73f56a68c0aedf6..8b490fe96b160933b311f91f212ca12ce53cc250 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.k_proj/_0.pt
index 6d14810bf4be10cf2ab4b044a98bb231069d9ea4..0e2b8b68f27e85321a8e7206ffab74d4f35e4bae 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.k_proj/_s.pt
index 46e73f3d0d1652099179fd477b5c3360c1b1575c..e6ee25752b4a0437f17901f30472883844c9fcdd 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.out_proj/_0.pt
index 7234c864a9549062448b17d85790111cad30d611..4b5e14f7b783237d81674ebf3f724d2e22009071 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.out_proj/_s.pt
index b2c51ec9fc09fe802cbce1855f5dabfb5dee9e0b..dbecde5745c34c7f309bbff35c27b0b5455e5f7b 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.q_proj/_0.pt
index 5b05fdbb92cd1d6d1a8ed085d2455f39d2814de9..242d72f8cd56ad4d7607dab2e70d38c36dc1bdc8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.q_proj/_s.pt
index b7518e475ed3c8eed139cb9186e6ee5edde949ef..86ebac329f6ebf482dc9e309ec7e73225e1c37ee 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.v_proj/_0.pt
index b3b6110a8e64b4412d8e940ef8086a936b27b003..19c98fed9ee676233ff73ac562aed9ae2514267f 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.v_proj/_s.pt
index 483f4d6bc47e46230c0606523f8ea89c66a7c8c0..810112d0dc11b41259afcb5158097478c8ae4c48 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.22.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.fc1/_0.pt
index fc788de092d7298e8c93ac3ea2c4c6f4630c492f..3a627c59ba1f4e3f1b97247a511516247caffcbb 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.fc1/_s.pt
index f90464eedc2085248796d77a49d2c7cbd46c42b6..0b3387a25a1b87b89adea397cbb924410e7b2d4a 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.fc2/_0.pt
index 99862c04df44ce00004f6776684b9a45e5c8e04a..72a35ed4072190f637956f5633cd63e7e87e46e5 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.fc2/_s.pt
index d88da45310d065dfc1b0b5a7b3f7a197978f2961..dabe8b41679762f82c05bea2303c2ae7008b5dc3 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.k_proj/_0.pt
index 1351889de752f584bcb761dcb6bb2e3f613b62b7..8b36aa5ef0d82a0abae9549af5c8e840f01f13d2 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.k_proj/_s.pt
index 838577cb4c3a066f187ae56b97d934be84917373..3ddd4ac28b7c255274715e5c4ec5e48eacce0a1a 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.out_proj/_0.pt
index db5b8d3d849656eb3548ea60412d1ae14734c173..8c28412229d2c5600c931741c640178366f86893 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.out_proj/_s.pt
index ae8f39c401acb95b3ea4e2eb35d826eab82b895b..98731fef2ad147c1d8ba8faf0e3ad93f83bfa7c0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.q_proj/_0.pt
index ae91a5d8ac52419ea52f24f099eea2d5859044b7..f2bc9cf0239eb572efbd2508bf4978c027e28dd0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.q_proj/_s.pt
index 6d3683b735b4cd1ec29ee4835522fceb9e4cba24..4b6920764ff5e7fc37402c18bd54c6a84a7171d0 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.v_proj/_0.pt
index e2110ab3042908e3f8123de8f9679a7c6357a883..efeb904260e16012fee9d065f0845f40e56613dd 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.v_proj/_s.pt
index 361c43ff3331fbfb31778f1081fc644e743506a8..f42d50415173a9f2f9ea1547e45a5283e471fab4 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.23.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.fc1/_0.pt
index cbc592f28ce0ba95603d0a1bbcf4a7b42a2a521b..077517c0184095048e24bbe1b003a27a1468b353 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.fc1/_s.pt
index 1d158cd833490f2b8752ba70b61f114b2bd1bb1f..06779511c3eb32a5a927c243a073282192ceb9f1 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.fc2/_0.pt
index 52fab0cc2aaa04ba2292b5f4c4120afd4fd319e9..64479c4635eb406b20c719b808c489f619b3783b 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.fc2/_s.pt
index b1fb3bbc044d4b971bbf47a4376d50fed9727c3b..7844c9294b317a428cc90fc79003fe1972c47a26 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.k_proj/_0.pt
index 3bd7f7ed02815f6a32129e62b6147bd4d9b05da4..e01258b7bd48938892d2f80c16fc16792983d8a7 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.k_proj/_s.pt
