diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/lm_head/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/lm_head/_0.pt index b26634d31627d1a0044c3f3651ff144ed5eb8f4c..257f78e580e29f2b211849ea7bc8ce9dac122cc9 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/lm_head/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/lm_head/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/lm_head/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/lm_head/_s.pt index 229fa2a280b7277e98ffe62da526267c9b27939c..132df72634062de2cbc73389d5efc07c2094f4ba 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/lm_head/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/lm_head/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.fc1/_s.pt index 8da9a3aa4a240252453c5b513f265bd8f7a523e5..4dd0b8a487f4cd15072e888c68ae3c9d6fa65de6 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.fc2/_s.pt index 64684459cff69ae140dd652f1a3cfce1c04db3f3..ec31a646939dd53653c2a8cedadfc546cc4b17df 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.k_proj/_0.pt index 489a6a063ef950f23ff1ff9861007f2dc64bb851..295b1d74f7ef78827adb0192c2f23ec1bf2a7f02 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.k_proj/_s.pt index 561f7bdd468cb95fd9c798a324062d1a8aafa3cd..9de20f86403112cd0891034ee48c79583cfc3eca 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.out_proj/_s.pt index 7e2aa3f8c5f633f3a9b8f8ff7847af7ba118d37b..5e3d7b39f8f7d294b4c55d3e1083dac4507e511a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.q_proj/_0.pt index e1d875889927a4a1b59276bf3151b47c319c68b9..c557e4a3c387a007eb7c078480304c54b0ceb410 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.q_proj/_s.pt index 703a367afc9c567e03c2486c4ee562ffbe0f6cb9..bd5f1cf3d7b64b482c5249c22b4fbd603708023e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.v_proj/_s.pt index 5d39ebc700a5351456013e6fd638785f708be581..07b76d1774cdf2cf3325227b07adaaa4a9c790a2 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.0.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.fc1/_s.pt index f4ee49caf36187da617d00cbb280a89b7d523fd0..73a097c92184a6235147cc44c10de7d3f4c4c16a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.fc2/_s.pt index 5ca551fa4cde963f54cf00ab4696ee01bcdf0127..09e9a5ce43b4a5704cdf66a5b287bd5fea71d3b1 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.k_proj/_s.pt index b60336640ee9fc4678c29288e9d2682e89a0fefa..80ec18ecbc50ac059193f45b3219719b22265231 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.out_proj/_s.pt index 558138af1c041fd4af32ced9cd65e1a9e3ffaec3..15c231e8e3cf85b54f7bfe528425ca36ac1fd5a9 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.q_proj/_s.pt index 2b16bc2b84d892bd7e6e4f32e7421c267f6015f6..35cf8c875147b1a2339ed38952129d412134e995 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.v_proj/_s.pt index 34a14421a354f7ce0e6c073227a4a1260a87013c..43f21210c3edbeac1622e7900df890dc1a18db97 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.1.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.fc1/_s.pt index 184df9d5c47144a4c30050b17e451935e63fa712..946470bfb5aa6e7ab1021747d6bdda0b0a60be7a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.fc2/_s.pt index f2856445b2b4b141ee892b270b4e5421e88c83a6..46307f242ad73be20526ceac03365bbdbcba8773 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.k_proj/_s.pt index a4162cc0584a46ae9ae4d092ff5c4bd18ba4274d..94c98157a74ca2ff2de17008582ed5aefa4f395b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.out_proj/_s.pt index 1f5459004165094ea5b11ca4fc170de03db43804..dd1624b0d1bc842a311cb34a631f665255e7c9f2 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.q_proj/_s.pt index efedecf0399d77e3f220870b5086aebddef28dd2..5eeb8c1840e15b07bb303fbeb1749dab22be3b37 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.v_proj/_s.pt index 7fc3a9750842d50bb2bb80ca0bf2f642b22fb286..91a6f13b543c280afff302485bf654b6d2ee1911 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.10.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.fc1/_s.pt index 157a445f98d0c38421a7d3609f5b54d19df9452a..58c9db639e4e0f8d8fc33350a2d216a554bf62d5 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.fc2/_s.pt index f5b6c53fe2d1513d579a9d756f15fbfc999af30a..cbb588bfb86f52e9224ad4ac0f5006fc747ae7ca 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.k_proj/_s.pt index a1b03bc49858c8c8edc63280f1dc8ae343ed8569..f19a100ffc738496075994c7d5f538e8c3748e09 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.out_proj/_s.pt index c5d6ed9848560e6344e0317fd169f8965471dfc8..6cec28512012a80b6f197c6dcc9adb085a8bc0c9 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.q_proj/_s.pt index bb4c3542a7227ef85c8f46d3445bea325c4091f9..b5117d9223ae21e7f28340cd87438d9c80a36318 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.v_proj/_s.pt index c17c9af3a9592d04ad932324cea8ab841d282d5b..dc50f361c0f6a84132f3d257cc0415b96d531f54 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.11.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.fc1/_s.pt index e5ded6d3ad04b41705b4aac3102aac3a8ac1a68d..7e7444dfdc0451dd9829568776b4a4c354f6c236 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.fc2/_s.pt index 63b70d86406e2ea102e50379b33b53ea45a20a51..8aa9d693a69ed93ea57dbc04f8a85dd13ec6b1cf 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.k_proj/_s.pt index c86b7ec12cae51ebd0d1d1a19c5ee8e9fa0a8e47..d93a6415e6ac051aca633c154359442493feda75 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.out_proj/_s.pt index ee1e5fb6b11f478490b68660374a27048c77cb26..0b38db81831b50a9d6bea47522ecbda905e554fc 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.q_proj/_s.pt index 367d43d3e46dc782bb7692186e4fd81ee3aa340b..3b81b8b21b15e05b3badc59c7918b9e587d63b44 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.v_proj/_s.pt index 81333d8480ab778a55e5b4280f078c94165a85dc..0610005cb5810ad326b981b1946210cefe2c329e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.12.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.fc1/_s.pt index 6a4b8cb2d34c02af671c2de0e6b88ae0c5e1ad28..53bd9c11d66f5cd9d0f41a69c343031c3c1aaa6a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.fc2/_s.pt index 36b46a2af6bee4fb7c926d9f100ea4fbe10606d4..0c1a7975cec68e1db51922209ccaa73ac778f4ba 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.k_proj/_s.pt index bc0b3bd67075d23df75606fc73cdd7e73e1a7279..2c643f743a820d821ca8a2c51a6e4acf198ab012 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.out_proj/_s.pt index 112feba17a03afc1ccfe1eb325b2562595e8ebf6..8421c8f884e81ab9bebf5d5379b1384676d1b43a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.q_proj/_s.pt index d17fec093e866b6f95e5062ec3523549cabcb8d0..6dde0a633fde9dfdd2d533013454231ec1ddad12 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.v_proj/_s.pt index ba2ad03fecdb77d96584fc362e76e802ac69cee9..889fe076c27070f557e147426c08d1005765196e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.13.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.fc1/_s.pt index 457ec8eaab5425ba258189acb8febc6c41f353b3..79a79f4faa3ecfe7d48b49a7dffafea47f61e85a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.fc2/_s.pt index c1a52a73642ae0040a159312a65a9635832e393f..1944b5a141fd06e04cff4e64961dc202f4481a56 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.k_proj/_s.pt index 546bf4af78d8adf7ee359faa9c070af895144fba..86f764642ba8726887cedcca2509da4ba675c10b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.out_proj/_s.pt index 2b93784f8a6c78f1a7c4988a5665161498b9f094..c892c2008e0d12599216492262d4dc2043421743 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.q_proj/_s.pt index f8e404886724c11d60c55e2661741766a20bbd2a..6aee61a560c9714ea0e11cc792d63fd19b3e9b98 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.v_proj/_s.pt index 218445400363aca2a8489e43e05d2f47e1347637..cdd842a4ca3da617177f0657c5acf7742a2c1a99 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.14.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.fc1/_s.pt index ea4b2f69dea4d6040bed69f6c5bccc043cf8cfc1..a48d6c906baa9a5a2708f87f959581e6a546bc7f 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.fc2/_s.pt index 9b4badba7a232c736611fd467a2241e761eedea3..94fbc6fa3f17f929e2e55c82c992cc4f907ebff7 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.k_proj/_s.pt index 534ce53a6d1391a94519e3adcaff0a5b90f23aaa..f10148fbfd26039bc29c447f8eb56f8b62c69a57 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.out_proj/_s.pt index 129099a749f994ed1d605412de4fa8650db01d1d..efa4a7081f23752326cb42ef95d036d37542683b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.q_proj/_s.pt index 8971008dcfc6a6f5c3026862fad7ac8ed55e01f9..614d2bd6a9f568a30489567c8928b74cfe10414c 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.v_proj/_s.pt index 09f781155cb85f14c2351044a16f5bf61994c912..2612db77441420289ec85930e369bf2c7f88288c 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.15.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.fc1/_s.pt index 35121819c77a3878ce35ff8a69f2fb21e3e04fbb..a86d1ed52bb6bd0ce415de1a0db181659b3e38a4 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.fc2/_s.pt index 3d150880243a392289c160b385e74755a09d1369..61b674460e7612c74bb6fb53220dcb26916c105d 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.k_proj/_s.pt index a53a8177f10874fbfdb19d8624af1b3332b83cfe..fb47ad3e745d3a11a643139f209eebbda8c776fa 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.out_proj/_s.pt index 851fbab0703fde6dc8338997b1bb2f1dc1989307..c4f40882177bd2e0265ca925f0db7c278351dc32 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.q_proj/_s.pt index 4ce20500ae4606efbe2d756b00cf163dab8540da..ff4413ccdfe58ce650836bcf989bfd76fe5d98d2 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.v_proj/_s.pt index ca766fd0016252c0d8ae2a2494a43ed7e48f4f2c..2e7fef6efd5f5dadfc18b5339d1e5ec6578cdbbc 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.16.