diff --git "a/ndarray-cache.json" "b/ndarray-cache.json" new file mode 100644--- /dev/null +++ "b/ndarray-cache.json" @@ -0,0 +1,4295 @@ +{ + "metadata": { + "ParamSize": 392, + "ParamBytes": 690311168.0, + "BitsPerParam": 4.505168401162159 + }, + "records": [ + { + "dataPath": "params_shard_0.bin", + "format": "raw-shard", + "nbytes": 50462720, + "records": [ + { + "name": "transformer.wte.q_weight", + "shape": [ + 49280, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 50462720, + "byteOffset": 0 + } + ], + "md5sum": "37752cf9591772a22574b2b1fcb29a93" + }, + { + "dataPath": "params_shard_1.bin", + "format": "raw-shard", + "nbytes": 32608768, + "records": [ + { + "name": "transformer.wte.q_scale", + "shape": [ + 49280, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6307840, + "byteOffset": 0 + }, + { + "name": "transformer.wpe.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 6307840 + }, + { + "name": "transformer.wpe.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 8404992 + }, + { + "name": "transformer.h.0.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 8667136 + }, + { + "name": "transformer.h.0.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 8671232 + }, + { + "name": "transformer.h.0.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 8675328 + }, + { + "name": "transformer.h.0.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 11034624 + }, + { + "name": "transformer.h.0.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 11329536 + }, + { + "name": "transformer.h.0.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 11334144 + }, + { + "name": "transformer.h.0.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 13431296 + }, + { + "name": "transformer.h.0.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 13693440 + }, + { + "name": "transformer.h.0.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 13697536 + }, + { + "name": "transformer.h.0.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 13701632 + }, + { + "name": "transformer.h.0.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 13705728 + }, + { + "name": "transformer.h.0.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 22094336 + }, + { + "name": "transformer.h.0.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 23142912 + }, + { + "name": "transformer.h.0.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23159296 + }, + { + "name": "transformer.h.0.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 31547904 + }, + { + "name": "transformer.h.0.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 32596480 + }, + { + "name": "transformer.h.1.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 32600576 + }, + { + "name": "transformer.h.1.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 32604672 + } + ], + "md5sum": "e15870fe91891b563bec51bcddbefd35" + }, + { + "dataPath": "params_shard_2.bin", + "format": "raw-shard", + "nbytes": 28963840, + "records": [ + { + "name": "transformer.h.1.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 0 + }, + { + "name": "transformer.h.1.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 2359296 + }, + { + "name": "transformer.h.1.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "transformer.h.1.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 2658816 + }, + { + "name": "transformer.h.1.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 4755968 + }, + { + "name": "transformer.h.1.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5018112 + }, + { + "name": "transformer.h.1.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5022208 + }, + { + "name": "transformer.h.1.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5026304 + }, + { + "name": "transformer.h.1.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 5030400 + }, + { + "name": "transformer.h.1.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 13419008 + }, + { + "name": "transformer.h.1.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 14467584 + }, + { + "name": "transformer.h.1.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 14483968 + }, + { + "name": "transformer.h.1.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 22872576 + }, + { + "name": "transformer.h.1.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23921152 + }, + { + "name": "transformer.h.2.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23925248 + }, + { + "name": "transformer.h.2.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23929344 + }, + { + "name": "transformer.h.2.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23933440 + }, + { + "name": "transformer.h.2.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 26292736 + }, + { + "name": "transformer.h.2.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26587648 + }, + { + "name": "transformer.h.2.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26592256 + }, + { + "name": "transformer.h.2.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28689408 + }, + { + "name": "transformer.h.2.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28951552 + }, + { + "name": "transformer.h.2.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28955648 + }, + { + "name": "transformer.h.2.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28959744 + } + ], + "md5sum": "142d0a395f60dfe2c7a26f32f144f05f" + }, + { + "dataPath": "params_shard_3.bin", + "format": "raw-shard", + "nbytes": 33387008, + "records": [ + { + "name": "transformer.h.2.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "transformer.h.2.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "transformer.h.2.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 9437184 + }, + { + "name": "transformer.h.2.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 9453568 + }, + { + "name": "transformer.h.2.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 17842176 + }, + { + "name": "transformer.h.2.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18890752 + }, + { + "name": "transformer.h.3.