diff --git "a/ndarray-cache.json" "b/ndarray-cache.json" new file mode 100644--- /dev/null +++ "b/ndarray-cache.json" @@ -0,0 +1,3119 @@ +{ + "metadata": { + "ParamSize": 269, + "ParamBytes": 1033572352.0, + "BitsPerParam": 4.501551474039708 + }, + "records": [ + { + "dataPath": "params_shard_0.bin", + "format": "raw-shard", + "nbytes": 155582464, + "records": [ + { + "name": "lm_head.q_weight", + "shape": [ + 256, + 151936 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 155582464, + "byteOffset": 0 + } + ], + "md5sum": "55d817231a8cfb46e80892f42f7486ff" + }, + { + "dataPath": "params_shard_1.bin", + "format": "raw-shard", + "nbytes": 28901376, + "records": [ + { + "name": "lm_head.q_scale", + "shape": [ + 64, + 151936 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 19447808, + "byteOffset": 0 + }, + { + "name": "transformer.h.14.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19447808 + }, + { + "name": "transformer.h.14.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19460096 + }, + { + "name": "transformer.h.14.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25751552 + }, + { + "name": "transformer.h.14.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26537984 + }, + { + "name": "transformer.h.14.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28635136 + }, + { + "name": "transformer.h.14.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28897280 + } + ], + "md5sum": "544f6005e17ec487fd384be0115ea74c" + }, + { + "dataPath": "params_shard_2.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.14.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.14.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.14.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.14.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.15.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.15.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.15.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.15.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.15.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.15.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.15.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "003faeb5d32289296b8fb6f1c0a360ad" + }, + { + "dataPath": "params_shard_3.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.15.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.15.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.15.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.15.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.16.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.16.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.16.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.16.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.16.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.16.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.16.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "6389d3daaab1740671c0dcc0ae14e5af" + }, + { + "dataPath": "params_shard_4.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.16.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.16.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.16.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.16.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.17.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.17.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.17.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.17.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.17.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.17.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.17.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "8ce342cbb7717f5749828e5850d6dfc1" + }, + { + "dataPath": "params_shard_5.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.17.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.17.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.17.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.17.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.18.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.18.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.18.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.18.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.18.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.18.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.18.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "5e2a13838d1ec9b41a248975eff518cd" + }, + { + "dataPath": "params_shard_6.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.18.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.18.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.18.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.18.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.19.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.19.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.19.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.19.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.19.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.19.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.19.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "819864e69047f88ccf40b00d1110d54e" + }, + { + "dataPath": "params_shard_7.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.19.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.19.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.19.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.19.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.20.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.20.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.20.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.20.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.20.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.20.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.20.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "43bc6e38b4026d041510933119b6d841" + }, + { + "dataPath": "params_shard_8.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.20.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.20.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.20.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.20.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.21.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.21.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.21.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.21.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.21.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.21.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.21.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "515ebbab4e302aa668094c8b7841effe" + }, + { + "dataPath": "params_shard_9.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.21.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.21.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.21.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.21.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.22.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.22.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.22.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.22.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.22.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.22.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.22.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "ab04377343cac28463bbc4f1a6a1b3e8" + }, + { + "dataPath": "params_shard_10.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.22.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.22.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.22.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.22.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.23.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.23.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.23.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.23.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.23.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.23.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.23.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "0542605592ffbe38e318a1ad1e627a3e" + }, + { + "dataPath": "params_shard_11.bin", + "format": "raw-shard", + "nbytes": 28483584, + "records": [ + { + "name": "transformer.h.23.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.23.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.23.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.23.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.ln_f.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.0.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19025920 + }, + { + "name": "transformer.h.0.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19038208 + }, + { + "name": "transformer.h.0.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25329664 + }, + { + "name": "transformer.h.0.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26116096 + }, + { + "name": "transformer.h.0.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28213248 + }, + { + "name": "transformer.h.0.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + }, + { + "name": "transformer.h.0.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28479488 + } + ], + "md5sum": "37e60dcede917d8b567dcf183b8f6fc8" + }, + { + "dataPath": "params_shard_12.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.0.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.0.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.0.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.0.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.1.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.1.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.1.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.1.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.1.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.1.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.1.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "506d468c4be56c6931de2181c0e568ef" + }, + { + "dataPath": "params_shard_13.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.1.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.1.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.1.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.1.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.10.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.10.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.10.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.10.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.10.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.10.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.10.