diff --git "a/ndarray-cache-b16.json" "b/ndarray-cache-b16.json" new file mode 100644--- /dev/null +++ "b/ndarray-cache-b16.json" @@ -0,0 +1,5809 @@ +{ + "metadata": { + "ParamSize": 518, + "ParamBytes": 1248667648.0, + "BitsPerParam": 3.6071579515642074 + }, + "records": [ + { + "dataPath": "params_shard_0.bin", + "format": "raw-shard", + "nbytes": 50331648, + "records": [ + { + "name": "gpt_neox.embed_in.q_weight", + "shape": [ + 49152, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 50331648, + "byteOffset": 0 + } + ], + "md5sum": "822f7593d2c576534d6e7ceadeff80e9" + }, + { + "dataPath": "params_shard_1.bin", + "format": "raw-shard", + "nbytes": 29945856, + "records": [ + { + "name": "gpt_neox.embed_in.q_scale", + "shape": [ + 49152, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.0.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 6291456 + }, + { + "name": "gpt_neox.layers.0.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 6296576 + }, + { + "name": "gpt_neox.layers.0.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 6301696 + }, + { + "name": "gpt_neox.layers.0.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 6306816 + }, + { + "name": "gpt_neox.layers.0.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 6311936 + }, + { + "name": "gpt_neox.layers.0.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 14176256 + }, + { + "name": "gpt_neox.layers.0.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 15159296 + }, + { + "name": "gpt_neox.layers.0.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 15174656 + }, + { + "name": "gpt_neox.layers.0.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 17796096 + }, + { + "name": "gpt_neox.layers.0.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 18123776 + }, + { + "name": "gpt_neox.layers.0.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 18128896 + }, + { + "name": "gpt_neox.layers.0.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 28614656 + }, + { + "name": "gpt_neox.layers.0.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 29925376 + } + ], + "md5sum": "9288b062709d19a17625040914506636" + }, + { + "dataPath": "params_shard_2.bin", + "format": "raw-shard", + "nbytes": 23639040, + "records": [ + { + "name": "gpt_neox.layers.0.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.0.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.0.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.1.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11801600 + }, + { + "name": "gpt_neox.layers.1.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11806720 + }, + { + "name": "gpt_neox.layers.1.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.1.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.1.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 11822080 + }, + { + "name": "gpt_neox.layers.1.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 19686400 + }, + { + "name": "gpt_neox.layers.1.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 20669440 + }, + { + "name": "gpt_neox.layers.1.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 20684800 + }, + { + "name": "gpt_neox.layers.1.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 23306240 + }, + { + "name": "gpt_neox.layers.1.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + } + ], + "md5sum": "2ea18d4fbaf4e92f89e111778cc300ba" + }, + { + "dataPath": "params_shard_3.bin", + "format": "raw-shard", + "nbytes": 32501760, + "records": [ + { + "name": "gpt_neox.layers.1.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.1.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.1.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.1.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.1.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.1.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 23613440 + }, + { + "name": "gpt_neox.layers.2.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23618560 + }, + { + "name": "gpt_neox.layers.2.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23623680 + }, + { + "name": "gpt_neox.layers.2.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23628800 + }, + { + "name": "gpt_neox.layers.2.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + }, + { + "name": "gpt_neox.layers.2.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 23639040 + }, + { + "name": "gpt_neox.layers.2.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 31503360 + }, + { + "name": "gpt_neox.layers.2.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 32486400 + } + ], + "md5sum": "71e1129512f6242f9960ac3aa81c6c26" + }, + { + "dataPath": "params_shard_4.bin", + "format": "raw-shard", + "nbytes": 26593280, + "records": [ + { + "name": "gpt_neox.layers.2.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.2.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 2621440 + }, + { + "name": "gpt_neox.layers.2.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 2949120 + }, + { + "name": "gpt_neox.layers.2.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 2954240 + }, + { + "name": "gpt_neox.layers.2.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 13440000 + }, + { + "name": "gpt_neox.layers.2.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.2.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.2.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 25256960 + }, + { + "name": "gpt_neox.layers.2.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 26567680 + }, + { + "name": "gpt_neox.layers.3.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26572800 + }, + { + "name": "gpt_neox.layers.3.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26577920 + }, + { + "name": "gpt_neox.layers.3.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26583040 + }, + { + "name": "gpt_neox.layers.3.