diff --git "a/ndarray-cache.json" "b/ndarray-cache.json" new file mode 100644--- /dev/null +++ "b/ndarray-cache.json" @@ -0,0 +1,3521 @@ +{ + "metadata": { + "ParamSize": 283, + "ParamBytes": 1656834048.0, + "BitsPerParam": 4.069857841273854 + }, + "records": [ + { + "dataPath": "params_shard_0.bin", + "format": "raw-shard", + "nbytes": 197001216, + "records": [ + { + "name": "model.embed_tokens.q_weight", + "shape": [ + 128256, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 197001216, + "byteOffset": 0 + } + ], + "md5sum": "4a201b9f7c714fcbb2854eb2672468c2" + }, + { + "dataPath": "params_shard_1.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.0.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "83d8a2313ad0695a7f50eeecb4b94698" + }, + { + "dataPath": "params_shard_2.bin", + "format": "raw-shard", + "nbytes": 32913408, + "records": [ + { + "name": "model.embed_tokens.q_scale", + "shape": [ + 128256, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6156288, + "byteOffset": 0 + }, + { + "name": "model.layers.0.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 6156288 + }, + { + "name": "model.layers.0.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 6162432 + }, + { + "name": "model.layers.0.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 18745344 + }, + { + "name": "model.layers.0.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 19138560 + }, + { + "name": "model.layers.0.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 19924992 + }, + { + "name": "model.layers.0.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 19931136 + }, + { + "name": "model.layers.0.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 27795456 + }, + { + "name": "model.layers.0.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 28041216 + }, + { + "name": "model.layers.0.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 32759808 + }, + { + "name": "model.layers.1.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 32907264 + } + ], + "md5sum": "cabe79d16cfb9b59fb737a6f7888f3b1" + }, + { + "dataPath": "params_shard_3.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.1.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "7a7bc92d640bdca4d1be3ae049e8619b" + }, + { + "dataPath": "params_shard_4.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.1.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.1.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.1.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.1.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.1.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.1.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.1.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.1.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.10.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "87e8b7d1b6462571ff7d35497b323137" + }, + { + "dataPath": "params_shard_5.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.10.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "ae5404602d0120258964826cb874e8ed" + }, + { + "dataPath": "params_shard_6.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.10.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.10.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.10.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.10.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.10.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.10.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.10.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.10.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.11.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "144323332c43e6db5c50e10233fd9ff8" + }, + { + "dataPath": "params_shard_7.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.11.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "c2078c5907c9017fbbe7325d06385380" + }, + { + "dataPath": "params_shard_8.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.11.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.11.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.11.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.11.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.11.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.11.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.11.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.11.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.12.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "3a650d513ad957b6296da19f146cafd4" + }, + { + "dataPath": "params_shard_9.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.12.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "038a5a328b9c8f7028eb30ebec07c1ca" + }, + { + "dataPath": "params_shard_10.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.12.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.12.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.12.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.12.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.12.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.12.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.12.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.12.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.13.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "df5ac66d0bd07f933bea8ea4a539610e" + }, + { + "dataPath": "params_shard_11.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.13.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "3c7c40509d2e2ad66de5e6502812ca18" + }, + { + "dataPath": "params_shard_12.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.13.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.13.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.13.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.13.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.13.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.13.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.13.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.13.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.14.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "c41b61842f98c008bea1291f1b13f766" + }, + { + "dataPath": "params_shard_13.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.14.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "d6e7bc6073d917a3fab253ac60edd3a5" + }, + { + "dataPath": "params_shard_14.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.14.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.14.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.14.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.14.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.14.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.14.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.14.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.14.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.15.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "e1ad9750c18edc831317fb49fd2ceef8" + }, + { + "dataPath": "params_shard_15.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.15.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "c1f623134c5627159a109db7aa154e7f" + }, + { + "dataPath": "params_shard_16.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.15.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.15.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.15.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.15.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.15.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.15.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.15.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.15.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.16.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "8c1cd702c6f089afb0685cc1ec749190" + }, + { + "dataPath": "params_shard_17.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.16.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "6bdde96d2bd0e08cae8f7a26901088a1" + }, + { + "dataPath": "params_shard_18.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.16.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.16.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.16.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.16.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.16.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.16.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.16.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.16.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.17.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "68a45707fb9d2b5c4a9ead878de5c48a" + }, + { + "dataPath": "params_shard_19.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.17.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "0605028a2915853c76fb1b11ae677899" + }, + { + "dataPath": "params_shard_20.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.17.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.17.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.17.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.17.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.17.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.17.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.17.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.17.