diff --git "a/ndarray-cache.json" "b/ndarray-cache.json" new file mode 100644--- /dev/null +++ "b/ndarray-cache.json" @@ -0,0 +1,3697 @@ +{ + "metadata": { + "ParamSize": 315, + "ParamBytes": 1348379136.0, + "BitsPerParam": 4.076776454709711 + }, + "records": [ + { + "dataPath": "params_shard_0.bin", + "format": "raw-shard", + "nbytes": 294912000, + "records": [ + { + "name": "model.embed_tokens.q_weight", + "shape": [ + 256000, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 294912000, + "byteOffset": 0 + } + ], + "md5sum": "407df35db25b64d836eada4ebe0262ef" + }, + { + "dataPath": "params_shard_1.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.0.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "1114d35d6fc2112a521dda40a61b7b8f" + }, + { + "dataPath": "params_shard_2.bin", + "format": "raw-shard", + "nbytes": 28150272, + "records": [ + { + "name": "model.embed_tokens.q_scale", + "shape": [ + 256000, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 9216000, + "byteOffset": 0 + }, + { + "name": "model.layers.0.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 9216000 + }, + { + "name": "model.layers.0.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 9220608 + }, + { + "name": "model.layers.0.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 19837440 + }, + { + "name": "model.layers.0.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 20169216 + }, + { + "name": "model.layers.0.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 20832768 + }, + { + "name": "model.layers.0.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 20837376 + }, + { + "name": "model.layers.0.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 20841984 + }, + { + "name": "model.layers.0.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 20846592 + }, + { + "name": "model.layers.0.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 25565184 + }, + { + "name": "model.layers.0.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 25712640 + }, + { + "name": "model.layers.0.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 28071936 + }, + { + "name": "model.layers.1.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 28145664 + } + ], + "md5sum": "c6b97a239126c320f9c230cb2ba6a8b6" + }, + { + "dataPath": "params_shard_3.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.1.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.1.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.1.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.1.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 32182272 + }, + { + "name": "model.layers.1.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32845824 + }, + { + "name": "model.layers.1.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32850432 + }, + { + "name": "model.layers.1.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32855040 + } + ], + "md5sum": "5a71a0b4fbfde05829f93fa28b1fbf03" + }, + { + "dataPath": "params_shard_4.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.10.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "f67630250dace666a594d6db3e8736eb" + }, + { + "dataPath": "params_shard_5.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.1.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.1.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.1.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.1.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 7225344 + }, + { + "name": "model.layers.10.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 7299072 + }, + { + "name": "model.layers.10.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.10.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.10.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 18252288 + }, + { + "name": "model.layers.10.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18915840 + }, + { + "name": "model.layers.10.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18920448 + }, + { + "name": "model.layers.10.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18925056 + }, + { + "name": "model.layers.10.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.10.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.10.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.10.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 26155008 + }, + { + "name": "model.layers.11.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26228736 + } + ], + "md5sum": "21d3dd9dc290a6fe1e6eb565589640c8" + }, + { + "dataPath": "params_shard_6.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.11.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.11.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.11.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.11.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 32182272 + }, + { + "name": "model.layers.11.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32845824 + }, + { + "name": "model.layers.11.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32850432 + }, + { + "name": "model.layers.11.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32855040 + } + ], + "md5sum": "bb7d77cf0c252e0c5f046a8d0bb23712" + }, + { + "dataPath": "params_shard_7.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.12.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "963f331f4dd691e3002687d12266802e" + }, + { + "dataPath": "params_shard_8.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.11.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.11.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.11.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.11.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 7225344 + }, + { + "name": "model.layers.12.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 7299072 + }, + { + "name": "model.layers.12.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.12.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.12.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 18252288 + }, + { + "name": "model.layers.12.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18915840 + }, + { + "name": "model.layers.12.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18920448 + }, + { + "name": "model.layers.12.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18925056 + }, + { + "name": "model.layers.12.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.12.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.12.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.12.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 26155008 + }, + { + "name": "model.layers.13.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26228736 + } + ], + "md5sum": "070cdc703e66d41c13a65dae71f7cc31" + }, + { + "dataPath": "params_shard_9.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.13.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.13.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.13.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.13.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 32182272 + }, + { + "name": "model.layers.13.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32845824 + }, + { + "name": "model.layers.13.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32850432 + }, + { + "name": "model.layers.13.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32855040 + } + ], + "md5sum": "c8c7ff49dc00bf0525ad8c6d0dabc567" + }, + { + "dataPath": "params_shard_10.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.14.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "93092b87ed4c4b43fd153c3f4f9ae5fd" + }, + { + "dataPath": "params_shard_11.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.13.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.13.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.13.