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26588160 + } + ], + "md5sum": "1e2acfa39fafcf395b2d92acc4d301e8" + }, + { + "dataPath": "params_shard_5.bin", + "format": "raw-shard", + "nbytes": 23633920, + "records": [ + { + "name": "gpt_neox.layers.3.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.3.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 7864320 + }, + { + "name": "gpt_neox.layers.3.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 8847360 + }, + { + "name": "gpt_neox.layers.3.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 8862720 + }, + { + "name": "gpt_neox.layers.3.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 11484160 + }, + { + "name": "gpt_neox.layers.3.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.3.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.3.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.3.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 23613440 + } + ], + "md5sum": "875fc08daacaa4db8cb01ff341406b5c" + }, + { + "dataPath": "params_shard_6.bin", + "format": "raw-shard", + "nbytes": 23639040, + "records": [ + { + "name": "gpt_neox.layers.3.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.3.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.3.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.4.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11801600 + }, + { + "name": "gpt_neox.layers.4.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11806720 + }, + { + "name": "gpt_neox.layers.4.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.4.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.4.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 11822080 + }, + { + "name": "gpt_neox.layers.4.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 19686400 + }, + { + "name": "gpt_neox.layers.4.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 20669440 + }, + { + "name": "gpt_neox.layers.4.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 20684800 + }, + { + "name": "gpt_neox.layers.4.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 23306240 + }, + { + "name": "gpt_neox.layers.4.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + } + ], + "md5sum": "55a338b22f7869bbc2045915614ac642" + }, + { + "dataPath": "params_shard_7.bin", + "format": "raw-shard", + "nbytes": 32501760, + "records": [ + { + "name": "gpt_neox.layers.4.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.4.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.4.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.4.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.4.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.4.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 23613440 + }, + { + "name": "gpt_neox.layers.5.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23618560 + }, + { + "name": "gpt_neox.layers.5.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23623680 + }, + { + "name": "gpt_neox.layers.5.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23628800 + }, + { + "name": "gpt_neox.layers.5.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + }, + { + "name": "gpt_neox.layers.5.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 23639040 + }, + { + "name": "gpt_neox.layers.5.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 31503360 + }, + { + "name": "gpt_neox.layers.5.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 32486400 + } + ], + "md5sum": "f7e969fbb188fbbdd2eaf417826ee642" + }, + { + "dataPath": "params_shard_8.bin", + "format": "raw-shard", + "nbytes": 26593280, + "records": [ + { + "name": "gpt_neox.layers.5.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.5.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 2621440 + }, + { + "name": "gpt_neox.layers.5.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 2949120 + }, + { + "name": "gpt_neox.layers.5.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 2954240 + }, + { + "name": "gpt_neox.layers.5.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 13440000 + }, + { + "name": "gpt_neox.layers.5.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.5.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.5.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 25256960 + }, + { + "name": "gpt_neox.layers.5.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 26567680 + }, + { + "name": "gpt_neox.layers.6.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26572800 + }, + { + "name": "gpt_neox.layers.6.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26577920 + }, + { + "name": "gpt_neox.layers.6.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26583040 + }, + { + "name": "gpt_neox.layers.6.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26588160 + } + ], + "md5sum": "7b824c166abf378d9147a83e71365af6" + }, + { + "dataPath": "params_shard_9.bin", + "format": "raw-shard", + "nbytes": 23633920, + "records": [ + { + "name": "gpt_neox.layers.6.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.6.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 7864320 + }, + { + "name": "gpt_neox.layers.6.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 8847360 + }, + { + "name": "gpt_neox.layers.6.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 8862720 + }, + { + "name": "gpt_neox.layers.6.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 11484160 + }, + { + "name": "gpt_neox.layers.6.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.6.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.6.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.6.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 23613440 + } + ], + "md5sum": "6166d03cfc84aa17f32e12f473757a1d" + }, + { + "dataPath": "params_shard_10.bin", + "format": "raw-shard", + "nbytes": 23639040, + "records": [ + { + "name": "gpt_neox.layers.6.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.6.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.6.