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": "a17ddd4ac7972efd6b46a7fee9296a06" + }, + { + "dataPath": "params_shard_1.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.0.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "cbaa8a25f1b38d7377e9e2ab0625a0e5" + }, + { + "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": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 9220608 + }, + { + "name": "model.layers.0.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 19837440 + }, + { + "name": "model.layers.0.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 20846592 + }, + { + "name": "model.layers.0.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 25565184 + }, + { + "name": "model.layers.0.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 25712640 + }, + { + "name": "model.layers.0.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": "f17f44b7a57caa46ef50150c09687b68" + }, + { + "dataPath": "params_shard_3.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.1.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.1.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.1.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.1.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": "0ff2b5ccae9546b1616c4211ec5d56b6" + }, + { + "dataPath": "params_shard_4.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.10.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "af4db9d3c8dfb4dd2c937dcaee45d71f" + }, + { + "dataPath": "params_shard_5.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.1.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.1.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.1.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.1.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.10.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.10.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.10.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.10.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.10.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": "8255dafb9fa7fa45303738c7c3392a44" + }, + { + "dataPath": "params_shard_6.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.11.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.11.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.11.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.11.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": "bbb0dd7adc8acdcc9c020c6d93c6f5af" + }, + { + "dataPath": "params_shard_7.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.12.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "8b771152060edcf945c0c699e12f2c65" + }, + { + "dataPath": "params_shard_8.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.11.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.11.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.11.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.11.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.12.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.12.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.12.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.12.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.12.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": "f3e03474cd17c17b4b1b84de84c319ae" + }, + { + "dataPath": "params_shard_9.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.13.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.13.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.13.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.13.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": "7965a342e534f15113fd52b5ccb99123" + }, + { + "dataPath": "params_shard_10.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.14.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "0030d4916b4fb74886ee66102b02db12" + }, + { + "dataPath": "params_shard_11.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.13.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.13.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.13.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.13.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.14.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.14.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.14.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.14.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.14.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": "fd0d736d60c67f71a900edda60e427fa" + }, + { + "dataPath": "params_shard_12.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.15.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.15.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.15.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.15.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": "e57c7025c8d0328f02ea398ec6e30af8" + }, + { + "dataPath": "params_shard_13.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.16.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "31ef071e9af2d1a74c340d95b823ac00" + }, + { + "dataPath": "params_shard_14.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.15.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.15.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.15.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.15.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.16.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.16.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.16.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.16.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.16.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": "8a823c6445e7e622ca35b06fef941997" + }, + { + "dataPath": "params_shard_15.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.17.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.17.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.17.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.17.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": "9ca0ee4a889bdc918df7d9dfd6b1bbc7" + }, + { + "dataPath": "params_shard_16.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.18.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "f563df377a8315275b670cbff2ad5754" + }, + { + "dataPath": "params_shard_17.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.17.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.17.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.17.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.17.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.18.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.18.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.18.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.18.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.18.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": "121ca885c13970a5278e1754e834fed7" + }, + { + "dataPath": "params_shard_18.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.19.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.19.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.19.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.19.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": "eead420d35de4baa484f451fdd806c32" + }, + { + "dataPath": "params_shard_19.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.2.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "3ea3fcaa9f623eee0c072f8bfb4f3e46" + }, + { + "dataPath": "params_shard_20.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.19.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.19.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.19.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.19.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.2.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.2.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.2.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.2.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.2.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": "37d518d85a1d39d93ae8913292da07e5" + }, + { + "dataPath": "params_shard_21.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.20.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.20.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.20.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.20.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": "d19354d84d350c949694344f4bce7d75" + }, + { + "dataPath": "params_shard_22.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.21.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "dd5d6fc66f6ccf6fab336a9045dc1b6c" + }, + { + "dataPath": "params_shard_23.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.20.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.20.