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.fc1/_s.pt index 40695b649b76de6e2e6b1b46c1255fa69074c378..1674542afd1dc3e630a4fa8c213618a73dea204f 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.fc2/_s.pt index cbb91a8d1f0474ae3738a5e422a7c05e34da3950..550d78cc3a7b0115941af3854080571d3ade21e6 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.k_proj/_s.pt index 2f3132e27cc9b3461b2f8c60cd77da09f45b4c7f..6f3d38d569b3dd10f49c26bd2c6265a510ad106a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.out_proj/_s.pt index 5bcc040eebca3a2552ad9cc76ebedb297f33a10b..989cbe9aea7b9cdf8b58c1ddb7eda8c08b14b548 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.q_proj/_s.pt index 4442674710ad92f1899d5e672cab3ca3e838e7e7..7514d55512fbac9e5e0ea76a6d15b79f4db6bbfa 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.v_proj/_s.pt index 73af24acdb5266f03413380be9bfae4a61789c65..e1e12414b77cf8a5fbfb1a8d8f31b624bae8899f 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.17.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.fc1/_s.pt index a9a7cc8f8442acf4065d90cd1ecd5cea29417779..3bbf40b2d55c9568973b66daf0691c835cb6ae06 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.fc2/_s.pt index 299b7e2b641a0fab9ff66d355759eda1fe2c8d5c..d94002017242292e301bd0d5f5032147a166c5a0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.k_proj/_s.pt index 5553f5dca6a21cf93a5ecac88be1f698660f064c..e813b7345e9cbbe0dae8cad6e1cb1dbaabca96d1 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.out_proj/_s.pt index d442d7fba48f338fd212faf3a183ef881efeb3fa..23d48b44a179991dfb619ae495a35ac89c46b709 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.q_proj/_s.pt index 56f66e13ae351d6a7f18917efc8904586597225c..f5b37f4b6f08cbe2ea151e819601e7eb0a0266c2 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.v_proj/_s.pt index 51348dd314a623e59e5c741a1727124e469fef92..1bdd70c0c4123fbfb2e0909f00af081d94bb6e32 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.18.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.fc1/_s.pt index f841ef83c268e40d530c0c436cef7463d526eaa6..c821ae11d6d64e22c8fa7c374416e832e11a9d34 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.fc2/_s.pt index a300620b974398245258e5f2da46c1014d394a29..727a26ec066679f512892ce6641057e3686e7bc3 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.k_proj/_s.pt index 0efdd60e7d4118330eeb93b453d53c6236b33007..29fd4f3719599eacdecbaaddece8478e2cd2c29b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.out_proj/_s.pt index 9c7273f95e9a7a47448b14589fb699358777307d..ca3ae259358bdea7cadbc38aef293dcc438c7d78 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.q_proj/_s.pt index 566740a8d4a844e3365782660557cc0ad6f6f064..b440121b1eb375a53c6352040daae1bdca208ad2 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.v_proj/_s.pt index 92fa9bf121bc729ba8d47a6116c562f77b5a5c01..aaacc798189672ca4bda3d075b1ae2094e984675 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.19.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.fc1/_s.pt index b53ab651943de7e06ba38b3714ef9b73e9eee306..39838872e10e582736c6fffd89812b68a97fb560 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.fc2/_s.pt index 0e23186eec9b9133eebb9c5478facab0c06fb1c2..dfe9daef0a60bc41f8e175117e528202431af235 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.k_proj/_s.pt index 4a8c49cf6f5546b2e33d52d7bc226246b55b9cbe..a2c4dc04314cf7d5c9fc8c6bd2dee782b9b6ee75 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.out_proj/_s.pt index 8c1874b0eee1c0cc398fb3992e86e1c3835c8697..5b5ab0eb8bbd95af6bfe596ea4bfec619048d2a5 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.q_proj/_s.pt index eaa341d1013d0f4410f6b97e3ff6668cbfe5f1b2..6ee63e64e73cb451444ae7f3e218a20a0c3ab22f 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.v_proj/_s.pt index a71ff9be524188154bd28b3172fd3fdad7d18084..d8ba43bfc2e0d16d542adadd59d0ead0396f868a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.2.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.fc1/_s.pt index 002872135a4d1df0231b6debea4a37d8cfb26ba6..5d73c6ace022654c6d9182e71fefe9a9ae11837e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.fc2/_s.pt index 28cb5b28dd215dfed70effaf9c49e0608bf676dd..cb5252ddcd63508754ccbb8dee2ba36da028420d 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.k_proj/_s.pt index 5598ab3d28b201fad6d33bc04a5448ab1bf82b97..16a3309bc5c3a6a0acbbe632666120f356e9a9be 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.out_proj/_s.pt index d4a9857e3f858237642fcd94f2183ec919fb02bf..6970619416ab34cf1909a3b95ef550bbb1f1351d 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.q_proj/_s.pt index d3e43810fc42db1fbb01c62861f5400c9f62444e..ab808e2059cdfa78f0e67472fcdade2251db2005 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.v_proj/_s.pt index 0852be129a8d9636a6fd6f2b483e4b73fd6d8613..fc319049c9ad14a164090beeca744230d404bda9 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.20.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.fc1/_s.pt index d14c8cfe4c638fef3f346ce8e92af13523e30f8a..1e8fc7312b2a50cec999899bd8daa2b5919a4052 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.fc2/_s.pt index b5d84ff5d519b05d8f6461b96ffb54f8f03888f0..7b9bbe8b5c24913b0e6408b12557b44d02247cbe 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.k_proj/_s.pt index fb5b5f470ae9278d8c44b8fa697e215ac42ad6b7..b622e85c497ca74720be8ab0949150d7dafd838d 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.out_proj/_s.pt index 30d63a3b586ca9685fadf1d24a0cec1d03f78327..dc358e839cda2a8990a98f20afa8a0bbab9fce14 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.q_proj/_s.pt index 116ec7b4d10b81b5d761333ec02f1bc0304bb5b7..682a8efa27682ebdbe1a79b8fd26ec761e473660 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.v_proj/_s.pt index f1be18a5eccfdc2dbf2e0bcbc6814c9518f994dc..9d6ce6303db5348d2d25b527cf328430bdab0c74 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.21.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.fc1/_s.pt index 621dbed330437a2d696cb5760b6efd0693e2dc38..a2e4c53f9f49b2959263306bc431dc2d63ebbb16 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.fc2/_s.pt index 59251e6978ff5ab587e186447e6dcde7778abaf1..bca196580de8bcc453a5da054c840fce632b58a5 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.k_proj/_s.pt index 84c3270ee8c7e793623d97c544869bf814f64854..441cb2ab93d7c85c9876864ca369f4ee23ab379e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.out_proj/_s.pt index a13f986d792ef41773ef6c1c893fd9ac247452bc..32f2a21a5840307c9018f545c75f6abb570b4d21 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.q_proj/_s.pt index 93340bd9c67e816ad21be64b348f6a15bd4edfaf..b7868b93c55ce06848031b701f670e7606c549f2 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.v_proj/_s.pt index 219fcefbe0493255ee028cc4df6cc6e95ae4cc20..2c13f4104c6ccc9268847523ea01685727b6444e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.22.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.fc1/_s.pt index 5edb5f44b9061d096d6cc265df155a7727f7e263..ca29ec10a80e86a3b9b53c2ca8e9c95bdde7365d 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.fc2/_s.pt index 2e65ba432e79f6148ec0cdcc508edfb1b8f73d70..4742765ac5cbea7c52fad80ace9c1a247e17b6a1 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.k_proj/_s.pt index 51c13d437b3fc28e21d68b66143a0937b53e8c3f..b112419648752f3cf1420b0bffd625818d578575 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.out_proj/_s.pt index 52600f01c1b6c6f3a47297686f9d48a47f83c956..234c6d6b08c9e173d947f1f1ed4a9aa197a490c7 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.q_proj/_s.pt index fe275fa33d5c4a0969fe80e40c04a67e28c07f10..f4f64ba992788d92dbc3160492249391c216962b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.v_proj/_s.pt index 0376cd39b7c1ac17158022b218c1f9614d503363..2c63e144385254582d2f60b273525faa74efdf2d 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.23.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.fc1/_s.pt index 5ab7b0505df87c10e9a7d5e32b3c49d458d593ef..f1d61c3a2faaf6cd4f80051c226ecdef8a740343 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.fc2/_s.pt index 12207a370841ae2beebe4ad9d90550a5158ee9d6..8a399d6f4746b53ed0ee6ade161abafd5ffc3714 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.k_proj/_s.pt index 1b284573ade005148145de6f1b31c0be7e46d8b1..1ae16991da67bce92e4f6f637148be1c06227fbe 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.out_proj/_s.pt index 9c44417d0ca67db06ccdb7a10ed9da57f0820e75..ff27bf4c25d949e14a3481f66d07a69b2aa1228b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.q_proj/_s.pt index 5719e8d2d7b1829e11244dea9fdadf5ac859a06f..ad9be6c0347e410a9116dc108f369743b4887e43 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.v_proj/_s.pt index 00e0c86fbb31321f77bc5d0f74b8b0d4f7921dfe..d222b0cd866aa82d18f36554e1d9a16ce3d482de 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.3.