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.7.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11801600 + }, + { + "name": "gpt_neox.layers.7.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11806720 + }, + { + "name": "gpt_neox.layers.7.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.7.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.7.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 11822080 + }, + { + "name": "gpt_neox.layers.7.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 19686400 + }, + { + "name": "gpt_neox.layers.7.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 20669440 + }, + { + "name": "gpt_neox.layers.7.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 20684800 + }, + { + "name": "gpt_neox.layers.7.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 23306240 + }, + { + "name": "gpt_neox.layers.7.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + } + ], + "md5sum": "f5522e4cab061d9792d125bfbcf30639" + }, + { + "dataPath": "params_shard_11.bin", + "format": "raw-shard", + "nbytes": 32501760, + "records": [ + { + "name": "gpt_neox.layers.7.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.7.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.7.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.7.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.7.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.7.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 23613440 + }, + { + "name": "gpt_neox.layers.8.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23618560 + }, + { + "name": "gpt_neox.layers.8.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23623680 + }, + { + "name": "gpt_neox.layers.8.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23628800 + }, + { + "name": "gpt_neox.layers.8.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + }, + { + "name": "gpt_neox.layers.8.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 23639040 + }, + { + "name": "gpt_neox.layers.8.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 31503360 + }, + { + "name": "gpt_neox.layers.8.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 32486400 + } + ], + "md5sum": "e6c918edd5d038d767e929464899a86e" + }, + { + "dataPath": "params_shard_12.bin", + "format": "raw-shard", + "nbytes": 26593280, + "records": [ + { + "name": "gpt_neox.layers.8.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.8.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 2621440 + }, + { + "name": "gpt_neox.layers.8.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 2949120 + }, + { + "name": "gpt_neox.layers.8.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 2954240 + }, + { + "name": "gpt_neox.layers.8.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 13440000 + }, + { + "name": "gpt_neox.layers.8.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.8.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.8.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 25256960 + }, + { + "name": "gpt_neox.layers.8.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 26567680 + }, + { + "name": "gpt_neox.layers.9.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26572800 + }, + { + "name": "gpt_neox.layers.9.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26577920 + }, + { + "name": "gpt_neox.layers.9.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26583040 + }, + { + "name": "gpt_neox.layers.9.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26588160 + } + ], + "md5sum": "b86d6df4055de6a30d4b4686bd3f7307" + }, + { + "dataPath": "params_shard_13.bin", + "format": "raw-shard", + "nbytes": 23633920, + "records": [ + { + "name": "gpt_neox.layers.9.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.9.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 7864320 + }, + { + "name": "gpt_neox.layers.9.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 8847360 + }, + { + "name": "gpt_neox.layers.9.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 8862720 + }, + { + "name": "gpt_neox.layers.9.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 11484160 + }, + { + "name": "gpt_neox.layers.9.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.9.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.9.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.9.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 23613440 + } + ], + "md5sum": "f8c423fb7348b0ec08a7fddf0cc24959" + }, + { + "dataPath": "params_shard_14.bin", + "format": "raw-shard", + "nbytes": 23639040, + "records": [ + { + "name": "gpt_neox.layers.9.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.9.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.9.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.10.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11801600 + }, + { + "name": "gpt_neox.layers.10.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11806720 + }, + { + "name": "gpt_neox.layers.10.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.10.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.10.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 11822080 + }, + { + "name": "gpt_neox.layers.10.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 19686400 + }, + { + "name": "gpt_neox.layers.10.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 20669440 + }, + { + "name": "gpt_neox.layers.10.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 20684800 + }, + { + "name": "gpt_neox.layers.10.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 23306240 + }, + { + "name": "gpt_neox.layers.10.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + } + ], + "md5sum": "5e0e05fa2c98b4a4f9d81d4564a0fa36" + }, + { + "dataPath": "params_shard_15.bin", + "format": "raw-shard", + "nbytes": 32501760, + "records": [ + { + "name": "gpt_neox.layers.10.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.10.