diff --git "a/ndarray-cache.json" "b/ndarray-cache.json" new file mode 100644--- /dev/null +++ "b/ndarray-cache.json" @@ -0,0 +1,5227 @@ +{ + "metadata": { + "ParamSize": 455, + "ParamBytes": 1740533760.0, + "BitsPerParam": 5.009299935348043 + }, + "records": [ + { + "dataPath": "params_shard_0.bin", + "format": "raw-shard", + "nbytes": 65536000, + "records": [ + { + "name": "lm_head.linear.q_weight", + "shape": [ + 51200, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 65536000, + "byteOffset": 0 + } + ], + "md5sum": "2aa30ec63aaf1faace62a6544c9a06dd" + }, + { + "dataPath": "params_shard_1.bin", + "format": "raw-shard", + "nbytes": 23091200, + "records": [ + { + "name": "lm_head.linear.bias", + "shape": [ + 51200 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 102400, + "byteOffset": 0 + }, + { + "name": "lm_head.linear.q_scale", + "shape": [ + 51200, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 8192000, + "byteOffset": 102400 + }, + { + "name": "lm_head.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 8294400 + }, + { + "name": "lm_head.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 8299520 + }, + { + "name": "transformer.h.30.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 8304640 + }, + { + "name": "transformer.h.30.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 8320000 + }, + { + "name": "transformer.h.30.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 18150400 + }, + { + "name": "transformer.h.30.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 19379200 + }, + { + "name": "transformer.h.30.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 19384320 + }, + { + "name": "transformer.h.30.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 22661120 + }, + { + "name": "transformer.h.30.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 23070720 + } + ], + "md5sum": "bf1492c3f22f466e5f70dba01a757eae" + }, + { + "dataPath": "params_shard_2.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.30.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.30.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.30.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.30.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.30.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.31.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.31.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.31.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "c9690715109fca285b5615510f5ff0dc" + }, + { + "dataPath": "params_shard_3.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.31.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.31.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.31.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.31.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.31.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.31.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.31.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.31.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.31.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "debe276ee49ce4eb8824ef4f5158f6b5" + }, + { + "dataPath": "params_shard_4.bin", + "format": "raw-shard", + "nbytes": 65536000, + "records": [ + { + "name": "transformer.embd.q_weight", + "shape": [ + 51200, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 65536000, + "byteOffset": 0 + } + ], + "md5sum": "cd2ab4b418c2cd5fbd240da8d981f1d8" + }, + { + "dataPath": "params_shard_5.bin", + "format": "raw-shard", + "nbytes": 32793600, + "records": [ + { + "name": "transformer.h.31.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.31.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.embd.q_scale", + "shape": [ + 51200, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 8192000, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.0.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 22937600 + }, + { + "name": "transformer.h.0.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 22942720 + }, + { + "name": "transformer.h.0.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 22947840 + }, + { + "name": "transformer.h.0.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 22963200 + } + ], + "md5sum": "2e1c938ad8df4a077b56a701705cb5ac" + }, + { + "dataPath": "params_shard_6.bin", + "format": "raw-shard", + "nbytes": 32798720, + "records": [ + { + "name": "transformer.h.0.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 0 + }, + { + "name": "transformer.h.0.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 1228800 + }, + { + "name": "transformer.h.0.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 1233920 + }, + { + "name": "transformer.h.0.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 4510720 + }, + { + "name": "transformer.h.0.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 4920320 + }, + { + "name": "transformer.h.0.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 4940800 + }, + { + "name": "transformer.h.0.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 18048000 + }, + { + "name": "transformer.h.0.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 19686400 + }, + { + "name": "transformer.h.0.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 19691520 + } + ], + "md5sum": "2f3350c7c4a6e6f99c9ce6cb56e600e4" + }, + { + "dataPath": "params_shard_7.bin", + "format": "raw-shard", + "nbytes": 31185920, + "records": [ + { + "name": "transformer.h.0.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 0 + }, + { + "name": "transformer.h.1.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 1638400 + }, + { + "name": "transformer.h.1.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 1643520 + }, + { + "name": "transformer.h.1.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 1648640 + }, + { + "name": "transformer.h.1.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 1664000 + }, + { + "name": "transformer.h.1.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 11494400 + }, + { + "name": "transformer.h.1.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 12723200 + }, + { + "name": "transformer.h.1.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 12728320 + }, + { + "name": "transformer.h.1.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 16005120 + }, + { + "name": "transformer.h.1.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 16414720 + }, + { + "name": "transformer.h.1.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 16435200 + }, + { + "name": "transformer.h.1.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 29542400 + }, + { + "name": "transformer.h.1.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 31180800 + } + ], + "md5sum": "48f740afe8713ce24c2d0d4c4692323c" + }, + { + "dataPath": "params_shard_8.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.1.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.1.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.10.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.10.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.10.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 14755840 + }, + { + "name": "transformer.h.10.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.10.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.10.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25830400 + }, + { + "name": "transformer.h.10.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25835520 + }, + { + "name": "transformer.h.10.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.10.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "61e253a68cf297afbe1e7d67f7743ab7" + }, + { + "dataPath": "params_shard_9.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.10.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.10.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.10.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.10.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.10.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.11.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.11.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.11.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "5850498b40179f5b4665ac73e0b69d84" + }, + { + "dataPath": "params_shard_10.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.11.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.11.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.11.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.11.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.11.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.11.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.11.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.11.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.11.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "52183debc2f472aea1474a23ef2b0cf7" + }, + { + "dataPath": "params_shard_11.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.11.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.11.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.12.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.12.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.12.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 14755840 + }, + { + "name": "transformer.h.12.