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18894848 + }, + { + "name": "transformer.h.3.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18898944 + }, + { + "name": "transformer.h.3.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 18903040 + }, + { + "name": "transformer.h.3.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 21262336 + }, + { + "name": "transformer.h.3.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 21557248 + }, + { + "name": "transformer.h.3.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 21561856 + }, + { + "name": "transformer.h.3.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 23659008 + }, + { + "name": "transformer.h.3.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23921152 + }, + { + "name": "transformer.h.3.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23925248 + }, + { + "name": "transformer.h.3.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23929344 + }, + { + "name": "transformer.h.3.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23933440 + }, + { + "name": "transformer.h.3.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 32322048 + }, + { + "name": "transformer.h.3.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 33370624 + } + ], + "md5sum": "cf13d7f71dd1bc461d13faca89277ef3" + }, + { + "dataPath": "params_shard_4.bin", + "format": "raw-shard", + "nbytes": 33382912, + "records": [ + { + "name": "transformer.h.3.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "transformer.h.3.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "transformer.h.3.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9437184 + }, + { + "name": "transformer.h.4.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9441280 + }, + { + "name": "transformer.h.4.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9445376 + }, + { + "name": "transformer.h.4.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 9449472 + }, + { + "name": "transformer.h.4.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 11808768 + }, + { + "name": "transformer.h.4.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 12103680 + }, + { + "name": "transformer.h.4.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 12108288 + }, + { + "name": "transformer.h.4.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 14205440 + }, + { + "name": "transformer.h.4.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 14467584 + }, + { + "name": "transformer.h.4.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 14471680 + }, + { + "name": "transformer.h.4.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 14475776 + }, + { + "name": "transformer.h.4.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 14479872 + }, + { + "name": "transformer.h.4.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 22868480 + }, + { + "name": "transformer.h.4.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 23917056 + }, + { + "name": "transformer.h.4.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23933440 + }, + { + "name": "transformer.h.4.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 32322048 + }, + { + "name": "transformer.h.4.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33370624 + }, + { + "name": "transformer.h.5.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33374720 + }, + { + "name": "transformer.h.5.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33378816 + } + ], + "md5sum": "c5ba04382fe711770f84107be85a31c7" + }, + { + "dataPath": "params_shard_5.bin", + "format": "raw-shard", + "nbytes": 28963840, + "records": [ + { + "name": "transformer.h.5.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 0 + }, + { + "name": "transformer.h.5.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 2359296 + }, + { + "name": "transformer.h.5.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "transformer.h.5.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 2658816 + }, + { + "name": "transformer.h.5.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 4755968 + }, + { + "name": "transformer.h.5.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5018112 + }, + { + "name": "transformer.h.5.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5022208 + }, + { + "name": "transformer.h.5.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5026304 + }, + { + "name": "transformer.h.5.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 5030400 + }, + { + "name": "transformer.h.5.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 13419008 + }, + { + "name": "transformer.h.5.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 14467584 + }, + { + "name": "transformer.h.5.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 14483968 + }, + { + "name": "transformer.h.5.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 22872576 + }, + { + "name": "transformer.h.5.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23921152 + }, + { + "name": "transformer.h.6.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23925248 + }, + { + "name": "transformer.h.6.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23929344 + }, + { + "name": "transformer.h.6.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23933440 + }, + { + "name": "transformer.h.6.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 26292736 + }, + { + "name": "transformer.h.6.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26587648 + }, + { + "name": "transformer.h.6.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26592256 + }, + { + "name": "transformer.h.6.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28689408 + }, + { + "name": "transformer.h.6.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28951552 + }, + { + "name": "transformer.h.6.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28955648 + }, + { + "name": "transformer.h.6.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28959744 + } + ], + "md5sum": "3252c3f5c0fd31bbeb3cab51b7b712e3" + }, + { + "dataPath": "params_shard_6.bin", + "format": "raw-shard", + "nbytes": 33387008, + "records": [ + { + "name": "transformer.h.6.