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.10.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.10.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.10.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.10.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 23613440 + }, + { + "name": "gpt_neox.layers.11.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23618560 + }, + { + "name": "gpt_neox.layers.11.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23623680 + }, + { + "name": "gpt_neox.layers.11.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23628800 + }, + { + "name": "gpt_neox.layers.11.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + }, + { + "name": "gpt_neox.layers.11.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 23639040 + }, + { + "name": "gpt_neox.layers.11.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 31503360 + }, + { + "name": "gpt_neox.layers.11.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 32486400 + } + ], + "md5sum": "1f2b2b1b62d0d995a618e7f1592e972f" + }, + { + "dataPath": "params_shard_16.bin", + "format": "raw-shard", + "nbytes": 26593280, + "records": [ + { + "name": "gpt_neox.layers.11.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.11.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 2621440 + }, + { + "name": "gpt_neox.layers.11.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 2949120 + }, + { + "name": "gpt_neox.layers.11.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 2954240 + }, + { + "name": "gpt_neox.layers.11.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 13440000 + }, + { + "name": "gpt_neox.layers.11.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.11.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.11.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 25256960 + }, + { + "name": "gpt_neox.layers.11.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 26567680 + }, + { + "name": "gpt_neox.layers.12.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26572800 + }, + { + "name": "gpt_neox.layers.12.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26577920 + }, + { + "name": "gpt_neox.layers.12.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26583040 + }, + { + "name": "gpt_neox.layers.12.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26588160 + } + ], + "md5sum": "59bec034e2d0cb41c0176610358d99f8" + }, + { + "dataPath": "params_shard_17.bin", + "format": "raw-shard", + "nbytes": 23633920, + "records": [ + { + "name": "gpt_neox.layers.12.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.12.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 7864320 + }, + { + "name": "gpt_neox.layers.12.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 8847360 + }, + { + "name": "gpt_neox.layers.12.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 8862720 + }, + { + "name": "gpt_neox.layers.12.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 11484160 + }, + { + "name": "gpt_neox.layers.12.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.12.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.12.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.12.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 23613440 + } + ], + "md5sum": "1a08439ee41b054210554f08651b8220" + }, + { + "dataPath": "params_shard_18.bin", + "format": "raw-shard", + "nbytes": 23639040, + "records": [ + { + "name": "gpt_neox.layers.12.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.12.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.12.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.13.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11801600 + }, + { + "name": "gpt_neox.layers.13.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11806720 + }, + { + "name": "gpt_neox.layers.13.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.13.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.13.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 11822080 + }, + { + "name": "gpt_neox.layers.13.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 19686400 + }, + { + "name": "gpt_neox.layers.13.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 20669440 + }, + { + "name": "gpt_neox.layers.13.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 20684800 + }, + { + "name": "gpt_neox.layers.13.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 23306240 + }, + { + "name": "gpt_neox.layers.13.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + } + ], + "md5sum": "fe297a1c495b5b184cdadec1464037b9" + }, + { + "dataPath": "params_shard_19.bin", + "format": "raw-shard", + "nbytes": 32501760, + "records": [ + { + "name": "gpt_neox.layers.13.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.13.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.13.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.13.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.13.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.13.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 23613440 + }, + { + "name": "gpt_neox.layers.14.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23618560 + }, + { + "name": "gpt_neox.layers.14.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23623680 + }, + { + "name": "gpt_neox.layers.14.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23628800 + }, + { + "name": "gpt_neox.layers.14.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + }, + { + "name": "gpt_neox.layers.14.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 23639040 + }, + { + "name": "gpt_neox.layers.14.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 31503360 + }, + { + "name": "gpt_neox.layers.14.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 32486400 + } + ], + "md5sum": "c876235e99f47995b7fbdb78010e02c2" + }, + { + "dataPath": "params_shard_20.bin", + "format": "raw-shard", + "nbytes": 26593280, + "records": [ + { + "name": "gpt_neox.layers.14.