diff --git "a/ndarray-cache.json" "b/ndarray-cache.json" new file mode 100644--- /dev/null +++ "b/ndarray-cache.json" @@ -0,0 +1,2919 @@ +{ + "metadata": { + "ParamSize": 245, + "ParamBytes": 1062768640.0, + "BitsPerParam": 4.500610895072402 + }, + "records": [ + { + "dataPath": "params_shard_0.bin", + "format": "raw-shard", + "nbytes": 25956352, + "records": [ + { + "name": "model.layers.0.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 0 + }, + { + "name": "model.layers.0.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 2097152 + }, + { + "name": "model.layers.0.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 2359296 + }, + { + "name": "model.layers.0.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 6553600 + }, + { + "name": "model.layers.0.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 7077888 + }, + { + "name": "model.layers.0.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 7081984 + }, + { + "name": "model.layers.0.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 23859200 + } + ], + "md5sum": "a25496cf4674e9f6408794f4e50908aa" + }, + { + "dataPath": "params_shard_1.bin", + "format": "raw-shard", + "nbytes": 33300480, + "records": [ + { + "name": "model.layers.0.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "model.layers.0.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "model.layers.0.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9437184 + }, + { + "name": "model.layers.1.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 9441280 + }, + { + "name": "model.layers.1.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 11538432 + }, + { + "name": "model.layers.1.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 11800576 + }, + { + "name": "model.layers.1.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 15994880 + }, + { + "name": "model.layers.1.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 16519168 + }, + { + "name": "model.layers.1.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 16523264 + } + ], + "md5sum": "11cf18c3a0b38b086704550102e0ca49" + }, + { + "dataPath": "params_shard_2.bin", + "format": "raw-shard", + "nbytes": 16777216, + "records": [ + { + "name": "model.layers.10.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 0 + } + ], + "md5sum": "ca71c08215c84d53fc086777bf6951fe" + }, + { + "dataPath": "params_shard_3.bin", + "format": "raw-shard", + "nbytes": 32518144, + "records": [ + { + "name": "model.layers.1.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 0 + }, + { + "name": "model.layers.1.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 2097152 + }, + { + "name": "model.layers.1.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 10485760 + }, + { + "name": "model.layers.1.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 11534336 + }, + { + "name": "model.layers.10.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 11538432 + }, + { + "name": "model.layers.10.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 13635584 + }, + { + "name": "model.layers.10.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 13897728 + }, + { + "name": "model.layers.10.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 18092032 + }, + { + "name": "model.layers.10.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18616320 + }, + { + "name": "model.layers.10.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 18620416 + }, + { + "name": "model.layers.10.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 20717568 + }, + { + "name": "model.layers.10.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 29106176 + }, + { + "name": "model.layers.10.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 30154752 + }, + { + "name": "model.layers.11.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 30158848 + }, + { + "name": "model.layers.11.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 32256000 + } + ], + "md5sum": "6d3deaf75e9ccbbd2c9d20c51408586b" + }, + { + "dataPath": "params_shard_4.bin", + "format": "raw-shard", + "nbytes": 33038336, + "records": [ + { + "name": "model.layers.11.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 0 + }, + { + "name": "model.layers.11.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 4194304 + }, + { + "name": "model.layers.11.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 4718592 + }, + { + "name": "model.layers.11.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 4722688 + }, + { + "name": "model.layers.11.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 21499904 + }, + { + "name": "model.layers.11.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23597056 + }, + { + "name": "model.layers.11.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 31985664 + }, + { + "name": "model.layers.11.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33034240 + } + ], + "md5sum": "a127e1b009a46ec59c68500380550d47" + }, + { + "dataPath": "params_shard_5.bin", + "format": "raw-shard", + "nbytes": 16777216, + "records": [ + { + "name": "model.layers.2.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 0 + } + ], + "md5sum": "0ee2da82994f047cca7774c7ff9a01c5" + }, + { + "dataPath": "params_shard_6.bin", + "format": "raw-shard", + "nbytes": 33034240, + "records": [ + { + "name": "model.layers.12.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 0 + }, + { + "name": "model.layers.12.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 2097152 + }, + { + "name": "model.layers.12.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 2359296 + }, + { + "name": "model.layers.12.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 6553600 + }, + { + "name": "model.layers.12.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 7077888 + }, + { + "name": "model.layers.12.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 23855104 + }, + { + "name": "model.layers.2.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 25952256 + }, + { + "name": "model.layers.2.