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "transformer.h.6.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "transformer.h.6.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 9437184 + }, + { + "name": "transformer.h.6.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 9453568 + }, + { + "name": "transformer.h.6.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 17842176 + }, + { + "name": "transformer.h.6.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18890752 + }, + { + "name": "transformer.h.7.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18894848 + }, + { + "name": "transformer.h.7.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18898944 + }, + { + "name": "transformer.h.7.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 18903040 + }, + { + "name": "transformer.h.7.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 21262336 + }, + { + "name": "transformer.h.7.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 21557248 + }, + { + "name": "transformer.h.7.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 21561856 + }, + { + "name": "transformer.h.7.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 23659008 + }, + { + "name": "transformer.h.7.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23921152 + }, + { + "name": "transformer.h.7.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23925248 + }, + { + "name": "transformer.h.7.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23929344 + }, + { + "name": "transformer.h.7.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23933440 + }, + { + "name": "transformer.h.7.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 32322048 + }, + { + "name": "transformer.h.7.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 33370624 + } + ], + "md5sum": "25177ac9ee7018d30a389c8f71591565" + }, + { + "dataPath": "params_shard_7.bin", + "format": "raw-shard", + "nbytes": 33382912, + "records": [ + { + "name": "transformer.h.7.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "transformer.h.7.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "transformer.h.7.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9437184 + }, + { + "name": "transformer.h.8.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9441280 + }, + { + "name": "transformer.h.8.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9445376 + }, + { + "name": "transformer.h.8.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 9449472 + }, + { + "name": "transformer.h.8.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 11808768 + }, + { + "name": "transformer.h.8.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 12103680 + }, + { + "name": "transformer.h.8.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 12108288 + }, + { + "name": "transformer.h.8.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 14205440 + }, + { + "name": "transformer.h.8.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 14467584 + }, + { + "name": "transformer.h.8.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 14471680 + }, + { + "name": "transformer.h.8.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 14475776 + }, + { + "name": "transformer.h.8.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 14479872 + }, + { + "name": "transformer.h.8.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 22868480 + }, + { + "name": "transformer.h.8.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 23917056 + }, + { + "name": "transformer.h.8.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23933440 + }, + { + "name": "transformer.h.8.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 32322048 + }, + { + "name": "transformer.h.8.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33370624 + }, + { + "name": "transformer.h.9.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33374720 + }, + { + "name": "transformer.h.9.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33378816 + } + ], + "md5sum": "c47f82d20e857ad7aff61bf2dab8e61c" + }, + { + "dataPath": "params_shard_8.bin", + "format": "raw-shard", + "nbytes": 28963840, + "records": [ + { + "name": "transformer.h.9.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 0 + }, + { + "name": "transformer.h.9.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 2359296 + }, + { + "name": "transformer.h.9.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "transformer.h.9.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 2658816 + }, + { + "name": "transformer.h.9.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 4755968 + }, + { + "name": "transformer.h.9.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5018112 + }, + { + "name": "transformer.h.9.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5022208 + }, + { + "name": "transformer.h.9.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5026304 + }, + { + "name": "transformer.h.9.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 5030400 + }, + { + "name": "transformer.h.9.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 13419008 + }, + { + "name": "transformer.h.9.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 14467584 + }, + { + "name": "transformer.h.9.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 14483968 + }, + { + "name": "transformer.h.9.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 22872576 + }, + { + "name": "transformer.h.9.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23921152 + }, + { + "name": "transformer.h.10.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23925248 + }, + { + "name": "transformer.h.10.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23929344 + }, + { + "name": "transformer.h.10.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23933440 + }, + { + "name": "transformer.h.10.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 26292736 + }, + { + "name": "transformer.h.10.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26587648 + }, + { + "name": "transformer.h.10.