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.fc1/_s.pt index 09b60ea13080978a3af533655e1a468e860fc7a8..a963e66b722a9c6392b37ddbd3a87f862f0c8ce9 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.fc2/_s.pt index 06dd10ef864550ebcac375d3f9f93686e6a98ded..57068d16aaec3300cf34f2ec6ca066247b96679e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.k_proj/_s.pt index 7ac6fb2004e2c7d4b9e815ce091020f86022fc71..81743b8fc44cefa4037acb7de70d2c022f7626b3 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.out_proj/_s.pt index dc9301cfc087e9f9e32bee8c22e2f06039d4ba10..dc5542b2c2f7eb5a61ed18e4bcd6b854a6cf1d2c 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.q_proj/_s.pt index fbdf6f27d7a3a6c594f44a1a6189e9a9b6ea96a9..a00e8b4bbb49ad9fa2861b25776f6d759ed34c91 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.v_proj/_s.pt index 8749fb5831917ed2465aabe5a5fe4fa7d2477e06..5c1e36df10c3e107e4920c546c433b1fcc3ad50a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.4.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.fc1/_s.pt index d7bc9446b23552813f22aaec08deb5ccc9685955..a98e1252e667c8acb30edb95e4750bd6d98ff593 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.fc2/_s.pt index 7089bdd8e86d37e3ef630a3c54939ce9422390c8..32f1021aaba52f552543c10a0375587c71464f38 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.k_proj/_s.pt index 65b041c6ed9075a74014d195212bd3300e43c78c..03e55a90eac93b3e551b72db7cf78e45d1d6ccc2 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.out_proj/_s.pt index bf126a3d7364363918ee95ad8ed970cb169f0c60..1316b956871aa79881af815b0e09ebdb69278b92 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.q_proj/_s.pt index b81ae7d2d4d285165c786376ba1420d8eb844d8d..da0618ce3b153c875c8f368b36f62424f1d48569 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.v_proj/_s.pt index 22fa4c93b71f739a0132b321f185eebe56f6d06e..2405feb4f3fb3e9e65d97efa923464ea38b42710 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.5.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.fc1/_s.pt index dfe7f4ffd6e882773711ef19bd1bc799d4b81c14..47b8441d053efa7b58ed428f0c68c4a5c7265bab 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.fc2/_s.pt index 016598a8fde3091017da589e900fff080e32498e..e4d6e586a9a754e81826aea3a9901092e06c0beb 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.k_proj/_s.pt index 5c2509f04aeaa78ebcd73b1adb8491fb0c9a109f..d132eb6c2134d4b5058647ce249072b2532dd375 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.out_proj/_s.pt index 821320bbba24e2a509c94060c8a8a1429e90f173..7f3732e64e2b9cf433e1b085dd34ece2cb76e6f6 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.q_proj/_s.pt index 991820893c24721dc21dfd817d52ce1c5f6a5e0f..5cd1e3b95c0dd8f239661452d33ee8ca550de868 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.v_proj/_s.pt index 182cf173e85ccc686a949dbcd85bfc7086d1a8a5..68f814f5560b7bc0945a0bc340d480d5f3a5a843 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.6.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.fc1/_s.pt index 317dc34dfcdd526236edd3f518e13e8b86c4e15c..ffcdf93fc01d8e2e0bb92d75751b0120b1d6615e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.fc2/_s.pt index 0fdcb7a6fb1857b5d8cd3314a5d2443f7b353dcd..847ded7df666e8cbd89b0d93c1ee3034505d990a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.k_proj/_s.pt index 12e29476805205d46ba9b652ec1c81e657fc7fa6..51f0381caf127659ea5ff739c829d1b39e475067 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.out_proj/_s.pt index 072bbe114941548632e0a89430fa37afa4eb2519..84b4e203dcff1deaf5a4e0cd3f57883f261f7a06 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.q_proj/_s.pt index a622487c3ce2fd7707066f65ff79ddccfe9e0c28..40cf667db697bbd3f43166c546b7b48c4066df59 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.v_proj/_s.pt index dc6550dad21358f0d85977bec69182d5ce8addd6..271dc3742fbd6fbfccdd1818883d946726e20b9f 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.7.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.fc1/_s.pt index 51a991befe78017b73c92edf03751907cd127dd0..64e41f97c6508973bdb9b2aca47d14a6f2ad4303 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.fc2/_s.pt index 4f2eac0182920f219d05e38d82c06996e3f56141..c391c88d0a3d7fa7f864eca99d729e6bf84957ef 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.k_proj/_s.pt index bd3866a9f231c00c5429fd7a8eafa028a3dcd87d..ab50857685e689d5009e16b2dc441b033e9a5e20 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.out_proj/_s.pt index ba9d779804dc7e633246636153ef1159837866b0..c349bf86580c910349fef36dd5adac2831bd738e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.q_proj/_s.pt index aed566ffc97a24c342da18d76f2f3a3106df6b4a..032d8804dae033536b359103acf90b8ed67f851e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.v_proj/_s.pt index 3c271a74965c8cfea4daa314803106604f2e49c6..e7199c55dd48184a8980ebed0ff39915d9629cf6 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.8.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.fc1/_0.pt index 58c96248fd00550cc15d645205571ee67c337035..446eea5860a21090a5f7c7578829aa7ea1576bc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.fc1/_s.pt index 673082563cda10741d3f4e61a7ab47ada0b3d016..605b93c56ceb77dd7eef42f8af5b8f2f3efd55a2 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.fc2/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.fc2/_s.pt index bee9e3a1b90371d106129d8ef34ed971c49e6082..7c6836201cee42cf623601c6e4767f05cf2d4bda 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.k_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.k_proj/_s.pt index da8ff1ae80addbe235decec6b1997da7ab278737..65ceef9969f4049c1e7db933e9c9592ef8fa938e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.out_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.out_proj/_s.pt index e0b3c93648baeadb820764a80d3c22187ac52cea..9e9b3c3b77b204b73d0ac63540bc8b708a34f09e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.q_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.q_proj/_s.pt index 5deca8653c6abd4e5ec7d77ad4b715d9e2cfa134..0405464b4f253b5b4cc798f6458e312611e4808c 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.v_proj/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.v_proj/_s.pt index 2c8a4b24b30e081e0fd3981930d68fa766bdacf6..06124335bbb422d376aea73fe34d48de9df89233 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.layers.9.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.project_in/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.project_in/_0.pt index 8ea38b2f931134a804d1dc0f42d8395e117dd20d..7a571786dfe022608d7168a415a64654991d2549 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.project_in/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.project_in/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.project_in/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.project_in/_s.pt index 6922a75e14cbcdc16c1c07054874967527271554..c20ef6b7d2342c76f3509a632f4679077f51ee39 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.project_in/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.project_in/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.project_out/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.project_out/_0.pt index 391a1fc6aa3ee122bf79ce6d151f25dc95c2b348..fb58641923148487d653a9dc7f80a671352ffbb9 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.project_out/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.project_out/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.project_out/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.project_out/_s.pt index 84a099c9c06d210dc070796bfa13f1ac1b64e006..f0759f7d38de279b2901965d14759c9762f7fc9f 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.project_out/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/init/model.decoder.project_out/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/lm_head/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/lm_head/_0.pt index 77a554662ffb7e43e77b7f5fef560fe66be4fe18..c0cae60bc1efdb9be747892ca49b4c15cd3c8af3 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/lm_head/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/lm_head/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/lm_head/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/lm_head/_s.pt index 36f1ee826dc9204d67ac50b3d7eae981b22dacb4..3d82b088dc011288e959fcbfbf9da7d23005e5bf 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/lm_head/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/lm_head/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.fc1/_0.pt index 0c500df6134efa60411574806b620b7d4a273d0c..be426a104a79159bb048f7d3c569bcad9bea812e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.fc1/_s.pt index 16a80a81d8e9fc2ffa0e4409d92391c2ec51b2ef..a9864b9624ec25e85eb5fdbf0f1eb31ddb3881eb 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.fc2/_0.pt index bfd9d3756aba777e59e066e2c1107cacb4116f69..48e513f8a25edfeefd4a58493eb69e3c231a9e51 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.fc2/_s.pt index 7f3bb15c409af3d470269c57b6a2241ec604ec30..e50b6efb5fe2ac3050502ed02bbc933fa3714a9c 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.k_proj/_0.pt index 824fd3a33e7f262adb1548eca10cfd472b43d1c5..9d3558160d95a514968ae6e68a404ef1884bdd6f 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.k_proj/_s.pt index 62cebb7c537b3e9ac5cc4cd148fd26b551ff031f..2752856ed4b58abec85cd64d37b1a3668434fa0c 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.out_proj/_0.pt index 1c2e78042818ab259c5d9b28eb686013ca315c25..ef7b1e8778c10447c9ee00aba99fb75420fdaab1 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.out_proj/_s.pt index cf596881d768067a6f2120344fb13e7f91a93dce..04b1e7e227d159980d9f077360ae429b1ed42fc5 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.q_proj/_0.pt index 8b1c14aea30cd03058f3211ad39c8364aea0ab0a..a2bb3dfe41fcbe70efde41c06acffeb5f66266ca 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.q_proj/_s.pt index 224352742808fe38428c5b21acc76759482be0ce..8419cd55f26b039484cd24c6eff0b018f3f4abd2 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.v_proj/_0.pt index 9db1eef53a2fc211c41cc975d665ce5078beff57..581634559e39091b2c87ed4f851960a49a68cc54 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.v_proj/_s.pt index d09badb76de5b5f9fd9672f4de5673ff47c18dc2..90efe9d934c4df9f7efea5910aed55f681b658b8 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.0.