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 28049408 + }, + { + "name": "model.layers.2.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 28311552 + }, + { + "name": "model.layers.2.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 32505856 + }, + { + "name": "model.layers.2.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33030144 + } + ], + "md5sum": "baf5749f017706a655f21b9dc35d9590" + }, + { + "dataPath": "params_shard_7.bin", + "format": "raw-shard", + "nbytes": 16777216, + "records": [ + { + "name": "model.layers.3.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 0 + } + ], + "md5sum": "f18e1f5460765d8397529d4333b90afe" + }, + { + "dataPath": "params_shard_8.bin", + "format": "raw-shard", + "nbytes": 32518144, + "records": [ + { + "name": "model.layers.2.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 0 + }, + { + "name": "model.layers.2.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 2097152 + }, + { + "name": "model.layers.2.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 10485760 + }, + { + "name": "model.layers.2.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 11534336 + }, + { + "name": "model.layers.3.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 11538432 + }, + { + "name": "model.layers.3.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 13635584 + }, + { + "name": "model.layers.3.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 13897728 + }, + { + "name": "model.layers.3.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 18092032 + }, + { + "name": "model.layers.3.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18616320 + }, + { + "name": "model.layers.3.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 18620416 + }, + { + "name": "model.layers.3.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 20717568 + }, + { + "name": "model.layers.3.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 29106176 + }, + { + "name": "model.layers.3.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 30154752 + }, + { + "name": "model.layers.4.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 30158848 + }, + { + "name": "model.layers.4.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 32256000 + } + ], + "md5sum": "92a3c7ef838fb432156d8b46e363299f" + }, + { + "dataPath": "params_shard_9.bin", + "format": "raw-shard", + "nbytes": 33038336, + "records": [ + { + "name": "model.layers.4.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 0 + }, + { + "name": "model.layers.4.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 4194304 + }, + { + "name": "model.layers.4.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 4718592 + }, + { + "name": "model.layers.4.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 4722688 + }, + { + "name": "model.layers.4.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 21499904 + }, + { + "name": "model.layers.4.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23597056 + }, + { + "name": "model.layers.4.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 31985664 + }, + { + "name": "model.layers.4.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33034240 + } + ], + "md5sum": "d5aea50d5fe99aae844914286db3a8ed" + }, + { + "dataPath": "params_shard_10.bin", + "format": "raw-shard", + "nbytes": 25956352, + "records": [ + { + "name": "model.layers.5.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 0 + }, + { + "name": "model.layers.5.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 2097152 + }, + { + "name": "model.layers.5.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 2359296 + }, + { + "name": "model.layers.5.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 6553600 + }, + { + "name": "model.layers.5.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 7077888 + }, + { + "name": "model.layers.5.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 7081984 + }, + { + "name": "model.layers.5.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 23859200 + } + ], + "md5sum": "1e043d657e519fe0e8e8bf1ebf19dbb4" + }, + { + "dataPath": "params_shard_11.bin", + "format": "raw-shard", + "nbytes": 33300480, + "records": [ + { + "name": "model.layers.5.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "model.layers.5.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "model.layers.5.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9437184 + }, + { + "name": "model.layers.6.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 9441280 + }, + { + "name": "model.layers.6.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 11538432 + }, + { + "name": "model.layers.6.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 11800576 + }, + { + "name": "model.layers.6.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 15994880 + }, + { + "name": "model.layers.6.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 16519168 + }, + { + "name": "model.layers.6.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 16523264 + } + ], + "md5sum": "2f0654640cdb6350fe8ea65ed5ddd112" + }, + { + "dataPath": "params_shard_12.bin", + "format": "raw-shard", + "nbytes": 16777216, + "records": [ + { + "name": "model.layers.7.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 0 + } + ], + "md5sum": "eeb2d5dc7018c7f5cdc7d4fab4fb6065" + }, + { + "dataPath": "params_shard_13.bin", + "format": "raw-shard", + "nbytes": 32518144, + "records": [ + { + "name": "model.layers.6.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 0 + }, + { + "name": "model.layers.6.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 2097152 + }, + { + "name": "model.layers.6.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 10485760 + }, + { + "name": "model.layers.6.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 11534336 + }, + { + "name": "model.layers.7.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 11538432 + }, + { + "name": "model.layers.7.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 13635584 + }, + { + "name": "model.layers.7.