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.14.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 2621440 + }, + { + "name": "gpt_neox.layers.14.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 2949120 + }, + { + "name": "gpt_neox.layers.14.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 2954240 + }, + { + "name": "gpt_neox.layers.14.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 13440000 + }, + { + "name": "gpt_neox.layers.14.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.14.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.14.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 25256960 + }, + { + "name": "gpt_neox.layers.14.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 26567680 + }, + { + "name": "gpt_neox.layers.15.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26572800 + }, + { + "name": "gpt_neox.layers.15.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26577920 + }, + { + "name": "gpt_neox.layers.15.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26583040 + }, + { + "name": "gpt_neox.layers.15.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26588160 + } + ], + "md5sum": "1d02fd6181cfec36e44972daa9933ddb" + }, + { + "dataPath": "params_shard_21.bin", + "format": "raw-shard", + "nbytes": 23633920, + "records": [ + { + "name": "gpt_neox.layers.15.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.15.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 7864320 + }, + { + "name": "gpt_neox.layers.15.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 8847360 + }, + { + "name": "gpt_neox.layers.15.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 8862720 + }, + { + "name": "gpt_neox.layers.15.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 11484160 + }, + { + "name": "gpt_neox.layers.15.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.15.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.15.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.15.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 23613440 + } + ], + "md5sum": "7491044c17ff6189819088333e698c76" + }, + { + "dataPath": "params_shard_22.bin", + "format": "raw-shard", + "nbytes": 23639040, + "records": [ + { + "name": "gpt_neox.layers.15.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.15.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.15.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.16.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11801600 + }, + { + "name": "gpt_neox.layers.16.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11806720 + }, + { + "name": "gpt_neox.layers.16.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.16.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.16.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 11822080 + }, + { + "name": "gpt_neox.layers.16.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 19686400 + }, + { + "name": "gpt_neox.layers.16.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 20669440 + }, + { + "name": "gpt_neox.layers.16.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 20684800 + }, + { + "name": "gpt_neox.layers.16.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 23306240 + }, + { + "name": "gpt_neox.layers.16.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + } + ], + "md5sum": "8bbf15d3ff6a31448b6b6534b6fd8373" + }, + { + "dataPath": "params_shard_23.bin", + "format": "raw-shard", + "nbytes": 32501760, + "records": [ + { + "name": "gpt_neox.layers.16.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.16.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.16.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.16.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.16.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.16.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 23613440 + }, + { + "name": "gpt_neox.layers.17.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23618560 + }, + { + "name": "gpt_neox.layers.17.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23623680 + }, + { + "name": "gpt_neox.layers.17.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23628800 + }, + { + "name": "gpt_neox.layers.17.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + }, + { + "name": "gpt_neox.layers.17.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 23639040 + }, + { + "name": "gpt_neox.layers.17.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 31503360 + }, + { + "name": "gpt_neox.layers.17.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 32486400 + } + ], + "md5sum": "8ec19bba3750983c0b367a075f8ff500" + }, + { + "dataPath": "params_shard_24.bin", + "format": "raw-shard", + "nbytes": 26593280, + "records": [ + { + "name": "gpt_neox.layers.17.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.17.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 2621440 + }, + { + "name": "gpt_neox.layers.17.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 2949120 + }, + { + "name": "gpt_neox.layers.17.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 2954240 + }, + { + "name": "gpt_neox.layers.17.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 13440000 + }, + { + "name": "gpt_neox.layers.17.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.17.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.17.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 25256960 + }, + { + "name": "gpt_neox.layers.17.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 26567680 + }, + { + "name": "gpt_neox.layers.18.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26572800 + }, + { + "name": "gpt_neox.layers.18.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26577920 + }, + { + "name": "gpt_neox.layers.18.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26583040 + }, + { + "name": "gpt_neox.layers.18.