index 263fdff28215b69f1580a55aeb8737e1930184fe..4706a735442085b25b46edc2c0f6e51c8d2b4b2f 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.out_proj/_0.pt
index 126d78033070adb1fa960f40d6e4d321afc064d6..fa6eedf658ef742f3ba51ba06327069e24628114 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.out_proj/_s.pt
index 7feff7fde65baadec10fea901ac9987f3b9dcab7..a0f6c887201048be55470fd21dc0f08b199c0ba6 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.q_proj/_0.pt
index 2cf6aa2d9912b83725c490f8522078d75d990959..655fe47decbc7abb8ce7b7a9be450cf282a37802 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.q_proj/_s.pt
index f551c11069b7317bda195826eed1e205814e02fc..ce64397e98c6aa8256f971b739c6fa323ac8d7a2 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.v_proj/_0.pt
index 94c08094155f60863df222c3546a1f61b7368ef4..e456e68d58e11e033a22007ba4cb0db8d19247ef 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.v_proj/_s.pt
index ae71fcac6fbb9ec4e705d3bccafde50e956f0f38..b8c01bccc24eb5194d31fd144497bafc7d990fd6 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.3.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.fc1/_0.pt
index b95c1d5f7c6972bfddbafab3a4c805063bf9bfab..dcfcf50edb4400364eed6e388e56d6851033934a 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.fc1/_s.pt
index 5e6340f3d87c34cb932a3000eb2c48f4b497331e..81b14b1976cae1951e52464e06ca55e55ea7641a 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.fc2/_0.pt
index d648448a3bec8e9b8fc495423810ee1f1c3e1255..57a2f78b30b378bcff7b0eb7df1bab5038d274ad 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.fc2/_s.pt
index ac386579e10a733f0d35b9b67083225a6e3c45b3..9a22b54f8237d595ab010cc49689f50813f28f83 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.k_proj/_0.pt
index f32ebac3680635adeb84a3dc007d9edddbeb39c7..49109a03067f353e4cb35619644167a974f0ece2 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.k_proj/_s.pt
index 846357523970444afbf27c73929adfba2712cdec..2d29166a0fd7cc0339ae352a64e55713e7eacf1b 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.out_proj/_0.pt
index cfbc9f5b3d00a622929de8d5b043f7c9c888ca12..1a3622a97903f6ea6e99f8b2d1c9e3723cd65c33 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.out_proj/_s.pt
index 6926adaaac5b24ecefcbe77f23b8d811433474c4..86d7523e3196e52a4a0e3dd1697a819e7d64a23e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.q_proj/_0.pt
index 9b48267a44a8da4b940b4925f81a9e37fefd1533..13dad817adddba75d3cad4a9b514c1a2c0dbb1f1 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.q_proj/_s.pt
index 063b69b31fb333993abcbdb25ad8adeb4e4f5192..ad04f2c8869801df58773f40edbd8adc32ee2f6f 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.v_proj/_0.pt
index fc520f4028881bddc59723fb34444daffa2ea86f..c744b94f56ff1784b51dc06b490b44a70420cc73 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.v_proj/_s.pt
index ed802d9f1778209e2a0584b328c755bc97011f70..0b450dab74cdb43e0a81b3e5ffd4fdd7743cfbd3 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.4.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.fc1/_0.pt
index 61d1244a7df7da3c9aa408c739fccc6d32815a9e..c900a7e9822915300aabc1418853091922c35f50 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.fc1/_s.pt
index 91decc0f957a70e06ba4f933027f1151a6c8dc2b..73146e32b98c61fcc4caebc5532048cd72dcf88f 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.fc2/_0.pt
index 3fd165c0532a169be8cdb2a3ce6d0634b3a49248..0ba654f209cf938371a057559c35c825f6fb3eb9 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.fc2/_s.pt
index 338cec7bc21339188c12a258409a327ae962f3db..d58f9a4379ae1a24191805681dba6ab4a5eaaf6f 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.k_proj/_0.pt
index 42bea849a7f56316b17cce4f7ebb0eb0d44ba0bf..5a215aa5c32cd506051e6c792d403e484fbe4a87 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.k_proj/_s.pt
index 5d6a57189d88a7e372c277b8a0d0f56a911361e4..a5d86e58846edab257f1d69f8f8b6ce66e3ad9ba 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.out_proj/_0.pt