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.fc1/_0.pt index 3996d48fad605068e8a6705f89799b2004cf68dd..9e6b2b7a5d7163219f6e55da7089ec12815b2577 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.fc1/_s.pt index 5a54029550bbd6e62c491450a9eb93fed1bb05b1..5577a32d14e8b59b1b49bf7ec73e6990e222cce9 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.fc2/_0.pt index b5c681cb5ca343837e530c3106440b8a16c1adf5..89aff4abc55b94bd7c58bba6ee3b2abb825a045c 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.fc2/_s.pt index 5aa9fbcbdc24009fc42841032cc4a773d8e03f17..7e724a377d0374fba01bdbac0b1820b7b1b8301a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.k_proj/_0.pt index 36219ac2fca26643f65345272fa3ec45d932de4d..ebf5dff8d5ac189cbc50d939c6a8f102b6e41643 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.k_proj/_s.pt index 4c5e4f236990f684ce7d09743c141721cef63cba..0d95b73be01f917a7e1bacb1c109627ef7f0fd46 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.out_proj/_0.pt index 84f4853f0b26403fbb54988631f1f72137607b20..b2d952f0ea176cd798d63a078d99c359f0531ddf 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.out_proj/_s.pt index 66f44d683a5c39ba8bf4e550d7a82bf2137764e3..5bf05e38867173504cbe270e4cc9aa57e6a13ea1 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.q_proj/_0.pt index 5cf3f21feb8af0ea63abb308234b4b55f417ad72..07c30657d92c9a8543be49621df97923f5444ffb 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.q_proj/_s.pt index e58fc2b1456f9e0f1f840464e3c2e40edc47e37f..3bb667062dbee14d449d0f63eb2e1f7d46295e54 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.v_proj/_0.pt index 97c21c68325097214d925743629981d743da3854..97acd20f77b9bba54608f45d6db97381ec96ea91 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.v_proj/_s.pt index a905b4d859a53313bd001bce928dfb4bcb223791..64bbc10137ddd3cc794fd0d94a4ebd6dac7b9dc9 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.1.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.fc1/_0.pt index b0710c4069633913c95dbc7a2d8a05faa1102cf4..4a28c95389c0ca304ec8d23f6674eae2ea56e70c 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.fc1/_s.pt index 732b7cf0bc43b3b191cd0788a240527194dd163a..a222b28f737bc115d1f603b75a394439288ce35c 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.fc2/_0.pt index e7a7e8a40400e225dcd6224562c0ae291af93082..fdf1bfef501c58eccb8e70228ea13c14f6acfb38 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.fc2/_s.pt index f0830a439471fb4fe357b46e87064bd37215c9d7..1e15719412f6eaf869d0c3c8214f0ecf0462529e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.k_proj/_0.pt index a5f95caa64a978de033960f07a8ac5569eefbb37..59dfc1d7dc8ad183e24cf861bea9b9d3fc0f53a9 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.k_proj/_s.pt index e4818d49786779a00683842a520cdb756f17cd67..84b0ddb236bf3784fd1ee9ac7910a609473cc4d4 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.out_proj/_0.pt index 3ac8a6bef78d79427212b402ac4010ad2f329695..ac3e9ca35d6f5beff7479ce92818abe9b24a795f 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.out_proj/_s.pt index fa411a4458133ef446622adacc152402e0ef37c5..7486d1adab70d446cc12e5d933d69b0c813b971b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.q_proj/_0.pt index 437f89f9b7b0ea2b77408be640ab65dba4f05fd9..fbe7faa5b003cac75d143d158e7d8d122fae80df 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.q_proj/_s.pt index 276bdfb42dc964cc17087e0af6e7e3954b3e42bd..ed89fab9d3cf74266767678f476ac7caa36fd9dc 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.v_proj/_0.pt index 0683102f9e35272f4841a2456b20c5a2ebab8dd0..e5078f4872ce8acf541bd99c29abd04c29e8d7b6 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.v_proj/_s.pt index 1fc04c9bd25ffc19b5bc3ad1f856174b6d1f25ec..b642bfeb26c9cea03508b608ca47bd9eabd32e1d 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.10.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.fc1/_0.pt index e2625a176c232921caa914e3e0040c457e436a3b..70730873435dfbd8a71f6a6926f5450fc6a9a2ac 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.fc1/_s.pt index 7e88b18fd91c413306340d31f829c921bd08b32f..e52dd29590ca18e0ed906436bbd1f6c024ee5752 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.fc2/_0.pt index e06aa74d341f5c749f7c5b0dfade18afda32d263..e3e4b50a8db3ac02ab73783b5bd21b61b18938ab 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.fc2/_s.pt index 72d8188d8d82370b45a8ed435cfc1caa19937ab8..2e72127e51cc477cefa5207e33c33aeb5a91323b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.k_proj/_0.pt index 5bffc81b8e837eef1fc2800cb0ba9c2bdf910513..163c29af36ad61aa668c03c71a92031817d6c799 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.k_proj/_s.pt index 30341555de6079f58460299c0db75edbcfa445cd..af446a155d35637d2616d33e0cb920f34790485b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.out_proj/_0.pt index 6f1f4199b3511ca9d4516472a078da990440ffd0..5bc90c20fc43a8fa27bb1bb6a84248b8946038bc 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.out_proj/_s.pt index 5df23775909be8e7f088b828c3bb1c3912e2b32e..1b7a38e1a45c471152f0a975088e87b452374216 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.q_proj/_0.pt index b8b5272ca2b4d58e2ba75d3928e842b262298eaf..350add31fd0b98a2005ccdf5890a348c4ad18b47 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.q_proj/_s.pt index 77cb565615cbdbfaa193528ac61922b0a92f786f..e2c3778ebdb83e679c4223589c565655e6759d25 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.v_proj/_0.pt index 32e8ad4e7a98fa5018b8f3782d0b537868e73d8c..258e938f4cd3345ffcdcc51de396d99a2d995212 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.v_proj/_s.pt index c7ac1d1a88cb0e06bfdb8546e16032fc33494680..cd5cee601bdd16ada9ed94c325e18b01a99e0756 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.11.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.fc1/_0.pt index 107c0ed012ae1880911692f717ce0f39dfdac351..500424f2934ab75135595c5d63f5afff55539b08 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.fc1/_s.pt index b9f517b13987552800b36dde40f52d12f1186d15..32cf354b755cad7ac89990710a3c2faa945a1ea8 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.fc2/_0.pt index 2eb99bae4dee96726375434f3a82f9c6ce8417c5..11802e4ac8c0c5656051c434c2f0c7e0363b46e1 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.fc2/_s.pt index 8da0f46d74c1fe954c72d99d058110cbf07f6185..f4417104e5493cc9462bbc6a1b85b7ddd9990d07 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.k_proj/_0.pt index d73b47d3eaa5c719dd335847f34ca50b7f30d523..be2319cd1eb70d854e8a29d59aaa6e5ab223ba3b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.k_proj/_s.pt index a223f9940500e592b124e0cd800e626cff37d76c..846cfcead1ccfc4d28e689f35773ac36e370402e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.out_proj/_0.pt index 78c2e5c10f383a807046101293458b86e7b8787e..52b09745f4aa74248cf23267dd64df3cea7e7351 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.out_proj/_s.pt index 6f29047693bbc93040263f51ec4d59fc6183411c..f7bfff6ae105afcafb36c6146a60cd8a3d199bd8 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.q_proj/_0.pt index 094dafeba25e85c4605c6fb8ac8fb69e2bfda04a..51e89c65be1eef1de77975e8a3810a26ace6478b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.q_proj/_s.pt index 1509c80c984dd2e8a2348c07d155033cb402f620..72017fc43f2302a82611a0fc0ff96fd980ac62af 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.v_proj/_0.pt index 131120e6fdd6ac0f382a41ba67589919a5cf3be1..3215d6a3aa9bdca2e792ca7e3ee2a800a3028089 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.v_proj/_s.pt index 181906522e0a6ea0fc065eb81ea13cf13f322794..41efe39470148e78459a528dd29142053197532e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.12.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.fc1/_0.pt index 8f88585e02720806e2b86408fac796fcede9bd43..95a742bb81e063d714df313c4d56c57eae5afc76 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.fc1/_s.pt index 0135e961eefa5b8ef3221f91027e1551503ec8c9..2e3a45fdfdd4abae3cd6d00d387c22890ee33db8 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.fc2/_0.pt index cf55ca3f21eb480c736e3040dd794777a1f81d11..dea98ec8d2d3860623a881ab8f2a0b315f3e0368 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.fc2/_s.pt index 21458ccfac6d23d41b50c2fbf9276db6a7140374..ccb516c50830b308ca67af37da3339ee17d54f52 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.k_proj/_0.pt index 8aee44659eca28916be4975922606bd2ac3f24e1..4778337f8ceda11c471676cbc5d1f92e75d8db8e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.k_proj/_s.pt index 0c46cb17fee1b9eb530d145104ead128d3760015..e5108e8677780ac4fa4107e890612f01869aec48 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.out_proj/_0.pt index 603531a81b06ddb66b9691d3080634fba2740178..ff4f88e8ecd5a1b008764c3d550f06b615e6d826 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.out_proj/_s.pt index ffd9c3fae7230a26ed6f3ee138b79073f03effe5..e0f4c6b47aaf247f4e63ce2b09026a27725cc5ba 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.q_proj/_0.pt index 622caaf96ccd97eb2c82d3d91b0e116a06abb70a..d739ebbc196a0de42c63026f686c44fd375e37e0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.q_proj/_s.pt index 842d4c07cbff8186390afd1fa8f167e29644a199..149211026d4677591ae2e0318127248d859321ca 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.v_proj/_0.pt index 78f303f190038bbcc5516feb0fa177b073699657..159a071f53e371cde5da708c7b3f9bea74bbf01c 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.v_proj/_s.pt index 1c0812bb40517603c9091d3cfcca942923e4b47e..40e2f67ac6fc058c352c351b58620817e19d89a7 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.13.