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.12.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.12.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25830400 + }, + { + "name": "transformer.h.12.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25835520 + }, + { + "name": "transformer.h.12.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.12.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "b517671b76bcc12c5e6c40ea1e97befc" + }, + { + "dataPath": "params_shard_12.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.12.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.12.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.12.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.12.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.12.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.13.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.13.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.13.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "ca0cd79d7891f5cde81fdbb3eda9b6ba" + }, + { + "dataPath": "params_shard_13.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.13.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.13.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.13.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.13.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.13.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.13.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.13.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.13.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.13.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "f229fc130442264ae9c8ca6ddc6d056e" + }, + { + "dataPath": "params_shard_14.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.13.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.13.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.14.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.14.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.14.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 14755840 + }, + { + "name": "transformer.h.14.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.14.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.14.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25830400 + }, + { + "name": "transformer.h.14.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25835520 + }, + { + "name": "transformer.h.14.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.14.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "ff565c15a46831676d433d65daec3de4" + }, + { + "dataPath": "params_shard_15.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.14.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.14.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.14.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.14.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.14.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.15.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.15.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.15.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "c38bfbab564fc53b317ba9c361edba5d" + }, + { + "dataPath": "params_shard_16.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.15.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.15.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.15.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.15.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.15.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.15.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.15.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.15.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.15.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "725fe5c31a7e17d897901149c9026440" + }, + { + "dataPath": "params_shard_17.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.15.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.15.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.16.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.16.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.16.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 14755840 + }, + { + "name": "transformer.h.16.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.16.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.16.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25830400 + }, + { + "name": "transformer.h.16.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25835520 + }, + { + "name": "transformer.h.16.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.16.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "b563d9b94a7b9bd50cc72d500de8972e" + }, + { + "dataPath": "params_shard_18.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.16.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.16.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.16.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.16.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.16.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.17.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.17.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.17.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "42055bf81bf9d29f387f82c06b48541b" + }, + { + "dataPath": "params_shard_19.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.17.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.17.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.17.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.17.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.17.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.17.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.17.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.17.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.17.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "d915687d75b0b9d1a2073068492d6a11" + }, + { + "dataPath": "params_shard_20.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.17.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.17.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.18.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.18.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.18.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 14755840 + }, + { + "name": "transformer.h.18.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.18.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.18.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25830400 + }, + { + "name": "transformer.h.18.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25835520 + }, + { + "name": "transformer.h.18.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.18.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "10374705213e6ca8f5b07d81784cd972" + }, + { + "dataPath": "params_shard_21.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.18.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.18.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.18.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.18.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.18.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.19.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.19.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.19.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "2c0679dddcd9d6234277cccc64332f29" + }, + { + "dataPath": "params_shard_22.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.19.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.19.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.19.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.19.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.19.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.19.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.19.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.19.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.19.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "08c3667f7832926907754b0c03e1b730" + }, + { + "dataPath": "params_shard_23.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.19.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.19.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.2.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.2.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.2.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 14755840 + }, + { + "name": "transformer.h.2.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.2.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.2.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25830400 + }, + { + "name": "transformer.h.2.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25835520 + }, + { + "name": "transformer.h.2.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.2.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "1b7bb8d1b13ff72affea458cd0c194c2" + }, + { + "dataPath": "params_shard_24.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.2.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.2.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.2.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.2.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.2.