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26588160 + } + ], + "md5sum": "20bb926de9fca83d2702a1f1a68d8296" + }, + { + "dataPath": "params_shard_25.bin", + "format": "raw-shard", + "nbytes": 23633920, + "records": [ + { + "name": "gpt_neox.layers.18.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.18.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 7864320 + }, + { + "name": "gpt_neox.layers.18.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 8847360 + }, + { + "name": "gpt_neox.layers.18.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 8862720 + }, + { + "name": "gpt_neox.layers.18.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 11484160 + }, + { + "name": "gpt_neox.layers.18.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.18.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.18.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.18.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 23613440 + } + ], + "md5sum": "97d1f6f7142c8fb540f33179c70abae0" + }, + { + "dataPath": "params_shard_26.bin", + "format": "raw-shard", + "nbytes": 23639040, + "records": [ + { + "name": "gpt_neox.layers.18.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.18.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.18.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.19.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11801600 + }, + { + "name": "gpt_neox.layers.19.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11806720 + }, + { + "name": "gpt_neox.layers.19.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.19.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.19.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 11822080 + }, + { + "name": "gpt_neox.layers.19.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 19686400 + }, + { + "name": "gpt_neox.layers.19.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 20669440 + }, + { + "name": "gpt_neox.layers.19.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 20684800 + }, + { + "name": "gpt_neox.layers.19.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 23306240 + }, + { + "name": "gpt_neox.layers.19.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + } + ], + "md5sum": "a015efdc822cd39ede7d74ab95c35ee0" + }, + { + "dataPath": "params_shard_27.bin", + "format": "raw-shard", + "nbytes": 32501760, + "records": [ + { + "name": "gpt_neox.layers.19.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.19.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.19.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.19.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.19.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.19.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 23613440 + }, + { + "name": "gpt_neox.layers.20.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23618560 + }, + { + "name": "gpt_neox.layers.20.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23623680 + }, + { + "name": "gpt_neox.layers.20.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23628800 + }, + { + "name": "gpt_neox.layers.20.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + }, + { + "name": "gpt_neox.layers.20.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 23639040 + }, + { + "name": "gpt_neox.layers.20.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 31503360 + }, + { + "name": "gpt_neox.layers.20.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 32486400 + } + ], + "md5sum": "5dc20b9cedc6c942e9dc841afbc41513" + }, + { + "dataPath": "params_shard_28.bin", + "format": "raw-shard", + "nbytes": 26593280, + "records": [ + { + "name": "gpt_neox.layers.20.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.20.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 2621440 + }, + { + "name": "gpt_neox.layers.20.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 2949120 + }, + { + "name": "gpt_neox.layers.20.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 2954240 + }, + { + "name": "gpt_neox.layers.20.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 13440000 + }, + { + "name": "gpt_neox.layers.20.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.20.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.20.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 25256960 + }, + { + "name": "gpt_neox.layers.20.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 26567680 + }, + { + "name": "gpt_neox.layers.21.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26572800 + }, + { + "name": "gpt_neox.layers.21.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26577920 + }, + { + "name": "gpt_neox.layers.21.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26583040 + }, + { + "name": "gpt_neox.layers.21.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26588160 + } + ], + "md5sum": "11c42681c2ac09ff3fd228ffb9ca1052" + }, + { + "dataPath": "params_shard_29.bin", + "format": "raw-shard", + "nbytes": 23633920, + "records": [ + { + "name": "gpt_neox.layers.21.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.21.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 7864320 + }, + { + "name": "gpt_neox.layers.21.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 8847360 + }, + { + "name": "gpt_neox.layers.21.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 8862720 + }, + { + "name": "gpt_neox.layers.21.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 11484160 + }, + { + "name": "gpt_neox.layers.21.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.21.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.21.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.21.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 23613440 + } + ], + "md5sum": "e6cf11c9971c5f3cff69c49a1fc52432" + }, + { + "dataPath": "params_shard_30.bin", + "format": "raw-shard", + "nbytes": 23639040, + "records": [ + { + "name": "gpt_neox.layers.21.