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.fc1/_0.pt index c5ca4607fbe3d2797309f29dfcd4cf00942bb81c..1d4fd84c491289d23104b2c2b3b8ffeb67c676dc 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.fc1/_s.pt index e2662d66a38db4afaf943b6ab7f37cc5d811dd9d..caab9ad7c2385d454bed8087c4eaa32ba48a83cf 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.fc2/_0.pt index 88273264e36fda72efe2d61f25e045c5e3d2b010..09eb4c1340f42d51bee53d2a9eb3ed4656f5c2b1 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.fc2/_s.pt index d5ebd17404755c2a240809fadfb1268fdb054d6e..089028fd1dc11cc4a40023cec27ba1dbe4c34590 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.k_proj/_0.pt index 5c4e5064f9d0b431443f525f78a0a1fa69d9c1bd..91fb044d088b4be75d1d605331d6fde2f9715dac 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.k_proj/_s.pt index 701d18162a7cd1e91842b66de9a7f3c0c3deb335..e1098436845d4c20f28c06a37d138e55283749dd 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.out_proj/_0.pt index 818002bec5c1dd4a316436b8c14430542d18975e..78cd490d693ad17bec4ca9a788c9aa457c7c6600 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.out_proj/_s.pt index 9732af707c4c9099f192c08d01a057053a468df2..8f42066627586106cc2f1bfa5c8e2e93f4d712f3 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.q_proj/_0.pt index dee977b287fe6447ee3c1e1bc274bd6a58d2e44c..a2f0b0214cf3a20a8c6a1d5b9de06d96873bf78e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.q_proj/_s.pt index 1f611ca4b478cb5e1928237dcad6e94eea4e1d94..c6404cd36b26cbc100f4191f6d82d20d8a977dd9 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.v_proj/_0.pt index 3e4bb01258d554b52b596e4140c621709bb774a6..11922310d3b1f830bc68b29f2421f124f65d7fb1 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.v_proj/_s.pt index 24f8051dd39a4644e4628be85ebe1f296512c66d..559f59b3bb4fb7eee014a0c4620d3b039f9dad37 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.14.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.fc1/_0.pt index 234653b79f60fae6c55453aab759d90e1bf25b25..db680516917b48c1145a42fe38b3a9f67a1dba81 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.fc1/_s.pt index 495ab98a4160aa224bf899eec23014b6e3bea075..822143ae2ce3c93497eaf6e61f5a25483b67ae17 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.fc2/_0.pt index f0b823ce89e8edcd1af3993992f1ce692eb0c73b..495e959ad98600dcb9b8f13061ce7dde668686ea 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.fc2/_s.pt index 7b99ab5ee3c8fa78ebc30d0d6adbc055681450bf..7bc00620647e5a7a517c2db0385ec2cee45f9402 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.k_proj/_0.pt index ef497de0fd78a234d51ad34598a943f8c0bdb465..e1c09ead670932b619bfc6007c5d2f5d1d987380 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.k_proj/_s.pt index dd188045449094e72acb5216f76c518b2b53528c..f8e8173595ab94e7cdce79b2bf1b83426790411b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.out_proj/_0.pt index b3fd37445a9557931ef83d763087ba5f7334f358..da288d6acbad51db046eb3971c48f000339dc7ad 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.out_proj/_s.pt index ca71d03a9bbad20f8322b65d75c4a34c66020c9f..3c18b768819d03ea8e3e4f51e77979526770c7ce 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.q_proj/_0.pt index 12c1b3785f114f419e4b72a5ad93e46b88e7a63b..55327914daa988f0274022f7f1e871b8887d2b0e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.q_proj/_s.pt index a65e9de21490a3395959dcfe607aa36786191bcc..8b08519d68f3233bf076712918cb89f2c8061398 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.v_proj/_0.pt index c155fe5a840f179169badce8c7ff4ff382b1509c..9ad86ee5a2ab0005f9e793ceb04a1131f6b2c4fc 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.v_proj/_s.pt index bc836602ebdcf6c84b721390487ac3f7afa2e05f..5da5202c048ea07a2b007f203f0ae0ccd68ab5a4 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.15.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.fc1/_0.pt index c39aaeee912f86433c8eb380dab50c4a1eb7a88b..1cb687eb28e1365be20bf8d9f304bf3e3ecf9fda 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.fc1/_s.pt index 036ff24fd6aa6aed3b5e93b76be99181ee5f24da..c56fd24c0cdc25aa43c9b7a53fafc3077b6b694d 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.fc2/_0.pt index 87c78e77c2eb9682bffb8884684c9ab8a0940cd3..53265da826e6a29cb74562167543968347e5791e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.fc2/_s.pt index 68a9b24ff4aa06387a2182cdb472d612b811e4a1..670d056453bb54a45f53ffdfd0ea3b36a11c0aaf 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.k_proj/_0.pt index d84965e1b441976804318b331264e85a761d9410..f1c9f175866bb3a8c3e0bdf7364b0b51a0c0a58e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.k_proj/_s.pt index f48ae4de2fd4f544cd48214dc6279ab8c19cff2c..09fbd212f3213defcdd1566838c3347c01ab6e02 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.out_proj/_0.pt index 5d49d948277dd9076d7edb250e76cdc0e6d60511..a7921894e217338c6e005f72ab298d0bc01e35a2 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.out_proj/_s.pt index cba9702abe262857836adce6c1a4eabbedf59846..dd4de03dd94deb62dae7180e26c309233f45cc7a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.q_proj/_0.pt index f44ca3f8f00cc7e21d52c4779802a90c60f25ca3..c6f6c6021199fd3d63c1d34aa2d962e2d0e13f93 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.q_proj/_s.pt index a3d2dfb914f51adccd166f3815fcd5bb835f4858..fae839515ff08fe238650ba60cb9f3817b0b47d0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.v_proj/_0.pt index 7e5f4532df5d438bdb912acc02641950fab8ce61..5fc41e7bb5c9edd39318d4a1e66cc3baafddef7b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.v_proj/_s.pt index 38aefc71e0af9e09e632a1838ebd4e82ea1a2c76..ae6c979ca9a1d55fa639970311909e5e13f4cebe 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.16.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.fc1/_0.pt index 8ad1f528043363be19bd50deacaae61bf92b48e3..5e8a614865ef484cb5b0d0936a1c104e7419bd1b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.fc1/_s.pt index 7db711aa3bfe2e640a62eb148a78a7cff4a207ef..9040c5726c7af1137eef3327700ecf9ca81abb48 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.fc2/_0.pt index 22daf94b741d36e766d78dd4b99e8319a838c7e9..18b3ec55f99ba3048fd42054ad2484554c9a71f3 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.fc2/_s.pt index 86e0cd1962e3476199b8315e059b37a9e0d0604d..27b88af0899d2fb21dfa0ea67db796c060c20501 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.k_proj/_0.pt index 1c08f1237d3b2a1b1ad4016ebe6da3183eebb7df..36aef486dbb488f4762d5a28b54a24a20e357b05 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.k_proj/_s.pt index 12d597724421a632419803652125dd7b88ed1adc..f4070586bacb56a2647c5001d7527ca4d4675dfd 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.out_proj/_0.pt index 8a322e5a82f6dd819c000df7fb288cbab9b0e93c..35470a8c26732d0be82091e4a223d1b82a33eb34 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.out_proj/_s.pt index 8a4abcced4565fa09d66aa9b663f997fe04c3659..205cf932701c3e7a909a9f9b07d1aa5d98c113d1 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.q_proj/_0.pt index b8e043dab805683f70f84e45616fe92c6ae8db3e..622a572f318e6adbd3ca12dd29a0609d1c8cc4ef 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.q_proj/_s.pt index 4f608210849ceb8809a3063d8e103d8af418cd3b..b36f4512771f90c4615c9fffe7ceb8d8a03fc4bd 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.v_proj/_0.pt index 04c7140f8c79bfc41ae30cc4e3b81943f75d0de9..123fa170fcc3d15bc149d844bbbe92e396925eff 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.v_proj/_s.pt index 5606e36512a8c58f97c1927bc3125e19c24b192e..806d36d32061b52bfaec0b61793f8dea9671605c 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.17.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.fc1/_0.pt index 6197843044b2e4629176a3a0823de3efa701ffc0..85b6ba4b25aead77f49dd24f50d5b22bef972bc6 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.fc1/_s.pt index 3a460b6657048ab5da3f38fe9a690f14d2ef9bcf..93c742873908c442dce0d246aab077f9e957759f 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.fc2/_0.pt index 412716007695d95938644a557d27137b6d521c06..e919524782c0ec6ec053cb5624e47879142ea029 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.fc2/_s.pt index 6bef1d74bd277d417cfb638c7d0588cd2a7b4a9e..80c724277b9ad08d76eb9669c5c1942f57d7951a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.k_proj/_0.pt index 9e9140c0d66d36210ab4533deff00e9588c14554..0dccc63c3d8c9d16804ecf1f05ad65303f875030 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.k_proj/_s.pt index 6a0ef7241f6fb5a65307893d7959b9e25e53349f..b50eaba277719c2e901159b8913d9eab91936399 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.out_proj/_0.pt index e8c109feeba3d7f1e450fe00d069a132d2f05815..83616d67ec70f9887cd9331e8a112dcacbd61142 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.out_proj/_s.pt index 499a30c347eda022fd286ba41579fa4dedfe5620..927ec8c4d5982fcf9aab5cba1ffcb4a33eaeba56 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.q_proj/_0.pt index c230db172a0259ff986ae726cefa0bde8e1ed1af..6e54f315c15d1805915bdc5ac3ff83f4a101b2b3 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.q_proj/_s.pt index 341543515b4a2e24db1d544eb4b8daf78ee4fb8a..a99399d16d59ab7e089859e7d356b295904cf80d 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.v_proj/_0.pt index fc088d953e617c229d3e446e46212e3688a4618f..e2cfdea96749c88170f3fe88b083131b2c2462c2 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.v_proj/_s.pt index 63ffe5156872221bb3c9ecd0e50f76bec408c6d3..0e612eb2f4dcd83dca00b0cc4c7659d54ac882e8 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.18.