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.20.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.20.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.20.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "806297d91bed14d687c298c5079b07a6" + }, + { + "dataPath": "params_shard_25.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.20.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.20.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.20.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.20.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.20.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.20.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.20.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.20.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.20.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "a7d231a13483fcc565d30ed028f89f0e" + }, + { + "dataPath": "params_shard_26.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.20.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.20.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.21.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.21.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.21.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 14755840 + }, + { + "name": "transformer.h.21.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.21.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.21.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25830400 + }, + { + "name": "transformer.h.21.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25835520 + }, + { + "name": "transformer.h.21.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.21.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "60742fb790f2d48490ca7d3244769ef5" + }, + { + "dataPath": "params_shard_27.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.21.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.21.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.21.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.21.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.21.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.22.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.22.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.22.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "dc82ce3796371aa1228e20b14eabe9d7" + }, + { + "dataPath": "params_shard_28.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.22.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.22.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.22.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.22.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.22.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.22.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.22.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.22.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.22.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "d7679ca4bdfa87309149819f27b6dc4f" + }, + { + "dataPath": "params_shard_29.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.22.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.22.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.23.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.23.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.23.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 14755840 + }, + { + "name": "transformer.h.23.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.23.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.23.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25830400 + }, + { + "name": "transformer.h.23.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25835520 + }, + { + "name": "transformer.h.23.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.23.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "bc4a32a329a689393587004dca333e56" + }, + { + "dataPath": "params_shard_30.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.23.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.23.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.23.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.23.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.23.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.24.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.24.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.24.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "9e2040a7b5aac397dec3b4875705bc47" + }, + { + "dataPath": "params_shard_31.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.24.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.24.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.24.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.24.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.24.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.24.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.24.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.24.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.24.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "7b4686167336d17eb561bb7345137a2a" + }, + { + "dataPath": "params_shard_32.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.24.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.24.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.25.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.25.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.25.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 14755840 + }, + { + "name": "transformer.h.25.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.25.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.25.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25830400 + }, + { + "name": "transformer.h.25.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25835520 + }, + { + "name": "transformer.h.25.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.25.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "3bb4b940a32be592af167952295f36bc" + }, + { + "dataPath": "params_shard_33.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.25.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.25.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.25.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.25.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.25.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.26.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.26.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.26.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "1aeaea0b8274641081213cd7f8d3af4f" + }, + { + "dataPath": "params_shard_34.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.26.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.26.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.26.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.26.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.26.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.26.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.26.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.26.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.26.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "f22a5372b560fae12c9c304a1032921f" + }, + { + "dataPath": "params_shard_35.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.26.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.26.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.27.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.27.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.27.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 14755840 + }, + { + "name": "transformer.h.27.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.27.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.27.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25830400 + }, + { + "name": "transformer.h.27.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25835520 + }, + { + "name": "transformer.h.27.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.27.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "241e687610b6f9d9081f41ccbc71fe56" + }, + { + "dataPath": "params_shard_36.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.27.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.27.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.27.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.27.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.27.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.28.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.28.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.28.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "bb6372f3a8e2f5a6353deb58d0babf0f" + }, + { + "dataPath": "params_shard_37.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.28.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.28.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.28.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.28.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.28.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.28.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.28.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.28.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.28.