index 35973b7a2562f97313bc1e5c4569e0bbcebd0609..155ea2364b4124750f2000e54c1757a3153d7621 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.out_proj/_s.pt
index 4f16205c1b2449013f2a76ede50bf2af2f38302c..8137a75cbbae39a8cdf47836a320d3092d690e86 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.q_proj/_0.pt
index 18f33e70378954e69234b5f327c23234752c3bee..f2ed44f7ab1f90d86ce6a05475c2fbb2a7e76786 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.q_proj/_s.pt
index d694272340958f13e3a36a22b8b4706210d8a372..22245fe1df8237dd3afb9f76f88ab808300f54c1 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.v_proj/_0.pt
index 9605e2c52e16aa7e81484d109447bb41c483849e..4dc1576df4df3217e614bde9b918edfd9fe213f8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.v_proj/_s.pt
index e135b4f46f8cb0c212eadd0d37dcbe5d44ca9033..3e02ea7e6d5e193fb739f30532b9f3b24bf6ea4f 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.5.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.fc1/_0.pt
index 867836badd0239104b59c436aae4962518f4f1b4..f06a7b864fe4862321fb270f8a857ce615cff223 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.fc1/_s.pt
index 0db4135390c8c6b738522db2b62d5e8aeb2077fc..7ba8df619e49cf972879e9631b09f0252b4dfc16 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.fc2/_0.pt
index 49a6aa86961e43648e91cdbbeca9726d56f307f1..24f8cc9fb061a400039dc7ea43dab5df3e20841b 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.fc2/_s.pt
index b016d686af93b71d9c120229587331682470b213..d0e61a966257e3dd496cfb025530ff4668413006 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.k_proj/_0.pt
index 923c642260701844c8da570db1e7f9aa68228a46..4a4763dafcc75488c8bb5edc21b1580f0d4b1b20 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.k_proj/_s.pt
index 7df8e7b76e985df60134a0812395cc220950b5a5..a13ad46d697d3351bcecfaaed40a6e3d76a59f0e 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.out_proj/_0.pt
index 7f32bb440921ace456c96dcb936c6c1c9f7725a7..bdfde4ff8527bb08d427be719dc1bb95bc195f82 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.out_proj/_s.pt
index 8304f509f6c7580d950706adb4c42d6646f8569f..be5c751a1f8d3f5a189b94d1419557f7edd6bb01 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.q_proj/_0.pt
index 46f45b115e34f10bfac4196e3ff806ed20a30cdd..0f17796016c7ccb7f53b821df5b968eec6c7979a 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.q_proj/_s.pt
index b0b8b44f1bacffb054c49057bb24b5bd1cda0fb2..2b89e131490ed590c43d5a5611c425cbc93bb2f7 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.v_proj/_0.pt
index bdfc8dc99551b2af3cbf61c2e2a325af19f5a64f..51c4be588ef7711bcdfdb0ee468e698e72ed7af1 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.v_proj/_s.pt
index 31f230c015806d37f3590e05b91a8457a62c548e..368cdbff1c880a19b6f92f8a49cdeac71a77b9ea 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.6.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.fc1/_0.pt
index 5fb4ef77f1420ce409155458adc640ef832be84f..38741dffd252c904b10e75af5f736fb6845f83a3 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.fc1/_s.pt
index 2e0d342073ea9799897a62cb18edf6544f1fcf24..96e98e81ad230e3d6b6362ad3f06782e487cfd47 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.fc2/_0.pt
index 79e2941645e08f3505f2a58e180f1f26261fe606..937a528b70431640064a6364188755c086404ff6 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.fc2/_s.pt
index 542818c8eed1fba28c165fa4dc85076c6d7fdfd7..4451964eab0fa074d267dc48f667c2b9d58cdd61 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.k_proj/_0.pt
index 41ce3117e452fc57cf6b9113730a236067a63ccb..acdf0268a456b37a4363b754cf6e469565ec7f6a 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.k_proj/_s.pt
index 218c053fe97795ec23a128f98a25fe7013e260cc..38a2d89843c230a5ab319328f8b3d7e5938d1db3 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.out_proj/_0.pt
index 7727c0603626ed1079de428e68c4c4385808697a..c135f5da489ce184d566e5ac2c5b2a3440e7fe0d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.out_proj/_s.pt
