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": 1564948480.0, + "BitsPerParam": 4.503961083574167 + }, + "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": "9ca9bebcc171f92526eeb3f53a4f82d2" + }, + { + "dataPath": "params_shard_1.bin", + "format": "raw-shard", + "nbytes": 23091200, + "records": [ + { + "name": "lm_head.linear.bias", + "shape": [ + 51200 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 102400, + "byteOffset": 0 + }, + { + "name": "lm_head.linear.q_scale", + "shape": [ + 51200, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 8192000, + "byteOffset": 102400 + }, + { + "name": "lm_head.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 8294400 + }, + { + "name": "lm_head.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 8299520 + }, + { + "name": "transformer.h.30.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 18150400 + }, + { + "name": "transformer.h.30.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 22661120 + }, + { + "name": "transformer.h.30.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 23070720 + } + ], + "md5sum": "03237478c3c98239f24ae9905c725036" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.30.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.31.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.31.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.31.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "9cccf6fd7d090ad2c84234b90c7b72f6" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.31.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.31.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.31.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "b9c419b02d0f38c1071b6226b69e0f19" + }, + { + "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": "478ad7ae97b8be71515834fd36010ef4" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.embd.q_scale", + "shape": [ + 51200, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 8192000, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.0.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 22937600 + }, + { + "name": "transformer.h.0.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 22942720 + }, + { + "name": "transformer.h.0.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "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": "d655f454552f5c29e8d4abeba39c54c4" + }, + { + "dataPath": "params_shard_6.bin", + "format": "raw-shard", + "nbytes": 32798720, + "records": [ + { + "name": "transformer.h.0.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 0 + }, + { + "name": "transformer.h.0.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 4510720 + }, + { + "name": "transformer.h.0.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 18048000 + }, + { + "name": "transformer.h.0.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "e7455684cb15a692c378bd37560330af" + }, + { + "dataPath": "params_shard_7.bin", + "format": "raw-shard", + "nbytes": 31185920, + "records": [ + { + "name": "transformer.h.0.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 0 + }, + { + "name": "transformer.h.1.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 1638400 + }, + { + "name": "transformer.h.1.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 1643520 + }, + { + "name": "transformer.h.1.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 11494400 + }, + { + "name": "transformer.h.1.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 16005120 + }, + { + "name": "transformer.h.1.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 29542400 + }, + { + "name": "transformer.h.1.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 31180800 + } + ], + "md5sum": "a43ecda9b81b6bb0153151fad2487b7f" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.10.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.10.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.10.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.10.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.10.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "cc3eb9417c02089e2f2d2ab3bc41174e" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.10.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.11.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.11.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.11.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "ded2f665bc60b32ff2e96cbe7be0b80e" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.11.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.11.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.11.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "21243ab7239a153975029e7025c96d06" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.12.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.12.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.12.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.12.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.12.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "21788e365b4d22e6ff68f1834f4636d0" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.12.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.13.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.13.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.13.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "a878f81a129292a411a8f534d2145b57" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.13.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.13.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.13.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "cb1d8d68a0c5ec351a8d0446f401f90e" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.14.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.14.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.14.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.14.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.14.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "d22ebf2618e8270834d254e9e042c468" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.14.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.15.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.15.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.15.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "deacf0cb0b511a97c9fc429ee970c504" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.15.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.15.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.15.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "750f8c8a54b79c37056db408e60ff0d7" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.16.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.16.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.16.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.16.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.16.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "393201fe4ab981fd3e9733964cdbc178" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.16.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.17.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.17.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.17.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "9651d92fa46d4e6f554dfa4293e58b51" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.17.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.17.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.17.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "a6b0124b39c0b338a8723487d2eca8f1" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.18.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.18.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.18.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.18.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.18.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "dd075149ec56b2ab9beb42c002da64eb" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.18.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.19.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.19.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.19.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "3ec2d8d68bfe55c1e79e01add1c6d60f" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.19.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.19.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.19.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "367b5a6ea7850cdece9ef04af15fdd04" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.2.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.2.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.2.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.2.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.2.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "1da1b2fc6b5bc15a8213ecd740183826" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.2.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.20.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.20.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.20.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "508b37d66e6e62a42a9129c2db4c16b0" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.20.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.20.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.20.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "857014e7493bab928e791d720334677b" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.21.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.21.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.21.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.21.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.21.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "3f70587070147c2e95968387b8a665b2" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.21.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.22.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.22.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.22.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "0708ef292f67c9b926a6bbd5e733344c" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.22.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.22.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.22.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "c6c7a3a3e716a15762cdb02899192795" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.23.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.23.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.23.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.23.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.23.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "f9fc330e4f4e61c8df83c5b63f86cfb0" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.23.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.24.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.24.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.24.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "d95a99d1c186f99468e69e143c0a3a15" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.24.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.24.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.24.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "9ca703bf92090c8ddd000b00c0d82d88" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.25.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.25.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.25.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.25.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.25.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "341320fef0e63ad01414c03739394e83" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.25.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.26.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.26.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.26.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "a92aa48713cbae692d1d00d1de859ca9" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.26.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.26.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.26.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "bdfcac45992ec28208360e07ad970fa6" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.27.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.27.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.27.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.27.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.27.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "5b0b329921ccdf63b5669e394da3e396" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.27.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.28.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.28.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.28.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "faa927598370b5237a159e0288c34a29" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.28.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.28.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.28.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "c97097e15ae00eb7bf7206ffc15b51f6" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.29.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.29.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.29.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.29.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.29.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "3a0e996a71942b36d19298271dce974e" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.29.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.3.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.3.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.3.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "8a4a9663690232b14277ec2e3304b991" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.3.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.3.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.3.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "c06197020e01f89789d300ad93899d36" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.30.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.30.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.4.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "transformer.h.4.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "transformer.h.4.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24611840 + }, + { + "name": "transformer.h.4.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29122560 + }, + { + "name": "transformer.h.4.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29532160 + } + ], + "md5sum": "14bcf0af3e77c91f6f4ea26ffe3b26fb" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.4.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.5.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.5.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.5.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "e902b5ae0ea02551bf1306429e604850" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.5.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.5.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.5.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "ad4b42d607e349ca0c2bdce94cf493ad" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.6.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.6.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.6.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.6.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.6.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "22e1775804ab9f448f970371fcedd532" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.6.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.7.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.7.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.7.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "271a71bf693a77aa29f84b2ccaa2dae3" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.7.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.7.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.7.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "011873b7282ea72e5db38fd71c5c3667" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.8.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.8.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.8.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "transformer.h.8.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29112320 + }, + { + "name": "transformer.h.8.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "2532d2ba41fd199fb974980e92dcb68d" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.8.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.9.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.9.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.9.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 29506560 + } + ], + "md5sum": "a2922b00f26b65a91836464bd25f0a36" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.9.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14341120 + }, + { + "name": "transformer.h.9.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27878400 + }, + { + "name": "transformer.h.9.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + } + ], + "md5sum": "d20202d21c779c39e26a7e5001a13f00" + }, + { + "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": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + } + ], + "md5sum": "42fc9b09f47b02d874aa2a0ca1bb102e" + } + ] +} \ No newline at end of file