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.20.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.20.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.21.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.21.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.21.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.21.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.21.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": "9f0a760d533157391d0b1efd2207b095" + }, + { + "dataPath": "params_shard_24.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.22.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.22.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.22.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.22.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": "27a3090a221eff90f3c69ac535a60892" + }, + { + "dataPath": "params_shard_25.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.23.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "40033eb2772efa34eb7fdd18ac8b7b07" + }, + { + "dataPath": "params_shard_26.bin", + "format": "raw-shard", + "nbytes": 31099392, + "records": [ + { + "name": "model.layers.22.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.22.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.22.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.22.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.23.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.23.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.23.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.23.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.23.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 73728, + "byteOffset": 26155008 + }, + { + "name": "model.layers.24.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 26228736 + }, + { + "name": "model.layers.24.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "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": "dca1dc66a9a527dae35d06d6470cc322" + }, + { + "dataPath": "params_shard_27.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.3.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.3.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.3.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.3.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": "3aadb8876872911cc270847c935ead97" + }, + { + "dataPath": "params_shard_28.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.4.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "ed40e1ea7eacd283ed70d4d6de617b73" + }, + { + "dataPath": "params_shard_29.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.3.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.3.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.3.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.3.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.4.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.4.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.4.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.4.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.4.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": "b45110b32c65db614727af109244f94a" + }, + { + "dataPath": "params_shard_30.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.5.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.5.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.5.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.5.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": "8264ba26b37d4c17adead643786724ca" + }, + { + "dataPath": "params_shard_31.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.6.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "ffd3198c1693eade7c08f07e416d0fd7" + }, + { + "dataPath": "params_shard_32.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.5.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.5.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.5.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.5.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.6.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.6.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.6.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.6.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.6.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": "bb5290cf57ec21e1efaab5dbc38d580e" + }, + { + "dataPath": "params_shard_33.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.7.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.7.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.7.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.7.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": "02f2415a6ec6acac4cf72b0ab00d2920" + }, + { + "dataPath": "params_shard_34.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.8.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "2b806cc67dd65a746deb356d8978eca4" + }, + { + "dataPath": "params_shard_35.bin", + "format": "raw-shard", + "nbytes": 26233344, + "records": [ + { + "name": "model.layers.7.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.7.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.7.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.7.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.8.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.8.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 18929664 + }, + { + "name": "model.layers.8.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 23648256 + }, + { + "name": "model.layers.8.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 23795712 + }, + { + "name": "model.layers.8.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": "c39182529cc556908da35bb6d1ae877c" + }, + { + "dataPath": "params_shard_36.bin", + "format": "raw-shard", + "nbytes": 32859648, + "records": [ + { + "name": "model.layers.9.mlp.down_proj.q_weight", + "shape": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 0 + }, + { + "name": "model.layers.9.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 10616832 + }, + { + "name": "model.layers.9.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 10948608 + }, + { + "name": "model.layers.9.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": "83f57badb2564a2f15946076de1ab08d" + }, + { + "dataPath": "params_shard_37.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.24.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "671d691d49afb8e54bf52c2af3f427ec" + }, + { + "dataPath": "params_shard_38.bin", + "format": "raw-shard", + "nbytes": 21233664, + "records": [ + { + "name": "model.layers.25.mlp.gate_up_proj.q_weight", + "shape": [ + 18432, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 21233664, + "byteOffset": 0 + } + ], + "md5sum": "1ed390c829925345121914acd166792f" + }, + { + "dataPath": "params_shard_39.bin", + "format": "raw-shard", + "nbytes": 32993280, + "records": [ + { + "name": "model.layers.9.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.9.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.9.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.9.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 7303680 + }, + { + "name": "model.layers.24.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 17920512 + }, + { + "name": "model.layers.24.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 18929664 + }, + { + "name": "model.layers.24.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": [ + 2304, + 1152 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10616832, + "byteOffset": 21367296 + }, + { + "name": "model.layers.25.mlp.down_proj.q_scale", + "shape": [ + 2304, + 72 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 331776, + "byteOffset": 31984128 + }, + { + "name": "model.layers.25.mlp.gate_up_proj.q_scale", + "shape": [ + 18432, + 18 + ], + "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": "52c594148b6f44ea9de8b76757c9ee28" + }, + { + "dataPath": "params_shard_40.bin", + "format": "raw-shard", + "nbytes": 7303680, + "records": [ + { + "name": "model.layers.25.self_attn.qkv_proj.q_weight", + "shape": [ + 4096, + 288 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 4718592, + "byteOffset": 0 + }, + { + "name": "model.layers.25.self_attn.qkv_proj.q_scale", + "shape": [ + 4096, + 18 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 147456, + "byteOffset": 4718592 + }, + { + "name": "model.layers.25.self_attn.o_proj.q_weight", + "shape": [ + 2304, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2359296, + "byteOffset": 4866048 + }, + { + "name": "model.layers.25.self_attn.o_proj.q_scale", + "shape": [ + 2304, + 16 + ], + "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": "6c0ddfff4672377d1b98a5c528a711c8" + } + ] +} \ No newline at end of file