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.13.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 7225344 + }, + { + "name": "model.layers.14.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 7299072 + }, + { + "name": "model.layers.14.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.14.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.14.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 18252288 + }, + { + "name": "model.layers.14.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18915840 + }, + { + "name": "model.layers.14.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18920448 + }, + { + "name": "model.layers.14.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18925056 + }, + { + "name": "model.layers.14.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.14.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.14.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.14.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 26155008 + }, + { + "name": "model.layers.15.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26228736 + } + ], + "md5sum": "30b386d3bb0a022b9c7c89913761d11b" + }, + { + "dataPath": "params_shard_12.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.15.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.15.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.15.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.15.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 32182272 + }, + { + "name": "model.layers.15.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32845824 + }, + { + "name": "model.layers.15.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32850432 + }, + { + "name": "model.layers.15.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32855040 + } + ], + "md5sum": "f75abfc384cb29eeb455a572682d2913" + }, + { + "dataPath": "params_shard_13.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.16.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "bfc0b5de8601fa42acb0b596516e7737" + }, + { + "dataPath": "params_shard_14.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.15.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.15.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.15.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.15.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 7225344 + }, + { + "name": "model.layers.16.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 7299072 + }, + { + "name": "model.layers.16.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.16.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.16.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 18252288 + }, + { + "name": "model.layers.16.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18915840 + }, + { + "name": "model.layers.16.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18920448 + }, + { + "name": "model.layers.16.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18925056 + }, + { + "name": "model.layers.16.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.16.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.16.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.16.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 26155008 + }, + { + "name": "model.layers.17.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26228736 + } + ], + "md5sum": "1bc300878bf5fd88fc4180de5c86ab9d" + }, + { + "dataPath": "params_shard_15.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.17.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.17.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.17.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.17.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 32182272 + }, + { + "name": "model.layers.17.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32845824 + }, + { + "name": "model.layers.17.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32850432 + }, + { + "name": "model.layers.17.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32855040 + } + ], + "md5sum": "90056b30a2f5a6137ba6800505d1345e" + }, + { + "dataPath": "params_shard_16.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.18.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "190ff5aaf45dadc1db769d1f4d311f94" + }, + { + "dataPath": "params_shard_17.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.17.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.17.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.17.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.17.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 7225344 + }, + { + "name": "model.layers.18.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 7299072 + }, + { + "name": "model.layers.18.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.18.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.18.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 18252288 + }, + { + "name": "model.layers.18.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18915840 + }, + { + "name": "model.layers.18.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18920448 + }, + { + "name": "model.layers.18.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18925056 + }, + { + "name": "model.layers.18.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.18.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.18.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.18.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 26155008 + }, + { + "name": "model.layers.19.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26228736 + } + ], + "md5sum": "2fcb0b7f71fe0a9644a3861e98e819ce" + }, + { + "dataPath": "params_shard_18.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.19.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.19.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.19.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.19.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 32182272 + }, + { + "name": "model.layers.19.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32845824 + }, + { + "name": "model.layers.19.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32850432 + }, + { + "name": "model.layers.19.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32855040 + } + ], + "md5sum": "b46e5b15d9402e8fd4ff464965e57fba" + }, + { + "dataPath": "params_shard_19.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.2.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "dccdd229386c032b672d17ec830409e7" + }, + { + "dataPath": "params_shard_20.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.19.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.19.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.19.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.19.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 7225344 + }, + { + "name": "model.layers.2.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 7299072 + }, + { + "name": "model.layers.2.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.2.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.2.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 18252288 + }, + { + "name": "model.layers.2.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18915840 + }, + { + "name": "model.layers.2.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18920448 + }, + { + "name": "model.layers.2.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18925056 + }, + { + "name": "model.layers.2.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.2.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.2.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.2.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 26155008 + }, + { + "name": "model.layers.20.