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26592256 + }, + { + "name": "transformer.h.10.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28689408 + }, + { + "name": "transformer.h.10.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28951552 + }, + { + "name": "transformer.h.10.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28955648 + }, + { + "name": "transformer.h.10.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28959744 + } + ], + "md5sum": "00806b1f373302a9d68949200bdd1c42" + }, + { + "dataPath": "params_shard_9.bin", + "format": "raw-shard", + "nbytes": 33387008, + "records": [ + { + "name": "transformer.h.10.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "transformer.h.10.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "transformer.h.10.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 9437184 + }, + { + "name": "transformer.h.10.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 9453568 + }, + { + "name": "transformer.h.10.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 17842176 + }, + { + "name": "transformer.h.10.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18890752 + }, + { + "name": "transformer.h.11.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18894848 + }, + { + "name": "transformer.h.11.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18898944 + }, + { + "name": "transformer.h.11.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 18903040 + }, + { + "name": "transformer.h.11.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 21262336 + }, + { + "name": "transformer.h.11.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 21557248 + }, + { + "name": "transformer.h.11.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 21561856 + }, + { + "name": "transformer.h.11.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 23659008 + }, + { + "name": "transformer.h.11.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23921152 + }, + { + "name": "transformer.h.11.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23925248 + }, + { + "name": "transformer.h.11.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23929344 + }, + { + "name": "transformer.h.11.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23933440 + }, + { + "name": "transformer.h.11.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 32322048 + }, + { + "name": "transformer.h.11.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 33370624 + } + ], + "md5sum": "968bad8365fd4a556283fdf34875974d" + }, + { + "dataPath": "params_shard_10.bin", + "format": "raw-shard", + "nbytes": 33382912, + "records": [ + { + "name": "transformer.h.11.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "transformer.h.11.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "transformer.h.11.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9437184 + }, + { + "name": "transformer.h.12.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9441280 + }, + { + "name": "transformer.h.12.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9445376 + }, + { + "name": "transformer.h.12.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 9449472 + }, + { + "name": "transformer.h.12.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 11808768 + }, + { + "name": "transformer.h.12.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 12103680 + }, + { + "name": "transformer.h.12.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 12108288 + }, + { + "name": "transformer.h.12.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 14205440 + }, + { + "name": "transformer.h.12.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 14467584 + }, + { + "name": "transformer.h.12.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 14471680 + }, + { + "name": "transformer.h.12.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 14475776 + }, + { + "name": "transformer.h.12.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 14479872 + }, + { + "name": "transformer.h.12.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 22868480 + }, + { + "name": "transformer.h.12.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 23917056 + }, + { + "name": "transformer.h.12.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23933440 + }, + { + "name": "transformer.h.12.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 32322048 + }, + { + "name": "transformer.h.12.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33370624 + }, + { + "name": "transformer.h.13.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33374720 + }, + { + "name": "transformer.h.13.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33378816 + } + ], + "md5sum": "be7dcf200906df6f86a9bfc3cb55f0b7" + }, + { + "dataPath": "params_shard_11.bin", + "format": "raw-shard", + "nbytes": 28963840, + "records": [ + { + "name": "transformer.h.13.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 0 + }, + { + "name": "transformer.h.13.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 2359296 + }, + { + "name": "transformer.h.13.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "transformer.h.13.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 2658816 + }, + { + "name": "transformer.h.13.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 4755968 + }, + { + "name": "transformer.h.13.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5018112 + }, + { + "name": "transformer.h.13.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5022208 + }, + { + "name": "transformer.h.13.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5026304 + }, + { + "name": "transformer.h.13.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 5030400 + }, + { + "name": "transformer.h.13.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 13419008 + }, + { + "name": "transformer.h.13.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 14467584 + }, + { + "name": "transformer.h.13.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 14483968 + }, + { + "name": "transformer.h.13.