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.fc1/_0.pt index 209b33513470b17c2ef7629c409fbcf867ab3567..9e3c15cbb06c85b52588ce26b44940132d985246 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.fc1/_s.pt index 6ac07d44c901a65b5426b1aca4807532a76a6ae2..2aa27f11c784504d874551b888a35125a942914b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.fc2/_0.pt index 674deef59a6bd274278d722f64c539beca2e570f..d0dbbb50466c70f3f2feccdc4a90f7a434a5387e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.fc2/_s.pt index 726a5a3bd94c642825a3aba5d87eece554b6ab41..7627c5edae72b528498c3ffc3a82212d1991d381 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.k_proj/_0.pt index d9a5cd814dbbc8c3d97867685d241176547eb13d..2bb244e432f4439809262ce85321e57dba35eac7 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.k_proj/_s.pt index 13518d5a43ff41d4622011c52b890949b020e435..5a2e5ba883ca63ff875ffb4ad6a9c258d3f4ff82 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.out_proj/_0.pt index d3ad78bdd07cecc2ab394a6b149c278c11298284..4a356f1aea1f5ad0e5b643ed3a74f685fac043fc 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.out_proj/_s.pt index f8861f16aca952fd4af1ac40bfd9c22ee8a7dda3..b03a4a226a638b2fdcd6d1ace9ffbe68c17c01f4 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.q_proj/_0.pt index 8f258d6d07b01a29d1269d6a1f587b95b20520ca..552b185ab072ec911aedc7b146a39bffcf40222f 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.q_proj/_s.pt index 6721206e28279d84215a0bf2f51ec6885809a203..bccdd5443d093b106d165048973677cad9c07811 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.v_proj/_0.pt index 0bbcd8b2f0d6ce1bdf1ae294f43954ac8e949d91..1e9cac394620b75a28fe0d5335fa8d791efbd324 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.v_proj/_s.pt index ac813175d3cead0b05a61b1f69ad0059820db911..11af0c221369b9fc5f2c6c355fd799eb34d3a307 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.19.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.fc1/_0.pt index 52acd16d962aceb049481a9c5dd386bc682f6708..54e68b21c3a4cc322e74986c319f98c7038f8c65 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.fc1/_s.pt index b6a1c53aa5c380e79302fd2304cf4bec82674614..e8b34a9f8e275c535bc23d756255d576852c47e8 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.fc2/_0.pt index a10555b8c852b2ce1b3a9b3baa6d8b1b2370927d..83f117d79cbe25632790261d88fed9d52c09b906 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.fc2/_s.pt index fef76a55c40dd711e5b321babf480e39ec243d1e..b98a1951c6fabbb2dd88a4a1c376d0f3d9c772a0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.k_proj/_0.pt index 654bb7ba7d8ceff28782772cd8ead998c96e375b..4ff4e885576e1544ff7daa87a73bb694bd6e24dd 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.k_proj/_s.pt index b4a4d2328b091ad883c6c019109afbed982e3057..5f974bcb7ca336c4e13ce2a310632a7aa1a88b62 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.out_proj/_0.pt index 572f0226cec59986accf051ad69aae59514cd138..c035a479705aa9da73a320fdbf4d05af7fb90235 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.out_proj/_s.pt index f47a93831370dbea72905430346a1d9148f81e98..89498fff55d91f80aaba3bb83eafb1a914f1a627 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.q_proj/_0.pt index e1894e83cf396dd37704344854fca99d215c1edd..c5a55f8fa2eaed7c70828bcf77a47d07bc831090 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.q_proj/_s.pt index 192548dfd3fbd63edab29c4a8662e13f7e9aefeb..646c00a7de93ad598a1e8a8138240102162305b4 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.v_proj/_0.pt index 71bcc3abb2316742e0c458c07878024d31042fd2..1232e82ef23f5071abfd140e9eeb5dfcf1afa896 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.v_proj/_s.pt index b4f1ec973ab040d3c205b2ad4017fdd1a430f86d..fd45d0e6702d7904bcda1c4d86caa198995221e7 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.2.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.fc1/_0.pt index 70a80426e82e83ee689ef4a44bdf169154b4cb66..ffbb67ce3fadb68aae2e12b3eac29216cc2e9bd0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.fc1/_s.pt index 25707e27e2f8d120cc21199d734d872c51df623b..19fc924c051a0a2e3c6b4122f9ff17a771c4a969 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.fc2/_0.pt index c0fd7fbdc9ad004081fd6ce134773453d0e1ca17..4bbcc02b1ad35c68ba902daefa0525864a2c93e8 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.fc2/_s.pt index 0bdb5c1a81f32d75cbfc5f7830a4a657a1c5e2e6..af9c272a139f3110308647ed6d64f8ba93bb12e8 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.k_proj/_0.pt index fa2a6007167859a61d89552980e019f9596eb981..8fc36da0ee754ae800e6361f88e85f6123bbe2b4 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.k_proj/_s.pt index 6cdbc94a258a0866a79a03a4dcb8631924881dbe..f5da52fa561f2fb06a69a721280b7d67921f6884 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.out_proj/_0.pt index ee353f4fb44aef7c4973a3a1fca49e947d191bf8..1be8ec4527dd5437ed523ba644d282f5f91c67ea 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.out_proj/_s.pt index 3f06fd0d9c3f2d9668062806b7893c2944529fa1..b9a131f75246f2d6359c7187c3dc03869ffaa0e7 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.q_proj/_0.pt index 05feb36ee008efcc19fb058f10da1a27ff00168b..ff2decb33ef6976f0c5c2ee161ac3ac53cab39e7 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.q_proj/_s.pt index 2afc8b275bd7b29565af775afd6241989ba39009..0019cb31561bceaaa1490268c92c219c0118aa34 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.v_proj/_0.pt index 0da7047de391bdaa75ae41b9adbeee3100fdac42..c687c2da68b256eacee5f65929e08d2be0bb9023 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.v_proj/_s.pt index 0e2fcf9e7283a027773d18e35b321461cc04bdf5..8f118b2eafaeb67d29b484db8df139db8545444a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.20.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.fc1/_0.pt index d0ef68229fc3b9ce68b111f6ec0a1b044a814acd..6ecdbbf8b6aa3e22869aa3ae9d2a64761bdcad2d 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.fc1/_s.pt index f5e8e523c69c6a937aa7faa0fecb58d2bc349e84..a75a616272ab2c69cacfa53fc91077675e2ed699 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.fc2/_0.pt index 1fea14d6ac63b9b9067922b661e9ade16f4b8d99..55bda095f71655a9a83b43cd56d3887d9fc20881 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.fc2/_s.pt index 091fd66ddf7a2c5337cd60a5667dcdb0d183eb1c..139bf960d648b3826a9f0fc930294b3b33edc288 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.k_proj/_0.pt index 7013cda5642e3e0f035906436edf022ef79f1ba4..c9f8109e33969981fe49991e80a41672c0f56d58 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.k_proj/_s.pt index 67ce3e88d0bc472b5bf05775a777f5aa5a060fb2..514ea525f587ab2e3fe16983d84bb228998324ea 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.out_proj/_0.pt index 08bf661dc76bac08ec7097d1d480c4cacfed1dad..8cda213f8485ee41b8053fd3c71e2ed55889b9f7 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.out_proj/_s.pt index 35acfbb644855b65ba7700e52ea03ad3656be709..9f511d0268d00e3aa1c9ec3c6b70a8be6af90b86 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.q_proj/_0.pt index 1ad59bdb53868f215a5c17a9f1fbcb03dbe1dd89..5ff9b2bdc52651137f9cc9235915302e22ecbce9 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.q_proj/_s.pt index 0a94b60bf60bf3dde5100c99b2efb14b0840499b..0450d5ac454bd6a8cb0905bf7cc36eca91216b7b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.v_proj/_0.pt index 55aa4e44e001840f86ceedcb448459f8e52104cc..d098cedc5d470146cf4791e8819352df56b05b38 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.v_proj/_s.pt index 2e42b0e570a1c0beeda31bbc17ae87a6926bdf4a..103f79bcf04ed9aecb1ddd00669130eeb6c795bf 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.21.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.fc1/_0.pt index 28bf9311d96965bf8788a71b45ad654a9bd6fb79..c3ec18d74a91683328ca0ff5aabf62134d07c57a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.fc1/_s.pt index a8b06a2a72289555536a899503fc65b961c13972..d4f9757d9773a70f622e3849cf37695e38d7ee9d 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.fc2/_0.pt index 8ed02f9f9969936373332575480a3d54c07407f4..18398e36af656d8c8b556b0223ba54477de2ccbf 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.fc2/_s.pt index d7522f3a2c0ac65079be5bdd25e281b7127f69e2..d3535566872cb4c92a5b103297625ce34177936d 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.k_proj/_0.pt index f21b66ab2dbc539c82ad8a7412d38fe223dd1fa3..c2e93b55ad341044eab2ad391c10bef5cd143f32 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.k_proj/_s.pt index 946417a14f83b6e9a55ade931700f113b89d17fc..d499113f59bf447327908ac6a25e3360d486cd64 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.out_proj/_0.pt index 55a065c8132672f3818f58f0426cab13f799b7fd..3ea9c1b8d7d4dd067a404dcb6b60ae37ab987bd2 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.out_proj/_s.pt index 7c5eb8c90a46e7b8530bad320a23151f45521a3c..4c731d7ba765ad67b4821fcd4c48d1f5ab51ff6a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.q_proj/_0.pt index 13fbba79d7abd05173f05d3dc384cf671aa6dc7e..f1f636743394afbc9a89613f6eea49dd43f88fa9 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.q_proj/_s.pt index 3c171da889bcee87e64e844ad95c0f070d907a08..5f70884112f6efe3043a205f0877b20be8291393 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.v_proj/_0.pt index 281cca433ee2c5dc5b77ff1672ee7fc8f87e6281..fdb49381541d5aec5a3dcfb862d9f4859cf5b672 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.v_proj/_s.pt index c7a7fc4e744e345254ae8c4e161632a6d20632c6..9b10591fa55db2603c3ff3b172dbf741038b36d1 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.22.