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26228736 + } + ], + "md5sum": "ee2f56a561534e91d2f863c1221dbf5c" + }, + { + "dataPath": "params_shard_21.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.20.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.20.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.20.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.20.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 32182272 + }, + { + "name": "model.layers.20.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32845824 + }, + { + "name": "model.layers.20.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32850432 + }, + { + "name": "model.layers.20.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32855040 + } + ], + "md5sum": "2873626e03071d12ef59a061e2648583" + }, + { + "dataPath": "params_shard_22.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.21.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "d7aed001e198d7b9984c28b571e512f5" + }, + { + "dataPath": "params_shard_23.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.20.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.20.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.20.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.20.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 7225344 + }, + { + "name": "model.layers.21.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 7299072 + }, + { + "name": "model.layers.21.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.21.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.21.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 18252288 + }, + { + "name": "model.layers.21.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18915840 + }, + { + "name": "model.layers.21.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18920448 + }, + { + "name": "model.layers.21.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18925056 + }, + { + "name": "model.layers.21.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.21.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.21.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.21.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 26155008 + }, + { + "name": "model.layers.22.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26228736 + } + ], + "md5sum": "c7dfb20d9df5cd5bf2822d78d359deab" + }, + { + "dataPath": "params_shard_24.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.22.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.22.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.22.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.22.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 32182272 + }, + { + "name": "model.layers.22.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32845824 + }, + { + "name": "model.layers.22.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32850432 + }, + { + "name": "model.layers.22.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32855040 + } + ], + "md5sum": "926c50510b0b77504590714e6bc310cb" + }, + { + "dataPath": "params_shard_25.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.23.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "3fa781e0ccace7e57431263fbce7dc68" + }, + { + "dataPath": "params_shard_26.bin", + "format": "raw-shard", + "nbytes": 31099392, + "records": [ + { + "name": "model.layers.22.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.22.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.22.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.22.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 7225344 + }, + { + "name": "model.layers.23.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 7299072 + }, + { + "name": "model.layers.23.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.23.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.23.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 18252288 + }, + { + "name": "model.layers.23.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18915840 + }, + { + "name": "model.layers.23.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18920448 + }, + { + "name": "model.layers.23.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18925056 + }, + { + "name": "model.layers.23.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.23.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.23.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.23.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 26155008 + }, + { + "name": "model.layers.24.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 26228736 + }, + { + "name": "model.layers.24.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 30947328 + }, + { + "name": "model.layers.3.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 31094784 + } + ], + "md5sum": "2fb9bd310d25cd4c40382b3ee555543d" + }, + { + "dataPath": "params_shard_27.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.3.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.3.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.3.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.3.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 32182272 + }, + { + "name": "model.layers.3.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32845824 + }, + { + "name": "model.layers.3.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32850432 + }, + { + "name": "model.layers.3.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32855040 + } + ], + "md5sum": "85984804f37a6d26d33fd5c7bddace84" + }, + { + "dataPath": "params_shard_28.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.4.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "87f6e07dfd2592ead68f4d3cef6b5282" + }, + { + "dataPath": "params_shard_29.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.3.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.3.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.3.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.3.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 7225344 + }, + { + "name": "model.layers.4.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 7299072 + }, + { + "name": "model.layers.4.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.4.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.4.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 18252288 + }, + { + "name": "model.layers.4.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18915840 + }, + { + "name": "model.layers.4.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18920448 + }, + { + "name": "model.layers.4.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18925056 + }, + { + "name": "model.layers.4.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.4.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.4.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.4.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 26155008 + }, + { + "name": "model.layers.5.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26228736 + } + ], + "md5sum": "f0c04a4ebb01241e3fc158eff7e6c220" + }, + { + "dataPath": "params_shard_30.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.5.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.5.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.5.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.5.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 32182272 + }, + { + "name": "model.layers.5.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32845824 + }, + { + "name": "model.layers.5.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32850432 + }, + { + "name": "model.layers.5.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32855040 + } + ], + "md5sum": "b38eef29ec602376171937994ded1167" + }, + { + "dataPath": "params_shard_31.