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 13897728 + }, + { + "name": "model.layers.7.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 18092032 + }, + { + "name": "model.layers.7.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18616320 + }, + { + "name": "model.layers.7.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 18620416 + }, + { + "name": "model.layers.7.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 20717568 + }, + { + "name": "model.layers.7.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 29106176 + }, + { + "name": "model.layers.7.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 30154752 + }, + { + "name": "model.layers.8.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 30158848 + }, + { + "name": "model.layers.8.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 32256000 + } + ], + "md5sum": "6248d9d4a3e15b049232b58b8f035bcd" + }, + { + "dataPath": "params_shard_14.bin", + "format": "raw-shard", + "nbytes": 33038336, + "records": [ + { + "name": "model.layers.8.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 0 + }, + { + "name": "model.layers.8.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 4194304 + }, + { + "name": "model.layers.8.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 4718592 + }, + { + "name": "model.layers.8.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 4722688 + }, + { + "name": "model.layers.8.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 21499904 + }, + { + "name": "model.layers.8.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23597056 + }, + { + "name": "model.layers.8.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 31985664 + }, + { + "name": "model.layers.8.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33034240 + } + ], + "md5sum": "6422716453104c4bab4563e604a164f8" + }, + { + "dataPath": "params_shard_15.bin", + "format": "raw-shard", + "nbytes": 25956352, + "records": [ + { + "name": "model.layers.9.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 0 + }, + { + "name": "model.layers.9.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 2097152 + }, + { + "name": "model.layers.9.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 2359296 + }, + { + "name": "model.layers.9.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 6553600 + }, + { + "name": "model.layers.9.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 7077888 + }, + { + "name": "model.layers.9.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 7081984 + }, + { + "name": "model.layers.9.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 23859200 + } + ], + "md5sum": "20b417c323d2aeef269e71f4f86f8ed0" + }, + { + "dataPath": "params_shard_16.bin", + "format": "raw-shard", + "nbytes": 94765056, + "records": [ + { + "name": "model.tok_embeddings.q_weight", + "shape": [ + 92544, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 94765056, + "byteOffset": 0 + } + ], + "md5sum": "a73e3b49851508b049183142411b54c4" + }, + { + "dataPath": "params_shard_17.bin", + "format": "raw-shard", + "nbytes": 33091584, + "records": [ + { + "name": "model.layers.9.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "model.layers.9.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "model.layers.9.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9437184 + }, + { + "name": "model.tok_embeddings.q_scale", + "shape": [ + 92544, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 11845632, + "byteOffset": 9441280 + }, + { + "name": "model.layers.12.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 21286912 + }, + { + "name": "model.layers.12.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 21291008 + }, + { + "name": "model.layers.12.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 29679616 + }, + { + "name": "model.layers.12.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 30728192 + }, + { + "name": "model.layers.13.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 30732288 + }, + { + "name": "model.layers.13.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 32829440 + } + ], + "md5sum": "5a45d367d750c4b0e369ac5a8ed2b41d" + }, + { + "dataPath": "params_shard_18.bin", + "format": "raw-shard", + "nbytes": 33038336, + "records": [ + { + "name": "model.layers.13.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 0 + }, + { + "name": "model.layers.13.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 4194304 + }, + { + "name": "model.layers.13.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 4718592 + }, + { + "name": "model.layers.13.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 4722688 + }, + { + "name": "model.layers.13.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 21499904 + }, + { + "name": "model.layers.13.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23597056 + }, + { + "name": "model.layers.13.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 31985664 + }, + { + "name": "model.layers.13.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33034240 + } + ], + "md5sum": "13e4dc09e3ec2748949f7bd5dc727f34" + }, + { + "dataPath": "params_shard_19.bin", + "format": "raw-shard", + "nbytes": 25956352, + "records": [ + { + "name": "model.layers.14.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 0 + }, + { + "name": "model.layers.14.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 2097152 + }, + { + "name": "model.layers.14.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 2359296 + }, + { + "name": "model.layers.14.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 6553600 + }, + { + "name": "model.layers.14.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 7077888 + }, + { + "name": "model.layers.14.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 7081984 + }, + { + "name": "model.layers.14.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 23859200 + } + ], + "md5sum": "f0a9961acf3ba5a5293eaf25b97bec72" + }, + { + "dataPath": "params_shard_20.bin", + "format": "raw-shard", + "nbytes": 33300480, + "records": [ + { + "name": "model.layers.14.