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.fc1/_0.pt index e0adfe1e0f40a254fd67d26f4b304b1d118340cc..56944ceeda9874d5e091106618422c855466a133 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.fc1/_s.pt index 6f11022c4132e5f141f051762feaa44b28caec53..8075e7aad2aa94502ea0ab38bc620fc48bc2a578 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.fc2/_0.pt index 51fbdc25a6ed6664deea56600ba2eff8dce7b947..8b388cd4fc14494323042c0653e678ef1e610d80 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.fc2/_s.pt index 31a251c0322698415f3152e0d14428f53cb7c49e..b55b9006d83c649d4b63dcf26e182243d06e9309 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.k_proj/_0.pt index 74dbb7ab7ec83b231aa065b27bbf9a58d1b02b7a..45b86a2411ab25863b8c10f44cf1129c761de36f 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.k_proj/_s.pt index 3a5bc06050d6a401d196b9df19971962d4f6767a..6ad6af84d2c1d02e6d00d9085627254c56e4afca 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.out_proj/_0.pt index 37f4eb34155e7f649acdccd9ae5da22608ea38ee..c9d0d0775a916e35f425f9bdfb75fa07c9c30ad3 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.out_proj/_s.pt index 416f557929fc4970a4a9e61dc11ba941b566f176..564f248bd0ff095943f4b49f7c2ba7241096b4ee 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.q_proj/_0.pt index 8637ef23d6d9cfdd2a9c8b5ce1471b9c1a41c365..6901652bc01ff4d5f937a45a9664cc7bce2d36d8 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.q_proj/_s.pt index a803f395e98796461f66d12fa3dbf90a2ef7418b..ea4bd08a750ea18cd4c4237e0a97bc47aa3d79cb 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.v_proj/_0.pt index 095bedc6f44c18d564b10dc20a36334edded7c55..0b833d2ebac5d7ad7e98aff88d76f15df5274300 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.v_proj/_s.pt index 1018fc7307496e702d0c12afc8d47ecb54d6ee25..f8053e4f5517ee7fedfd17b0ef08e46723079f95 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.23.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.fc1/_0.pt index 25576c9ff8106106395c2663257a2c50f9ffaff9..9535752482a315a500a2cc6b20384cd90d8a276b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.fc1/_s.pt index 483d0e14c9b22114fb07e4ca147fc51db4e14a64..f99bfaf31b9687f6381cf16a4103fcd35d08809b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.fc2/_0.pt index c4af2a25332e186f33e5b78577eeca60716706d9..d63b2a02cee1e73d746e52967f4a4e1560fac1b1 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.fc2/_s.pt index 690c249cfb6f94e927fd7e421e3cac72f903ca5a..99b8289ad6f9664317409fedaa8b681b90e80a0c 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.k_proj/_0.pt index 69a2c477490b05fac6b2ccac540b3506dd3dd31b..8bf635594ce3204288f41c6a48a7a0b0a3b2ae9a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.k_proj/_s.pt index 229a09eba14622c6817ec6f166507e4714646d6d..4fb28711ecac4eddc3cb11f01d180a1bb198940d 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.out_proj/_0.pt index 28a1d8f1d8735d5c878526b818f3d8be78fb9468..613159f6e8016b6effbc82b617c05fc1700c9078 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.out_proj/_s.pt index 3594a20b5a1e2eb13bbaea58cfaa7ed28d23e009..3abcbc4aba2dc466ba22e4de9445ea709a308b58 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.q_proj/_0.pt index dc59f90759bee6890b1c005755f2b763179d4e0b..931e325f4c395c57089b6944d8a293caa541b959 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.q_proj/_s.pt index 9b4fc0164f194ca77719207818c91bee4992201f..97111d7029819e5ec74e354b6bda9f0fb81b18d0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.v_proj/_0.pt index 2e466f3ef07d8d803e1fd51d455707e63aa64766..97c3e358d136788235cfb17feaa71ff242cd7ef6 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.v_proj/_s.pt index ebd3aaadc01d61a53610083ccf852366afe62bde..2fcdfe3704e43cc31dc0813d30ad2a6cea9192d1 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.3.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.fc1/_0.pt index 5685374b0bee6351dbcafa6569f33a37687d4d0c..f857f46def962e2ddfc809d94912864fd56477c3 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.fc1/_s.pt index c3438ba3f357ff00ebcd2743c94a5ce6b5082873..51462ba07dd78bc5d9496bc8d1c0682c0ab319eb 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.fc2/_0.pt index 66c6cc224cc4b9fee37b764f9dc501e8c63ac631..8840d9ea27c5abf2b7f68d07c07fb6f4352bdadb 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.fc2/_s.pt index 2b3f5b438f2ecf2c7a4a2a12dd55738c0d4c6732..0a6e86c79f68b1e01c76dbc190e27805ea03f8b6 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.k_proj/_0.pt index d388a512f88119a7a0aa4adf06299014be2900e2..d125655297a6e3567173c360ed74065d249b86a6 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.k_proj/_s.pt index ec32b29d8de8bff9e424055afbdd4c4867bfa005..4cc8a3b97315a72179370186d38114c2022821c2 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.out_proj/_0.pt index 7f11ed553bab837b898cf34c952a0d42ad743242..e7de2a2718f53c92ccdfa1f518fd1bb7f749268e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.out_proj/_s.pt index a7609745875ec5a5fef292238a50d10ff5e0a367..cae9ccd9abc1a4f3bcedefc19c9d847fa4a51352 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.q_proj/_0.pt index e63210122347c56c07873a7da490397b636e099d..33e4f033bbc0e1d902901b7cb1f550ea2ea7cb78 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.q_proj/_s.pt index dbe43dbbba75ca183817ab0c5e62c19647548b4d..27e61688f6de1689c45ec2b3786b24efeefe9ebc 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.v_proj/_0.pt index 5a945779bf187a4b4a7f46bf20f3c8b85f3e51e7..03914b0baedd4cec0bf9f0c3fb3c55983b575e47 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.v_proj/_s.pt index cc45e2d4084764a16a62d3b112db423e7ea32916..d993b112bb25ecd2f66d6996a55b065cdb72cfa7 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.4.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.fc1/_0.pt index 1966afacab4dc63e6a153215bddc1d57113ffaf1..59f0c5d18f0efdd06193e3af50c8666cfbfc4953 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.fc1/_s.pt index cb72ef37fab152e9c3e51e36432520d42a5e3e48..db252ca6885fd8829e6227894155f3aae9f1af1e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.fc2/_0.pt index 664410a06aba2a16738b2d411da422eabf94fdc6..36556ba8e4dbd24a73231c30cf9066fd9a3a5b91 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.fc2/_s.pt index 40f6b01e7ea54b35cf08f81fe38fb371374198e4..10002dd29dc92d7594dfe85caa20c2baef596594 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.k_proj/_0.pt index 5554e7177932577335497783b72fe4d4cb7bf2a9..b2f8dec82bf0325f9d4a550616609424a50bdda5 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.k_proj/_s.pt index b4ab4635706fbadfc977c6a154acb084a8254755..f909fffd80dee279b51fa56882127f19c528d8ad 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.out_proj/_0.pt index d959c0ba52c3847df560c9bff6ff1f29f672df6d..476a67404d9b6727857e4e5f5d9e35493f2d6ed2 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.out_proj/_s.pt index 887932d5a70698217fa2bfbe353ee03d677825fa..ef508928c5aece0d64ae2e7f106e2366529f58db 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.q_proj/_0.pt index 622daa75dc5c4b93c3ee6093094d8e816562434c..47bc240cb652c2e2af92ec5fcaa755cea128ddb3 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.q_proj/_s.pt index 92491c83332eecbdb954dbf9ff7e647d678f2f13..e050498fa9c53a5701c7b7b6a23f37af9e91075f 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.v_proj/_0.pt index 50cd2c76a2508176dab0d51f239ffb426bd2667a..40e16dc8ae2731aeff6f71f8619ba3efd3d632b1 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.v_proj/_s.pt index d385bdb298447e3d78b08a10d7187dbb45cc2b35..121831d16f5d150935fa1303163da0d7b9d6d5a6 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.5.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.fc1/_0.pt index ade82b7f9ec3a8819ab1488dd16d08450bad8eaf..000c8911cafd27015b2c8867293a74f63f33e2ab 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.fc1/_s.pt index d937ee92b1cbdb3eab650f49b172036e8ebf1519..208ff6174e6305245bb53d0d5e4b2180beaa27fb 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.fc2/_0.pt index 120d85cbd5bc8ea8a2567e4493c5da8dce4348bf..9b7944d187eb3ed451ac682263a98dcbc693a8a4 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.fc2/_s.pt index d646e3def636825a0560f1b931c2c6955159cc58..823223a4abc4c74a4de2b0759580d2dc9f15384e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.k_proj/_0.pt index ea7544c3eae012cea64a891accc280119e033139..17963c5725535ba826f0d9a4d4384943e695a50d 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.k_proj/_s.pt index 6d37cfaa425338dbad1e58c22a56a1ff53aba8a7..4150a8a10e74bf4243c1ac66f83fe818eb6dd8c3 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.out_proj/_0.pt index 904b12a1ed2db1f4fb68b23b8f3fc62b921d79de..7b52444c27fd2aadc8f1f8b101a1146de7548e38 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.out_proj/_s.pt index ab89ea36cbbf6dfdacb8090659ede6b10831a252..d327f38539c4f5e0463db4c03543fee274e3ea40 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.q_proj/_0.pt index 43235056c047b21ba5577d3440e116b06bf3a0f5..8097e0b29ab7ff2af3310bf55fedeb4190bca753 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.q_proj/_s.pt index 9daa8ee350bf28b77d8a3625d17a0fe9cb2a5e47..bb554b5861c5c36054b080d00f1a675905aed545 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.v_proj/_0.pt index ff7010ba6af6a6b9dade19137855a06a45455de9..58553934c846f2eb83df7c5a4dc6646126f83cea 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.v_proj/_s.pt index 84cb036089c7aa6a65f5073b82baae427c3d5414..79c5a55f44a90b2a3e7bfd46070cb809e5d32893 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.6.