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "1307087ad4d73ceac393bade117fb3dd" + }, + { + "dataPath": "params_shard_38.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.28.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.28.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.29.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.29.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.29.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 14755840 + }, + { + "name": "transformer.h.29.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.29.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.29.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25830400 + }, + { + "name": "transformer.h.29.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25835520 + }, + { + "name": "transformer.h.29.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.29.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "22a49734ac7361630d8586f7503ec1fe" + }, + { + "dataPath": "params_shard_39.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.29.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.29.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.29.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.29.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.29.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.3.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.3.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.3.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "d3f7c3921d5e2a68b181672c8a0c2f25" + }, + { + "dataPath": "params_shard_40.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.3.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.3.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.3.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.3.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.3.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.3.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.3.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.3.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.3.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "8996dcb8dd4330ea15ab360a65dd75ee" + }, + { + "dataPath": "params_shard_41.bin", + "format": "raw-shard", + "nbytes": 29552640, + "records": [ + { + "name": "transformer.h.3.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.3.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.30.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.30.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.4.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "transformer.h.4.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "transformer.h.4.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 14766080 + }, + { + "name": "transformer.h.4.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14781440 + }, + { + "name": "transformer.h.4.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24611840 + }, + { + "name": "transformer.h.4.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25840640 + }, + { + "name": "transformer.h.4.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25845760 + }, + { + "name": "transformer.h.4.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29122560 + }, + { + "name": "transformer.h.4.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29532160 + } + ], + "md5sum": "62ff08bc55e176726cdd96c6d3bf3b8b" + }, + { + "dataPath": "params_shard_42.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.4.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.4.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.4.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.4.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.4.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.5.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.5.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.5.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "5801dafb49c0432492e930a8b757da21" + }, + { + "dataPath": "params_shard_43.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.5.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.5.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.5.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.5.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.5.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.5.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.5.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.5.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.5.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "c3fc5169a9e07ed1acc959d0054267d5" + }, + { + "dataPath": "params_shard_44.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.5.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.5.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.6.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.6.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.6.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 14755840 + }, + { + "name": "transformer.h.6.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.6.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.6.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25830400 + }, + { + "name": "transformer.h.6.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25835520 + }, + { + "name": "transformer.h.6.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.6.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "8dcebcdd42a9b732be62d62f140d5a52" + }, + { + "dataPath": "params_shard_45.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.6.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.6.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.6.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.6.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.6.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.7.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.7.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.7.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "39bc14f05d2604911e9709b9c047247d" + }, + { + "dataPath": "params_shard_46.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.7.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.7.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.7.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.7.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.7.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.7.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.7.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.7.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.7.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "15e515f6f48182c88eb8c609545f39ec" + }, + { + "dataPath": "params_shard_47.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.7.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.7.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.8.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.8.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.8.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 14755840 + }, + { + "name": "transformer.h.8.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.8.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.8.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25830400 + }, + { + "name": "transformer.h.8.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25835520 + }, + { + "name": "transformer.h.8.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.8.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "6b3da487775f01bd94570cd5410ac94c" + }, + { + "dataPath": "params_shard_48.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.8.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.8.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.8.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.8.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.8.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.9.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.9.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.9.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "5a53a7517652b44334e1be290f59714d" + }, + { + "dataPath": "params_shard_49.bin", + "format": "raw-shard", + "nbytes": 29521920, + "records": [ + { + "name": "transformer.h.9.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.9.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.9.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.9.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.9.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.9.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.9.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14771200 + }, + { + "name": "transformer.h.9.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.9.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "edc2ebc3e2bdde118bc55533c9e8f3d2" + }, + { + "dataPath": "params_shard_50.bin", + "format": "raw-shard", + "nbytes": 14745600, + "records": [ + { + "name": "transformer.h.9.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.9.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float32", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + } + ], + "md5sum": "26ae37b226e76de82fb04ed4abf1328f" + } + ] +} \ No newline at end of file