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 22872576 + }, + { + "name": "transformer.h.13.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23921152 + }, + { + "name": "transformer.h.14.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23925248 + }, + { + "name": "transformer.h.14.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23929344 + }, + { + "name": "transformer.h.14.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23933440 + }, + { + "name": "transformer.h.14.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 26292736 + }, + { + "name": "transformer.h.14.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26587648 + }, + { + "name": "transformer.h.14.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26592256 + }, + { + "name": "transformer.h.14.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28689408 + }, + { + "name": "transformer.h.14.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28951552 + }, + { + "name": "transformer.h.14.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28955648 + }, + { + "name": "transformer.h.14.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28959744 + } + ], + "md5sum": "e82e9c4fda7cfa51ff997e46661bacdb" + }, + { + "dataPath": "params_shard_12.bin", + "format": "raw-shard", + "nbytes": 33387008, + "records": [ + { + "name": "transformer.h.14.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "transformer.h.14.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "transformer.h.14.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 9437184 + }, + { + "name": "transformer.h.14.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 9453568 + }, + { + "name": "transformer.h.14.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 17842176 + }, + { + "name": "transformer.h.14.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18890752 + }, + { + "name": "transformer.h.15.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18894848 + }, + { + "name": "transformer.h.15.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18898944 + }, + { + "name": "transformer.h.15.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 18903040 + }, + { + "name": "transformer.h.15.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 21262336 + }, + { + "name": "transformer.h.15.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 21557248 + }, + { + "name": "transformer.h.15.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 21561856 + }, + { + "name": "transformer.h.15.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 23659008 + }, + { + "name": "transformer.h.15.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23921152 + }, + { + "name": "transformer.h.15.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23925248 + }, + { + "name": "transformer.h.15.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23929344 + }, + { + "name": "transformer.h.15.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23933440 + }, + { + "name": "transformer.h.15.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 32322048 + }, + { + "name": "transformer.h.15.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 33370624 + } + ], + "md5sum": "f2fe56417ba5feb375bfbff3bb330777" + }, + { + "dataPath": "params_shard_13.bin", + "format": "raw-shard", + "nbytes": 33382912, + "records": [ + { + "name": "transformer.h.15.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "transformer.h.15.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "transformer.h.15.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9437184 + }, + { + "name": "transformer.h.16.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9441280 + }, + { + "name": "transformer.h.16.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9445376 + }, + { + "name": "transformer.h.16.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 9449472 + }, + { + "name": "transformer.h.16.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 11808768 + }, + { + "name": "transformer.h.16.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 12103680 + }, + { + "name": "transformer.h.16.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 12108288 + }, + { + "name": "transformer.h.16.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 14205440 + }, + { + "name": "transformer.h.16.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 14467584 + }, + { + "name": "transformer.h.16.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 14471680 + }, + { + "name": "transformer.h.16.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 14475776 + }, + { + "name": "transformer.h.16.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 14479872 + }, + { + "name": "transformer.h.16.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 22868480 + }, + { + "name": "transformer.h.16.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 23917056 + }, + { + "name": "transformer.h.16.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23933440 + }, + { + "name": "transformer.h.16.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 32322048 + }, + { + "name": "transformer.h.16.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33370624 + }, + { + "name": "transformer.h.17.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33374720 + }, + { + "name": "transformer.h.17.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33378816 + } + ], + "md5sum": "a52393c9d9025bf27ea12a597ca0ff04" + }, + { + "dataPath": "params_shard_14.bin", + "format": "raw-shard", + "nbytes": 28963840, + "records": [ + { + "name": "transformer.h.17.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 0 + }, + { + "name": "transformer.h.17.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 2359296 + }, + { + "name": "transformer.h.17.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "transformer.h.17.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 2658816 + }, + { + "name": "transformer.h.17.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 4755968 + }, + { + "name": "transformer.h.17.