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.fc1/_0.pt index 6e279499306524a8b04303818fdb045755ba0e63..d414512eb0571289d94f523ab23cd5439428621c 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.fc1/_s.pt index 04fe16ede778b296e48d1271ffb103dea1a2bbb0..7a40ff2bdf1262fffd7edc5609b517184ecd1d3a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.fc2/_0.pt index a3205532dfd7516a327173c3df36816aa77566ff..ff7bb7cd0a1c649abeb9a907987db8762f3c5b33 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.fc2/_s.pt index 9305cd8a9a8b0eb17ef8d9b0dce72a54b4066fa6..c9c2cecd97962cc958408e67fce98d5748dea061 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.k_proj/_0.pt index 4cc6eafae20be26b3587a161a15bec784ce0cb90..ee670ef48c4d940b2c83e4c9c54e6f2f542e0c84 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.k_proj/_s.pt index dd908ba5843db7f4241485d0bedba3286d4adf73..a98943ab46faf6a00a34089f1e1d0b01520a375e 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.out_proj/_0.pt index 42aef4220ecbf6846df2723c9d585a980ba56b08..e1e50ccee4e7cc29543e30acf36b8a6e4982a054 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.out_proj/_s.pt index 827f36c668e39c680db6f5e32ca6abbd4630fcf2..3d5807434dfbc4caaa7b1cf267879ee5473f9aa1 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.q_proj/_0.pt index ac858f621eee20851a3d267432b7733a8e57adc5..510993a03cdd9b72e252436de0eb052b1ebfaec6 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.q_proj/_s.pt index ce60090f677688ee2e2fdae52c665ed82f622bc7..3f88f1312ea3b130d30129567c2c771fb270248f 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.v_proj/_0.pt index b1500ae3e0e0be292e33f4d960683a652ac29f3d..bb26e360ca06497bdd63d1a61974a39634eb982b 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.v_proj/_s.pt index 2915f86720481ce5b1b14bd6745203a444d9186e..5581a0a0cc51cbe19b3fe57590e5f63451e79037 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.7.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.fc1/_0.pt index 9c2e70576b950ae6b3cbbc7a1935946c8868ff95..7c88c1e057f0f2b2f067987540b7f96aaad4d3ef 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.fc1/_s.pt index ab17159fd7315250e65446cc83894b7d6108aec2..11a5186659e0b244f187f3dbd0f74329001f3bce 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.fc2/_0.pt index ecaa8f40573540e2209879fdaa63ace7fdf564a5..abdcc39b445f6fe3dd00ad193ed76c16eb7647ca 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.fc2/_s.pt index 94718822942ffb5dc39107e33423ed76a930fe85..4e21014faca9c40eaf9be3aae39740be7692c075 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.k_proj/_0.pt index fa62f12d9e52716f4fec5717a1d19b68bbd4a18d..36b1d2f0acfeae40366b748ca1be9b93380d0517 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.k_proj/_s.pt index edbfb817bf48848420eb32ad6625de40c6e90d66..74b07329aff6d9ecf89e1e21d3afe25ae05c1934 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.out_proj/_0.pt index 885dd7fa7c8e054c70b5ca1c54c84ba16b85b848..82742dc40812d5df5d358fb310d6d89290c06c48 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.out_proj/_s.pt index 1a598438ac2ac87031c616aec0a62d1e7c830a78..8209a62eb098c724621b43787f550ea128de42ef 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.q_proj/_0.pt index 9dc35529c3e99186b1b94b04ee5e970ece0704ce..ed87196a2268b13d1992f64c6e0e3e554328be09 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.q_proj/_s.pt index baae087a1f8a4ea4bcf2f0b6ebdcd96c3f21878f..89e672ddd5fe0f8f95ce22f3908192d7b922d961 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.v_proj/_0.pt index 8f0a39bc50b6c2eb513859620828982402b5eeaa..7ff34f762b01a2ec816c227f603393eca895cae7 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.v_proj/_s.pt index 0908ef3aa0dc15e4c741c69d0694f8cb9a054d26..07bb84583b44509820606cd9af89aeb50d65b1ce 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.8.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.fc1/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.fc1/_0.pt index 07e4c74bf1b61bc98a00951c5884db91be2d11f3..d1fe45dc4bebd067912f98ef63fe24d582704199 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.fc1/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.fc1/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.fc1/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.fc1/_s.pt index 91c314bd6360995f508c2ce44aa58fb6de36872d..ef9151d4524866956b0a531d0ff954982e12f377 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.fc1/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.fc1/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.fc2/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.fc2/_0.pt index d4a5f92c8b8a009d9ec5ee39139834c867fe6b3c..52847cc13d534913ee55d8578d24f912276513ac 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.fc2/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.fc2/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.fc2/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.fc2/_s.pt index b7e854e5f0846492832f91f5890e1ad90b2f09fa..d15554189bdf9e2b9fdb46b7f88144227c78a026 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.fc2/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.fc2/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.k_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.k_proj/_0.pt index 1b04bd7e308cbd2bf259d25fd3b3ecf9a46868dd..d92724ce2a4949b9dd16593455cd0d69ccdc3196 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.k_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.k_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.k_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.k_proj/_s.pt index ae85deccbc9719633874237f2365fcaa2bac1d47..75191ba0cfdbce8eda10b05929c0d5a02b02d43d 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.k_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.k_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.out_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.out_proj/_0.pt index 5ecf466ed0d1421f973b863f7766b82a72525029..3bd462fd3865bb8b733472c99a88dd91cd780cc0 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.out_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.out_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.out_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.out_proj/_s.pt index a65a34d2cd13bef2a4f9630886fc46f18251c591..1579b5944c4e0fe9c97972090bac8266c3ec36a1 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.out_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.out_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.q_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.q_proj/_0.pt index a4607c5c32caddf22d3daabbae4cfa28427bf579..fb01405fb1c2eb126f4a56eddb5185ee0bb25e94 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.q_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.q_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.q_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.q_proj/_s.pt index b35cf25eeeb737460d6728064bf4fb879f9b7978..a8ca29df7ed0145c0e773004630f72a0e5162c57 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.q_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.q_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.v_proj/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.v_proj/_0.pt index 101f636ae31e70a1e90d4b55b4ae6e4c7079ccdf..efc12ab754ff66c14d96bb92245b4b36d6cf2fda 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.v_proj/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.v_proj/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.v_proj/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.v_proj/_s.pt index 8a734482261b2bdf7433daeefa66131fe945aa13..eec58ddd7a148d028e5c5dbc66f0ad841b9c1bc5 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.v_proj/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.layers.9.self_attn.v_proj/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.project_in/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.project_in/_0.pt index 07b2e0c730fc26b06b83963925b24c830db91722..267b92cd06092a5a3dbf7d1dcd4823ed5e73fb84 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.project_in/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.project_in/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.project_in/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.project_in/_s.pt index a4c43d3d5077458ac34024798c1570c06ba9e135..fb8bad94f64067dfeb18f151313f0c8f9b29081a 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.project_in/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.project_in/_s.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.project_out/_0.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.project_out/_0.pt index c867b2ba09ea81a74605b4badf2dc4e464f87a19..9c8a150e2d79c2a7d979c51059a6c32142ba0d05 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.project_out/_0.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.project_out/_0.pt differ diff --git a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.project_out/_s.pt b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.project_out/_s.pt index d94f70c7352407384cd18e6c72f2faab9075e283..aa13e05b4e2f051c6e3b4a7185b4fb3fb4227e47 100644 Binary files a/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.project_out/_s.pt and b/params/opt/4/rnd_search_t_cos/comb_1/trained/model.decoder.project_out/_s.pt differ