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.6.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "1e9ce621be14acd1b56f5e4c0b8f1bea" + }, + { + "dataPath": "params_shard_32.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.5.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.5.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.5.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.5.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 7225344 + }, + { + "name": "model.layers.6.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 7299072 + }, + { + "name": "model.layers.6.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.6.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.6.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 18252288 + }, + { + "name": "model.layers.6.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18915840 + }, + { + "name": "model.layers.6.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18920448 + }, + { + "name": "model.layers.6.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18925056 + }, + { + "name": "model.layers.6.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.6.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.6.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.6.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 26155008 + }, + { + "name": "model.layers.7.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26228736 + } + ], + "md5sum": "03526eaf0f7797c0f91e6a14ee060931" + }, + { + "dataPath": "params_shard_33.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.7.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.7.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.7.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.7.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 32182272 + }, + { + "name": "model.layers.7.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32845824 + }, + { + "name": "model.layers.7.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32850432 + }, + { + "name": "model.layers.7.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32855040 + } + ], + "md5sum": "8ebba7d1d76c324963f215b75e2ec767" + }, + { + "dataPath": "params_shard_34.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.8.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "a1f3e4b9945302dd1efe9d76d0ac9e8d" + }, + { + "dataPath": "params_shard_35.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.7.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.7.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.7.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.7.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 7225344 + }, + { + "name": "model.layers.8.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 7299072 + }, + { + "name": "model.layers.8.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.8.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.8.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 18252288 + }, + { + "name": "model.layers.8.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18915840 + }, + { + "name": "model.layers.8.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18920448 + }, + { + "name": "model.layers.8.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18925056 + }, + { + "name": "model.layers.8.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.8.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.8.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.8.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 26155008 + }, + { + "name": "model.layers.9.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 26228736 + } + ], + "md5sum": "9bb24eaf6c7496b71462817d5176bbf0" + }, + { + "dataPath": "params_shard_36.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.9.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.9.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.9.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.9.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 32182272 + }, + { + "name": "model.layers.9.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32845824 + }, + { + "name": "model.layers.9.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32850432 + }, + { + "name": "model.layers.9.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32855040 + } + ], + "md5sum": "07499b356cebd57c5d33f9ff558fcfdb" + }, + { + "dataPath": "params_shard_37.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.24.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "f88e7295c088a58091cd4fe3d5cec20d" + }, + { + "dataPath": "params_shard_38.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.25.mlp.gate_up_proj.q_weight", + "shape": [ + 288, + 18432 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "c3aa207eaf00c36cf1ff244f375e2f42" + }, + { + "dataPath": "params_shard_39.bin", + "format": "raw-shard", + "nbytes": 32993280, + "records": [ + { + "name": "model.layers.9.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.9.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.9.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.9.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 7225344 + }, + { + "name": "model.layers.24.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 7299072 + }, + { + "name": "model.layers.24.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.24.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.24.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 18252288 + }, + { + "name": "model.layers.24.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18915840 + }, + { + "name": "model.layers.24.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18920448 + }, + { + "name": "model.layers.24.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 18925056 + }, + { + "name": "model.layers.24.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 18929664 + }, + { + "name": "model.layers.24.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 21288960 + }, + { + "name": "model.layers.25.input_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 21362688 + }, + { + "name": "model.layers.25.mlp.down_proj.q_weight", + "shape": [ + 1152, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 21367296 + }, + { + "name": "model.layers.25.mlp.down_proj.q_scale", + "shape": [ + 72, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 31984128 + }, + { + "name": "model.layers.25.mlp.gate_up_proj.q_scale", + "shape": [ + 18, + 18432 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 663552, + "byteOffset": 32315904 + }, + { + "name": "model.layers.25.post_attention_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32979456 + }, + { + "name": "model.layers.25.post_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32984064 + }, + { + "name": "model.layers.25.pre_feedforward_layernorm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 32988672 + } + ], + "md5sum": "aeb2221c0029d0d16efe9a8db88e95d5" + }, + { + "dataPath": "params_shard_40.bin", + "format": "raw-shard", + "nbytes": 7303680, + "records": [ + { + "name": "model.layers.25.self_attn.qkv_proj.q_weight", + "shape": [ + 288, + 4096 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.25.self_attn.qkv_proj.q_scale", + "shape": [ + 18, + 4096 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.25.self_attn.o_proj.q_weight", + "shape": [ + 256, + 2304 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.25.self_attn.o_proj.q_scale", + "shape": [ + 16, + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 7225344 + }, + { + "name": "model.norm.weight", + "shape": [ + 2304 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 4608, + "byteOffset": 7299072 + } + ], + "md5sum": "c4716d6d513ce243bf46d29b6b753050" + } + ] +} \ No newline at end of file