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.18.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "26f3bc43be779bc1bb59aee28fa5c2c6" + }, + { + "dataPath": "params_shard_21.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.18.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "b8637153c946844ff07b0397adba54df" + }, + { + "dataPath": "params_shard_22.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.18.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.18.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.18.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.18.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.18.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.18.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.18.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.18.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.19.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "99096fc4593ad6f1094f323d05b2250d" + }, + { + "dataPath": "params_shard_23.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.19.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "8a2047f1ee95d5e846506664632f9149" + }, + { + "dataPath": "params_shard_24.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.19.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.19.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.19.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.19.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.19.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.19.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.19.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.19.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.2.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "4738ffd625c7b65bd28d523f8aa61d04" + }, + { + "dataPath": "params_shard_25.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.2.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "d994307fa762cae0692dad5001f3c783" + }, + { + "dataPath": "params_shard_26.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.20.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "05e83127e0c5202754ca7195ad5eff7b" + }, + { + "dataPath": "params_shard_27.bin", + "format": "raw-shard", + "nbytes": 27531264, + "records": [ + { + "name": "model.layers.2.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.2.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.2.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.2.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.2.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.2.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.2.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.2.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.20.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 26744832 + } + ], + "md5sum": "0aaa89301a1016c0887f66b1176cc952" + }, + { + "dataPath": "params_shard_28.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.3.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "57d1b74c7c00cc9c32fd7bccc75101cf" + }, + { + "dataPath": "params_shard_29.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.20.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "model.layers.20.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 7864320 + }, + { + "name": "model.layers.20.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 8110080 + }, + { + "name": "model.layers.20.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 12828672 + }, + { + "name": "model.layers.3.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 12976128 + }, + { + "name": "model.layers.3.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 12982272 + }, + { + "name": "model.layers.3.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 25565184 + }, + { + "name": "model.layers.3.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25958400 + }, + { + "name": "model.layers.3.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "ff4cfcad4cfb96262641351f2d0c7a3d" + }, + { + "dataPath": "params_shard_30.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.4.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "77d9293ea9e211739f37b3a667988ae9" + }, + { + "dataPath": "params_shard_31.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.3.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "model.layers.3.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 7864320 + }, + { + "name": "model.layers.3.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 8110080 + }, + { + "name": "model.layers.3.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 12828672 + }, + { + "name": "model.layers.4.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 12976128 + }, + { + "name": "model.layers.4.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 12982272 + }, + { + "name": "model.layers.4.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 25565184 + }, + { + "name": "model.layers.4.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25958400 + }, + { + "name": "model.layers.4.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "b69a94dc7ffa376879b99beb2d43c56d" + }, + { + "dataPath": "params_shard_32.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.5.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "18f85cedec776ddc90d268b9edf81f16" + }, + { + "dataPath": "params_shard_33.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.4.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "model.layers.4.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 7864320 + }, + { + "name": "model.layers.4.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 8110080 + }, + { + "name": "model.layers.4.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 12828672 + }, + { + "name": "model.layers.5.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 12976128 + }, + { + "name": "model.layers.5.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 12982272 + }, + { + "name": "model.layers.5.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 25565184 + }, + { + "name": "model.layers.5.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25958400 + }, + { + "name": "model.layers.5.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "1701f4f12a75168b81f730d866a2fa08" + }, + { + "dataPath": "params_shard_34.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.6.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "7ad9202b57b0417b34d96c700605a4b5" + }, + { + "dataPath": "params_shard_35.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.5.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "model.layers.5.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 7864320 + }, + { + "name": "model.layers.5.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 8110080 + }, + { + "name": "model.layers.5.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 12828672 + }, + { + "name": "model.layers.6.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 12976128 + }, + { + "name": "model.layers.6.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 12982272 + }, + { + "name": "model.layers.6.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 25565184 + }, + { + "name": "model.layers.6.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25958400 + }, + { + "name": "model.layers.6.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "aa912fde4b4a6dfe412243653ff0c698" + }, + { + "dataPath": "params_shard_36.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.7.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "b70d5f38f59835f49701105df685328d" + }, + { + "dataPath": "params_shard_37.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.6.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "model.layers.6.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 7864320 + }, + { + "name": "model.layers.6.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 8110080 + }, + { + "name": "model.layers.6.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 12828672 + }, + { + "name": "model.layers.7.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 12976128 + }, + { + "name": "model.layers.7.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 12982272 + }, + { + "name": "model.layers.7.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 25565184 + }, + { + "name": "model.layers.7.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25958400 + }, + { + "name": "model.layers.7.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "58dafd0de8ff53d418e839496e44043d" + }, + { + "dataPath": "params_shard_38.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.8.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "0edb641b5c82cdc6cf0dc56b12db78a7" + }, + { + "dataPath": "params_shard_39.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.7.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "model.layers.7.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 7864320 + }, + { + "name": "model.layers.7.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 8110080 + }, + { + "name": "model.layers.7.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 12828672 + }, + { + "name": "model.layers.8.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 12976128 + }, + { + "name": "model.layers.8.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 12982272 + }, + { + "name": "model.layers.8.