index 872505296eb78ad420eb3715851a6d7155b39d2e..7d86b915cb3c15bfae103f290e5a2be0672ca40d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.q_proj/_0.pt
index ade8f40e422952def2194f31355c82fbc8d3ba90..c369acdf8f35855c5cd395c27bf8b36a1dc1e676 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.q_proj/_s.pt
index fea0ee3e5d2eb90c4bf8031c0891b08cbe32328a..a549fdbbdd0aa4353a9227faeb25431327890d03 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.v_proj/_0.pt
index dbacf539ade5d9526091c8c6a2607ea2ef30f614..08ccb86e2b41c9ec438b9aa0c2e4e3ed761afe86 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.v_proj/_s.pt
index 8903cece18c32f2a5dcafc870f3b1bed76ec72bb..0dd54c09343647712fd919dea02475a0998be28f 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.7.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.fc1/_0.pt
index ae69e2b7122c6aaa0bcd6e0e43e3f4b2fba81df7..96d27bc72be902b49d7ad079c264625284327985 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.fc1/_s.pt
index f52306c7c7cafa54ff8c1cf692715c6885d3da1b..3df8658943ac22369d4dd765390c78dbb5a21164 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.fc2/_0.pt
index 07c20391f8e7b4a7bb5c8a2842bcdf7cd38a5fb6..1f393c5095b4f362588187ae349eb1ace803f479 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.fc2/_s.pt
index 2e57f6dea4a23be525e3b7baf65bab895bce43cd..bee471fcc60c108905a6ffc706fe78957ddc59d5 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.k_proj/_0.pt
index 044955a1202804d3fa9e5ff0e469c5639590396e..72e073e3ca3e2126f93ba6c859d9999b3fff4ba9 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.k_proj/_s.pt
index 923b92fc518e966e1129652c27ad0ac93790f368..c42eb129e64a597a106d8a9ab31daed14fdc14b9 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.out_proj/_0.pt
index fec5182a3995b154be1f4960e434f6b108604846..816fbdb80c405275d493bc95e5d35ec81870bed5 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.out_proj/_s.pt
index 5350eb8fefe5c786be95315713ec6d873ee09735..c0ab8415805007830507341a6ccde9e50daf35dc 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.q_proj/_0.pt
index c20ab2c5919f853f0702bb66555e9d27bf9f4bc0..20093b2e553f56ed8e4dd058451e0530f5611842 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.q_proj/_s.pt
index 558e6c58bfd094a76131da8c7b46776e362edeb2..c546ab78781108f321ed172cb106dfd7067adfe8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.v_proj/_0.pt
index c89e464a53adef6a300a816ba9fd13ed8417b024..db1f967984b547aa093cc63f8c25c590eb288518 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.v_proj/_s.pt
index c81643ed1e0fd4ac09b37a978e35706b4862d220..649a68bedeba02a178397b3d4e9a1998eafb04ad 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.8.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.fc1/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.fc1/_0.pt
index d3a2f3e734b48bd731d33898018791731e751671..662fb27610d4a6cf87069b8ce8232c3b9c91bfbf 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.fc1/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.fc1/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.fc1/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.fc1/_s.pt
index 445b02293cdba2d836b8fc6f88dea10d850cb89a..760822a67915f3891cc457dceabdbe907a6c1725 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.fc1/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.fc1/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.fc2/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.fc2/_0.pt
index 5f5ad55c59f7d74b616fbde6b7a1dbd3435eb948..5ff561e0d9d83b076648dace0339f64f45dfba67 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.fc2/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.fc2/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.fc2/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.fc2/_s.pt
index a7314bff94c8b2d9a56ba81fa5ba06bd7f64962b..ca2c5f4de182f2618f25c852e0ecd653cba8e774 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.fc2/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.fc2/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.k_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.k_proj/_0.pt