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.21.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.21.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.22.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11801600 + }, + { + "name": "gpt_neox.layers.22.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11806720 + }, + { + "name": "gpt_neox.layers.22.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.22.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.22.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 11822080 + }, + { + "name": "gpt_neox.layers.22.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 19686400 + }, + { + "name": "gpt_neox.layers.22.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 20669440 + }, + { + "name": "gpt_neox.layers.22.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 20684800 + }, + { + "name": "gpt_neox.layers.22.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 23306240 + }, + { + "name": "gpt_neox.layers.22.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + } + ], + "md5sum": "ebe8336b9896ac2125a1345e119f0217" + }, + { + "dataPath": "params_shard_31.bin", + "format": "raw-shard", + "nbytes": 32501760, + "records": [ + { + "name": "gpt_neox.layers.22.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.22.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.22.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.22.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.22.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.22.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 23613440 + }, + { + "name": "gpt_neox.layers.23.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23618560 + }, + { + "name": "gpt_neox.layers.23.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23623680 + }, + { + "name": "gpt_neox.layers.23.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23628800 + }, + { + "name": "gpt_neox.layers.23.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + }, + { + "name": "gpt_neox.layers.23.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 23639040 + }, + { + "name": "gpt_neox.layers.23.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 31503360 + }, + { + "name": "gpt_neox.layers.23.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 32486400 + } + ], + "md5sum": "5f82aeaa9e4a777f234e3e39ce837046" + }, + { + "dataPath": "params_shard_32.bin", + "format": "raw-shard", + "nbytes": 26593280, + "records": [ + { + "name": "gpt_neox.layers.23.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.23.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 2621440 + }, + { + "name": "gpt_neox.layers.23.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 2949120 + }, + { + "name": "gpt_neox.layers.23.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 2954240 + }, + { + "name": "gpt_neox.layers.23.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 13440000 + }, + { + "name": "gpt_neox.layers.23.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.23.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.23.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 25256960 + }, + { + "name": "gpt_neox.layers.23.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 26567680 + }, + { + "name": "gpt_neox.layers.24.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26572800 + }, + { + "name": "gpt_neox.layers.24.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26577920 + }, + { + "name": "gpt_neox.layers.24.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26583040 + }, + { + "name": "gpt_neox.layers.24.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26588160 + } + ], + "md5sum": "6e8ed3b79bf665729eef4bad7d759c1f" + }, + { + "dataPath": "params_shard_33.bin", + "format": "raw-shard", + "nbytes": 23633920, + "records": [ + { + "name": "gpt_neox.layers.24.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.24.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 7864320 + }, + { + "name": "gpt_neox.layers.24.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 8847360 + }, + { + "name": "gpt_neox.layers.24.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 8862720 + }, + { + "name": "gpt_neox.layers.24.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 11484160 + }, + { + "name": "gpt_neox.layers.24.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.24.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.24.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.24.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 23613440 + } + ], + "md5sum": "0b931da1f3ec481aecffd7c3053953d3" + }, + { + "dataPath": "params_shard_34.bin", + "format": "raw-shard", + "nbytes": 23639040, + "records": [ + { + "name": "gpt_neox.layers.24.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.24.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.24.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.25.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11801600 + }, + { + "name": "gpt_neox.layers.25.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11806720 + }, + { + "name": "gpt_neox.layers.25.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.25.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.25.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 11822080 + }, + { + "name": "gpt_neox.layers.25.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 19686400 + }, + { + "name": "gpt_neox.layers.25.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 20669440 + }, + { + "name": "gpt_neox.layers.25.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 20684800 + }, + { + "name": "gpt_neox.layers.25.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 23306240 + }, + { + "name": "gpt_neox.layers.25.