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5018112 + }, + { + "name": "transformer.h.17.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5022208 + }, + { + "name": "transformer.h.17.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5026304 + }, + { + "name": "transformer.h.17.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 5030400 + }, + { + "name": "transformer.h.17.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 13419008 + }, + { + "name": "transformer.h.17.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 14467584 + }, + { + "name": "transformer.h.17.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 14483968 + }, + { + "name": "transformer.h.17.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 22872576 + }, + { + "name": "transformer.h.17.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23921152 + }, + { + "name": "transformer.h.18.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23925248 + }, + { + "name": "transformer.h.18.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23929344 + }, + { + "name": "transformer.h.18.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23933440 + }, + { + "name": "transformer.h.18.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 26292736 + }, + { + "name": "transformer.h.18.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26587648 + }, + { + "name": "transformer.h.18.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26592256 + }, + { + "name": "transformer.h.18.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28689408 + }, + { + "name": "transformer.h.18.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28951552 + }, + { + "name": "transformer.h.18.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28955648 + }, + { + "name": "transformer.h.18.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28959744 + } + ], + "md5sum": "fd15a3c071525656f07424171cbb46e2" + }, + { + "dataPath": "params_shard_15.bin", + "format": "raw-shard", + "nbytes": 33387008, + "records": [ + { + "name": "transformer.h.18.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "transformer.h.18.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "transformer.h.18.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 9437184 + }, + { + "name": "transformer.h.18.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 9453568 + }, + { + "name": "transformer.h.18.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 17842176 + }, + { + "name": "transformer.h.18.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18890752 + }, + { + "name": "transformer.h.19.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18894848 + }, + { + "name": "transformer.h.19.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18898944 + }, + { + "name": "transformer.h.19.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 18903040 + }, + { + "name": "transformer.h.19.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 21262336 + }, + { + "name": "transformer.h.19.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 21557248 + }, + { + "name": "transformer.h.19.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 21561856 + }, + { + "name": "transformer.h.19.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 23659008 + }, + { + "name": "transformer.h.19.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23921152 + }, + { + "name": "transformer.h.19.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23925248 + }, + { + "name": "transformer.h.19.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23929344 + }, + { + "name": "transformer.h.19.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23933440 + }, + { + "name": "transformer.h.19.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 32322048 + }, + { + "name": "transformer.h.19.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 33370624 + } + ], + "md5sum": "e333e89d3d0e04646d09a103c11d9725" + }, + { + "dataPath": "params_shard_16.bin", + "format": "raw-shard", + "nbytes": 33382912, + "records": [ + { + "name": "transformer.h.19.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "transformer.h.19.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "transformer.h.19.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9437184 + }, + { + "name": "transformer.h.20.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9441280 + }, + { + "name": "transformer.h.20.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9445376 + }, + { + "name": "transformer.h.20.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 9449472 + }, + { + "name": "transformer.h.20.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 11808768 + }, + { + "name": "transformer.h.20.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 12103680 + }, + { + "name": "transformer.h.20.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 12108288 + }, + { + "name": "transformer.h.20.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 14205440 + }, + { + "name": "transformer.h.20.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 14467584 + }, + { + "name": "transformer.h.20.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 14471680 + }, + { + "name": "transformer.h.20.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 14475776 + }, + { + "name": "transformer.h.20.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 14479872 + }, + { + "name": "transformer.h.20.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 22868480 + }, + { + "name": "transformer.h.20.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 23917056 + }, + { + "name": "transformer.h.20.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23933440 + }, + { + "name": "transformer.h.20.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 32322048 + }, + { + "name": "transformer.h.20.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33370624 + }, + { + "name": "transformer.h.21.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33374720 + }, + { + "name": "transformer.h.21.