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 25565184 + }, + { + "name": "model.layers.8.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25958400 + }, + { + "name": "model.layers.8.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "c94d38af642ce1b55cf9bd143c5ecbd9" + }, + { + "dataPath": "params_shard_40.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.9.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "4fe615e1b1cfeff782003a0c7e3e6e01" + }, + { + "dataPath": "params_shard_41.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.8.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "model.layers.8.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 7864320 + }, + { + "name": "model.layers.8.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 8110080 + }, + { + "name": "model.layers.8.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 12828672 + }, + { + "name": "model.layers.9.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 12976128 + }, + { + "name": "model.layers.9.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 12982272 + }, + { + "name": "model.layers.9.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 25565184 + }, + { + "name": "model.layers.9.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25958400 + }, + { + "name": "model.layers.9.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "f0ecbaa7191336e67980c663f6d6d208" + }, + { + "dataPath": "params_shard_42.bin", + "format": "raw-shard", + "nbytes": 25970688, + "records": [ + { + "name": "model.layers.9.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "model.layers.9.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 7864320 + }, + { + "name": "model.layers.9.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 8110080 + }, + { + "name": "model.layers.9.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 12828672 + }, + { + "name": "model.layers.20.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 12976128 + }, + { + "name": "model.layers.20.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 12982272 + }, + { + "name": "model.layers.20.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 25565184 + }, + { + "name": "model.layers.20.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 25958400 + }, + { + "name": "model.layers.21.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 25964544 + } + ], + "md5sum": "fe8f72a9aff4da784bb8f35ef8c1f885" + }, + { + "dataPath": "params_shard_43.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.21.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "1021522a19a496d6a6d9e0033e995511" + }, + { + "dataPath": "params_shard_44.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.21.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.21.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.21.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.21.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.21.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.21.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.21.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.21.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.22.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "eaedf91426feedc357266fb4aa5076d1" + }, + { + "dataPath": "params_shard_45.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.22.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "44a42cdd0428b6d034619d666e78c457" + }, + { + "dataPath": "params_shard_46.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.22.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.22.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.22.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.22.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.22.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.22.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.22.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.22.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.23.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "1e11708d29c4da30db3fa72a80b0fd97" + }, + { + "dataPath": "params_shard_47.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.23.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "da29a2b9fefe1a5b62e7381be56353ac" + }, + { + "dataPath": "params_shard_48.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.23.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.23.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.23.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.23.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.23.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.23.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.23.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.23.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.24.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "ca0c5cd994984e0aee5b03afdf8ff45f" + }, + { + "dataPath": "params_shard_49.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.24.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "b3f748ef4f6121c0617a75f250ff9ab2" + }, + { + "dataPath": "params_shard_50.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.24.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.24.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.24.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.24.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.24.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.24.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.24.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.24.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.25.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "ad99ae7f7c87d6a7321e78ba56d9a8fe" + }, + { + "dataPath": "params_shard_51.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.25.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "a0298474fdcfc7f501f310744c32cc17" + }, + { + "dataPath": "params_shard_52.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.25.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.25.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.25.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.25.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.25.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.25.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.25.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.25.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.26.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "f6ff882325497a003b55e3aec32469ef" + }, + { + "dataPath": "params_shard_53.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.26.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "e3253ff3570ac1fd71e84b0c601f379c" + }, + { + "dataPath": "params_shard_54.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.26.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.26.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.26.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.26.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.26.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.26.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.26.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.26.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.layers.27.input_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "e8ff8a556e1c7de9bc75dd3fc755f76b" + }, + { + "dataPath": "params_shard_55.bin", + "format": "raw-shard", + "nbytes": 25165824, + "records": [ + { + "name": "model.layers.27.mlp.gate_up_proj.q_weight", + "shape": [ + 16384, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 25165824, + "byteOffset": 0 + } + ], + "md5sum": "7b48efb69ade26a9e983c6aaea13e986" + }, + { + "dataPath": "params_shard_56.bin", + "format": "raw-shard", + "nbytes": 26750976, + "records": [ + { + "name": "model.layers.27.mlp.down_proj.q_weight", + "shape": [ + 3072, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 12582912, + "byteOffset": 0 + }, + { + "name": "model.layers.27.mlp.down_proj.q_scale", + "shape": [ + 3072, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 393216, + "byteOffset": 12582912 + }, + { + "name": "model.layers.27.mlp.gate_up_proj.q_scale", + "shape": [ + 16384, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 12976128 + }, + { + "name": "model.layers.27.post_attention_layernorm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 13762560 + }, + { + "name": "model.layers.27.self_attn.qkv_proj.q_weight", + "shape": [ + 5120, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 13768704 + }, + { + "name": "model.layers.27.self_attn.qkv_proj.q_scale", + "shape": [ + 5120, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 245760, + "byteOffset": 21633024 + }, + { + "name": "model.layers.27.self_attn.o_proj.q_weight", + "shape": [ + 3072, + 384 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 21878784 + }, + { + "name": "model.layers.27.self_attn.o_proj.q_scale", + "shape": [ + 3072, + 24 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 26597376 + }, + { + "name": "model.norm.weight", + "shape": [ + 3072 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6144, + "byteOffset": 26744832 + } + ], + "md5sum": "325dae8f30aaff18456072a7d32fce73" + } + ] +} \ No newline at end of file