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "model.layers.14.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "model.layers.14.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9437184 + }, + { + "name": "model.layers.15.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 9441280 + }, + { + "name": "model.layers.15.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 11538432 + }, + { + "name": "model.layers.15.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 11800576 + }, + { + "name": "model.layers.15.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 15994880 + }, + { + "name": "model.layers.15.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 16519168 + }, + { + "name": "model.layers.15.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 16523264 + } + ], + "md5sum": "0d1857fe93acd20d78ff83a773fa8f56" + }, + { + "dataPath": "params_shard_21.bin", + "format": "raw-shard", + "nbytes": 16777216, + "records": [ + { + "name": "model.layers.16.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 0 + } + ], + "md5sum": "21ad26a976b5a3426a8902161fcee074" + }, + { + "dataPath": "params_shard_22.bin", + "format": "raw-shard", + "nbytes": 32518144, + "records": [ + { + "name": "model.layers.15.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 0 + }, + { + "name": "model.layers.15.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 2097152 + }, + { + "name": "model.layers.15.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 10485760 + }, + { + "name": "model.layers.15.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 11534336 + }, + { + "name": "model.layers.16.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 11538432 + }, + { + "name": "model.layers.16.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 13635584 + }, + { + "name": "model.layers.16.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 13897728 + }, + { + "name": "model.layers.16.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 18092032 + }, + { + "name": "model.layers.16.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18616320 + }, + { + "name": "model.layers.16.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 18620416 + }, + { + "name": "model.layers.16.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 20717568 + }, + { + "name": "model.layers.16.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 29106176 + }, + { + "name": "model.layers.16.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 30154752 + }, + { + "name": "model.layers.17.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 30158848 + }, + { + "name": "model.layers.17.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 32256000 + } + ], + "md5sum": "53b3bf10b0bfe514abd219ea55f4510b" + }, + { + "dataPath": "params_shard_23.bin", + "format": "raw-shard", + "nbytes": 33038336, + "records": [ + { + "name": "model.layers.17.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 0 + }, + { + "name": "model.layers.17.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 4194304 + }, + { + "name": "model.layers.17.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 4718592 + }, + { + "name": "model.layers.17.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 4722688 + }, + { + "name": "model.layers.17.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 21499904 + }, + { + "name": "model.layers.17.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23597056 + }, + { + "name": "model.layers.17.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 31985664 + }, + { + "name": "model.layers.17.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33034240 + } + ], + "md5sum": "3e39970d677e0979bc3edd422553e031" + }, + { + "dataPath": "params_shard_24.bin", + "format": "raw-shard", + "nbytes": 25956352, + "records": [ + { + "name": "model.layers.18.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 0 + }, + { + "name": "model.layers.18.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 2097152 + }, + { + "name": "model.layers.18.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 2359296 + }, + { + "name": "model.layers.18.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 6553600 + }, + { + "name": "model.layers.18.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 7077888 + }, + { + "name": "model.layers.18.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 7081984 + }, + { + "name": "model.layers.18.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 23859200 + } + ], + "md5sum": "0622afff97ca80e9b80f33b037f0b8a8" + }, + { + "dataPath": "params_shard_25.bin", + "format": "raw-shard", + "nbytes": 33300480, + "records": [ + { + "name": "model.layers.18.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "model.layers.18.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "model.layers.18.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9437184 + }, + { + "name": "model.layers.19.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 9441280 + }, + { + "name": "model.layers.19.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 11538432 + }, + { + "name": "model.layers.19.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 11800576 + }, + { + "name": "model.layers.19.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 15994880 + }, + { + "name": "model.layers.19.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 16519168 + }, + { + "name": "model.layers.19.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 16523264 + } + ], + "md5sum": "ff7e5302501238ce8e40772a82042969" + }, + { + "dataPath": "params_shard_26.bin", + "format": "raw-shard", + "nbytes": 16777216, + "records": [ + { + "name": "model.layers.20.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 0 + } + ], + "md5sum": "b570f7e2a24e0b9e5e982b3718752c98" + }, + { + "dataPath": "params_shard_27.bin", + "format": "raw-shard", + "nbytes": 32518144, + "records": [ + { + "name": "model.layers.19.