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33378816 + } + ], + "md5sum": "cf0962fb2fd7bc2c5a89a79692b56d06" + }, + { + "dataPath": "params_shard_17.bin", + "format": "raw-shard", + "nbytes": 28963840, + "records": [ + { + "name": "transformer.h.21.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 0 + }, + { + "name": "transformer.h.21.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 2359296 + }, + { + "name": "transformer.h.21.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 2654208 + }, + { + "name": "transformer.h.21.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 2658816 + }, + { + "name": "transformer.h.21.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 4755968 + }, + { + "name": "transformer.h.21.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5018112 + }, + { + "name": "transformer.h.21.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5022208 + }, + { + "name": "transformer.h.21.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 5026304 + }, + { + "name": "transformer.h.21.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 5030400 + }, + { + "name": "transformer.h.21.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 13419008 + }, + { + "name": "transformer.h.21.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 14467584 + }, + { + "name": "transformer.h.21.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 14483968 + }, + { + "name": "transformer.h.21.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 22872576 + }, + { + "name": "transformer.h.21.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23921152 + }, + { + "name": "transformer.h.22.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23925248 + }, + { + "name": "transformer.h.22.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23929344 + }, + { + "name": "transformer.h.22.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23933440 + }, + { + "name": "transformer.h.22.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 26292736 + }, + { + "name": "transformer.h.22.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26587648 + }, + { + "name": "transformer.h.22.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26592256 + }, + { + "name": "transformer.h.22.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28689408 + }, + { + "name": "transformer.h.22.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28951552 + }, + { + "name": "transformer.h.22.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28955648 + }, + { + "name": "transformer.h.22.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28959744 + } + ], + "md5sum": "cecccd0a0634f308c850742bbf1bc8d4" + }, + { + "dataPath": "params_shard_18.bin", + "format": "raw-shard", + "nbytes": 33387008, + "records": [ + { + "name": "transformer.h.22.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "transformer.h.22.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "transformer.h.22.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 9437184 + }, + { + "name": "transformer.h.22.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 9453568 + }, + { + "name": "transformer.h.22.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 17842176 + }, + { + "name": "transformer.h.22.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18890752 + }, + { + "name": "transformer.h.23.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18894848 + }, + { + "name": "transformer.h.23.ln_1.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18898944 + }, + { + "name": "transformer.h.23.attn.c_attn.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 18903040 + }, + { + "name": "transformer.h.23.attn.c_attn.q_scale", + "shape": [ + 2304, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 294912, + "byteOffset": 21262336 + }, + { + "name": "transformer.h.23.attn.c_attn.bias", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 21557248 + }, + { + "name": "transformer.h.23.attn.c_proj.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 21561856 + }, + { + "name": "transformer.h.23.attn.c_proj.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 23659008 + }, + { + "name": "transformer.h.23.attn.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23921152 + }, + { + "name": "transformer.h.23.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23925248 + }, + { + "name": "transformer.h.23.ln_2.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 23929344 + }, + { + "name": "transformer.h.23.mlp.c_fc.q_weight", + "shape": [ + 8192, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23933440 + }, + { + "name": "transformer.h.23.mlp.c_fc.q_scale", + "shape": [ + 8192, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 32322048 + }, + { + "name": "transformer.h.23.mlp.c_fc.bias", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 16384, + "byteOffset": 33370624 + } + ], + "md5sum": "ce80d822f52c4a96a4529a5f45671bc9" + }, + { + "dataPath": "params_shard_19.bin", + "format": "raw-shard", + "nbytes": 50462720, + "records": [ + { + "name": "lm_head.q_weight", + "shape": [ + 49280, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 50462720, + "byteOffset": 0 + } + ], + "md5sum": "37752cf9591772a22574b2b1fcb29a93" + }, + { + "dataPath": "params_shard_20.bin", + "format": "raw-shard", + "nbytes": 15757312, + "records": [ + { + "name": "transformer.h.23.mlp.c_proj.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "transformer.h.23.mlp.c_proj.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "transformer.h.23.mlp.c_proj.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9437184 + }, + { + "name": "transformer.ln_f.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9441280 + }, + { + "name": "transformer.ln_f.bias", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9445376 + }, + { + "name": "lm_head.q_scale", + "shape": [ + 49280, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6307840, + "byteOffset": 9449472 + } + ], + "md5sum": "c7d5887dd10810c0f22e810413f4c29a" + } + ] +} \ No newline at end of file