index 24b58051532aa821b005b24204cd3878aa296827..e633396b74687fdae2d95fc24d28fe1fffdff327 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.k_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.k_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.k_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.k_proj/_s.pt
index a5d73a0d15c2a248e60a76dbe563762044bbd6bc..0fe9218055f6df18921fced2500fb16ab36d4e3f 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.k_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.k_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.out_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.out_proj/_0.pt
index 8e1be843d6fc8011cbe7056b3a1344d1cb49e610..20ee5d00fa471f19c5a6455a9057bec77125d7d4 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.out_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.out_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.out_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.out_proj/_s.pt
index e917b9fdb76cd96401c07c2078e14309f493e61c..aae46e2dabc5eb2d55293c9180a0458c93a7dc38 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.out_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.out_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.q_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.q_proj/_0.pt
index de7a97ed06465bbfaaa35a28eb0a4f533328444b..10a64baff56bc88d69b2aac1f4748b598d6ff40d 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.q_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.q_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.q_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.q_proj/_s.pt
index aa98f390ab72493d3a0e3f12a3a64c3f187ba75a..90ea3b183498e2cc774b5cf943ecf43f1decab36 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.q_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.q_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.v_proj/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.v_proj/_0.pt
index 729a0c15cb39748cb0f9079f01bbddc17f4443f8..eedfd3c9f9b2f58015fd8bfd78a79a027b788fc1 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.v_proj/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.v_proj/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.v_proj/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.v_proj/_s.pt
index fc8098a08e541e33c4c087e37cb422d8d5e9142a..822f0f926cc04334c7a6bce5b022524599b409d6 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.v_proj/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.layers.9.self_attn.v_proj/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.project_in/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.project_in/_0.pt
index 234eb0f9ff177b0add4c77be856471962572cf19..6c91fcbd9188d5b7fa784b23f81c8440d98fd75b 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.project_in/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.project_in/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.project_in/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.project_in/_s.pt
index a7060810718bbae60c38386141b58c8c09d59c83..1f2475f0863c8c0a01d964d8e4cc75e97afb96c8 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.project_in/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.project_in/_s.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.project_out/_0.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.project_out/_0.pt
index e71ccf18c87e8afa7e9ad3fc73aaf1b0c27dd7f0..89473d880a73b9cc67a49ed0ac90d6d50b6ec2be 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.project_out/_0.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.project_out/_0.pt differ
diff --git a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.project_out/_s.pt b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.project_out/_s.pt
index 41db0411e3eb26b64e2d15797e221083d7e89ed5..d55131317d30c537f57f16e1e7b294451c1d8271 100644
Binary files a/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.project_out/_s.pt and b/params/opt/8/rnd_search_t_cos/comb_9/trained/model.decoder.project_out/_s.pt differ