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "3ef6f46eacceb2b8f0f6bfe3e0533728" + }, + { + "dataPath": "params_shard_14.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.10.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.10.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.10.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.10.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.11.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.11.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.11.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.11.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.11.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.11.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.11.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "11d8caa859eb3037bda85f770b6a3622" + }, + { + "dataPath": "params_shard_15.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.11.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.11.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.11.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.11.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.12.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.12.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.12.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.12.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.12.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.12.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.12.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "4c541ff06cef85d6a0e57ade16974aa8" + }, + { + "dataPath": "params_shard_16.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.12.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.12.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.12.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.12.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.13.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.13.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.13.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.13.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.13.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.13.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.13.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "52c0f30ffc4c941bd17767bb30260b6a" + }, + { + "dataPath": "params_shard_17.bin", + "format": "raw-shard", + "nbytes": 28483584, + "records": [ + { + "name": "transformer.h.13.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.13.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.13.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.13.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.14.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.2.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19025920 + }, + { + "name": "transformer.h.2.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19038208 + }, + { + "name": "transformer.h.2.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25329664 + }, + { + "name": "transformer.h.2.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26116096 + }, + { + "name": "transformer.h.2.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28213248 + }, + { + "name": "transformer.h.2.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + }, + { + "name": "transformer.h.2.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28479488 + } + ], + "md5sum": "4451758a6feb0da995bcff710427eb8a" + }, + { + "dataPath": "params_shard_18.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.2.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.2.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.2.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.2.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.3.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.3.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.3.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.3.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.3.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.3.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.3.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "067eba818ad7ce39c5b1a5c1c962a370" + }, + { + "dataPath": "params_shard_19.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.3.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.3.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.3.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.3.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.4.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.4.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.4.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.4.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.4.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.4.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.4.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "8f79e478fb710d8ba8083d492a6705cc" + }, + { + "dataPath": "params_shard_20.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.4.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.4.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.4.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.4.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.5.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.5.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.5.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.5.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.5.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.5.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.5.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "27a2164ab34e2d36ebe6ad2dfc7f52e9" + }, + { + "dataPath": "params_shard_21.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.5.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.5.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.5.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.5.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.6.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.6.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.6.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.6.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.6.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.6.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.6.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "26def18cdc64becee6079e2a16295b4c" + }, + { + "dataPath": "params_shard_22.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.6.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.6.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.6.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.6.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.7.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.7.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.7.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.7.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.7.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.7.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.7.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "7c0ac19942d64b79a4fd2ddd8f3f771e" + }, + { + "dataPath": "params_shard_23.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.7.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.7.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.7.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.7.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.8.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.8.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.8.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.8.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.8.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.8.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.8.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "dc0d14465ae623462b7ba730fc5a9f0e" + }, + { + "dataPath": "params_shard_24.bin", + "format": "raw-shard", + "nbytes": 28479488, + "records": [ + { + "name": "transformer.h.8.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.8.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.8.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.8.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + }, + { + "name": "transformer.h.9.attn.c_attn.bias", + "shape": [ + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 12288, + "byteOffset": 19021824 + }, + { + "name": "transformer.h.9.attn.c_attn.q_weight", + "shape": [ + 256, + 6144 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 19034112 + }, + { + "name": "transformer.h.9.attn.c_attn.q_scale", + "shape": [ + 64, + 6144 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 786432, + "byteOffset": 25325568 + }, + { + "name": "transformer.h.9.attn.c_proj.q_weight", + "shape": [ + 256, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 26112000 + }, + { + "name": "transformer.h.9.attn.c_proj.q_scale", + "shape": [ + 64, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28209152 + }, + { + "name": "transformer.h.9.ln_1.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28471296 + }, + { + "name": "transformer.h.9.ln_2.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 28475392 + } + ], + "md5sum": "1ed3df499c4a17db99c05a1e35062e88" + }, + { + "dataPath": "params_shard_25.bin", + "format": "raw-shard", + "nbytes": 155582464, + "records": [ + { + "name": "transformer.wte.q_weight", + "shape": [ + 151936, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 155582464, + "byteOffset": 0 + } + ], + "md5sum": "8b1ffdae0c8efb2938a2acbb5235e601" + }, + { + "dataPath": "params_shard_26.bin", + "format": "raw-shard", + "nbytes": 19447808, + "records": [ + { + "name": "transformer.wte.q_scale", + "shape": [ + 151936, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 19447808, + "byteOffset": 0 + } + ], + "md5sum": "ad620bed8cb3980cf1f2a9a6a43de187" + }, + { + "dataPath": "params_shard_27.bin", + "format": "raw-shard", + "nbytes": 19021824, + "records": [ + { + "name": "transformer.h.9.mlp.c_proj.q_weight", + "shape": [ + 688, + 2048 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 5636096, + "byteOffset": 0 + }, + { + "name": "transformer.h.9.mlp.c_proj.q_scale", + "shape": [ + 172, + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 704512, + "byteOffset": 5636096 + }, + { + "name": "transformer.h.9.mlp.gate_up_proj.q_weight", + "shape": [ + 256, + 11008 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 11272192, + "byteOffset": 6340608 + }, + { + "name": "transformer.h.9.mlp.gate_up_proj.q_scale", + "shape": [ + 64, + 11008 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1409024, + "byteOffset": 17612800 + } + ], + "md5sum": "95fee7353e4ba5d43350cfa8b51154ea" + } + ] +} \ No newline at end of file