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 0 + }, + { + "name": "model.layers.19.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 2097152 + }, + { + "name": "model.layers.19.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 10485760 + }, + { + "name": "model.layers.19.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 11534336 + }, + { + "name": "model.layers.20.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 11538432 + }, + { + "name": "model.layers.20.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 13635584 + }, + { + "name": "model.layers.20.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 13897728 + }, + { + "name": "model.layers.20.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 18092032 + }, + { + "name": "model.layers.20.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 18616320 + }, + { + "name": "model.layers.20.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 18620416 + }, + { + "name": "model.layers.20.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 20717568 + }, + { + "name": "model.layers.20.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 29106176 + }, + { + "name": "model.layers.20.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 30154752 + }, + { + "name": "model.layers.21.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 30158848 + }, + { + "name": "model.layers.21.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 32256000 + } + ], + "md5sum": "d6c40bba2b678bb1ab6c4d2dd511cee9" + }, + { + "dataPath": "params_shard_28.bin", + "format": "raw-shard", + "nbytes": 33038336, + "records": [ + { + "name": "model.layers.21.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 0 + }, + { + "name": "model.layers.21.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 4194304 + }, + { + "name": "model.layers.21.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 4718592 + }, + { + "name": "model.layers.21.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 4722688 + }, + { + "name": "model.layers.21.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 21499904 + }, + { + "name": "model.layers.21.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 23597056 + }, + { + "name": "model.layers.21.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 31985664 + }, + { + "name": "model.layers.21.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 33034240 + } + ], + "md5sum": "21056282748c3460f741069bc8b376df" + }, + { + "dataPath": "params_shard_29.bin", + "format": "raw-shard", + "nbytes": 25956352, + "records": [ + { + "name": "model.layers.22.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 0 + }, + { + "name": "model.layers.22.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 2097152 + }, + { + "name": "model.layers.22.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 2359296 + }, + { + "name": "model.layers.22.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 6553600 + }, + { + "name": "model.layers.22.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 7077888 + }, + { + "name": "model.layers.22.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 7081984 + }, + { + "name": "model.layers.22.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 23859200 + } + ], + "md5sum": "eb9d208b74ae8e431af56f9744347b8c" + }, + { + "dataPath": "params_shard_30.bin", + "format": "raw-shard", + "nbytes": 33300480, + "records": [ + { + "name": "model.layers.22.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "model.layers.22.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "model.layers.22.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 9437184 + }, + { + "name": "model.layers.23.attention.wo.q_weight", + "shape": [ + 2048, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 9441280 + }, + { + "name": "model.layers.23.attention.wo.q_scale", + "shape": [ + 2048, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 262144, + "byteOffset": 11538432 + }, + { + "name": "model.layers.23.attention.wqkv.q_weight", + "shape": [ + 4096, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4194304, + "byteOffset": 11800576 + }, + { + "name": "model.layers.23.attention.wqkv.q_scale", + "shape": [ + 4096, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 524288, + "byteOffset": 15994880 + }, + { + "name": "model.layers.23.attention_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 16519168 + }, + { + "name": "model.layers.23.feed_forward.gate_up_proj.q_weight", + "shape": [ + 16384, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 16777216, + "byteOffset": 16523264 + } + ], + "md5sum": "f3e79e9aa185d09a7fda0064826b2e4e" + }, + { + "dataPath": "params_shard_31.bin", + "format": "raw-shard", + "nbytes": 94765056, + "records": [ + { + "name": "output.q_weight", + "shape": [ + 92544, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 94765056, + "byteOffset": 0 + } + ], + "md5sum": "064f77babd659a8e5cffa486e796e41a" + }, + { + "dataPath": "params_shard_32.bin", + "format": "raw-shard", + "nbytes": 23388160, + "records": [ + { + "name": "model.layers.23.feed_forward.gate_up_proj.q_scale", + "shape": [ + 16384, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 2097152, + "byteOffset": 0 + }, + { + "name": "model.layers.23.feed_forward.w2.q_weight", + "shape": [ + 2048, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 8388608, + "byteOffset": 2097152 + }, + { + "name": "model.layers.23.feed_forward.w2.q_scale", + "shape": [ + 2048, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1048576, + "byteOffset": 10485760 + }, + { + "name": "model.layers.23.ffn_norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 11534336 + }, + { + "name": "model.norm.weight", + "shape": [ + 2048 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4096, + "byteOffset": 11538432 + }, + { + "name": "output.q_scale", + "shape": [ + 92544, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 11845632, + "byteOffset": 11542528 + } + ], + "md5sum": "25c3b22e8b1b199b6c3268fcaae64b62" + } + ] +} \ No newline at end of file