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + } + ], + "md5sum": "5cc788521f027d4109678976dc8b32ef" + }, + { + "dataPath": "params_shard_35.bin", + "format": "raw-shard", + "nbytes": 32501760, + "records": [ + { + "name": "gpt_neox.layers.25.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.25.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.25.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.25.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.25.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.25.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 23613440 + }, + { + "name": "gpt_neox.layers.26.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23618560 + }, + { + "name": "gpt_neox.layers.26.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23623680 + }, + { + "name": "gpt_neox.layers.26.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23628800 + }, + { + "name": "gpt_neox.layers.26.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + }, + { + "name": "gpt_neox.layers.26.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 23639040 + }, + { + "name": "gpt_neox.layers.26.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 31503360 + }, + { + "name": "gpt_neox.layers.26.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 32486400 + } + ], + "md5sum": "a81e5b77ef7b9c2c4e0eeaec6aaf896e" + }, + { + "dataPath": "params_shard_36.bin", + "format": "raw-shard", + "nbytes": 26593280, + "records": [ + { + "name": "gpt_neox.layers.26.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.26.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 2621440 + }, + { + "name": "gpt_neox.layers.26.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 2949120 + }, + { + "name": "gpt_neox.layers.26.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 2954240 + }, + { + "name": "gpt_neox.layers.26.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 13440000 + }, + { + "name": "gpt_neox.layers.26.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.26.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.26.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 25256960 + }, + { + "name": "gpt_neox.layers.26.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 26567680 + }, + { + "name": "gpt_neox.layers.27.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26572800 + }, + { + "name": "gpt_neox.layers.27.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26577920 + }, + { + "name": "gpt_neox.layers.27.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26583040 + }, + { + "name": "gpt_neox.layers.27.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26588160 + } + ], + "md5sum": "507cdddc701a2b55b04b36f1b86d56df" + }, + { + "dataPath": "params_shard_37.bin", + "format": "raw-shard", + "nbytes": 23633920, + "records": [ + { + "name": "gpt_neox.layers.27.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.27.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 7864320 + }, + { + "name": "gpt_neox.layers.27.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 8847360 + }, + { + "name": "gpt_neox.layers.27.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 8862720 + }, + { + "name": "gpt_neox.layers.27.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 11484160 + }, + { + "name": "gpt_neox.layers.27.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.27.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.27.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.27.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 23613440 + } + ], + "md5sum": "2d73849944415a91209fa3a7add97228" + }, + { + "dataPath": "params_shard_38.bin", + "format": "raw-shard", + "nbytes": 23639040, + "records": [ + { + "name": "gpt_neox.layers.27.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.27.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.27.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.28.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11801600 + }, + { + "name": "gpt_neox.layers.28.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11806720 + }, + { + "name": "gpt_neox.layers.28.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.28.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.28.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 11822080 + }, + { + "name": "gpt_neox.layers.28.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 19686400 + }, + { + "name": "gpt_neox.layers.28.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 20669440 + }, + { + "name": "gpt_neox.layers.28.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 20684800 + }, + { + "name": "gpt_neox.layers.28.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 23306240 + }, + { + "name": "gpt_neox.layers.28.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + } + ], + "md5sum": "22b84627863638058913843a07059c9b" + }, + { + "dataPath": "params_shard_39.bin", + "format": "raw-shard", + "nbytes": 32501760, + "records": [ + { + "name": "gpt_neox.layers.28.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.28.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.28.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.28.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.28.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.28.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 23613440 + }, + { + "name": "gpt_neox.layers.29.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23618560 + }, + { + "name": "gpt_neox.layers.29.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23623680 + }, + { + "name": "gpt_neox.layers.29.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23628800 + }, + { + "name": "gpt_neox.layers.29.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + }, + { + "name": "gpt_neox.layers.29.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 23639040 + }, + { + "name": "gpt_neox.layers.29.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 31503360 + }, + { + "name": "gpt_neox.layers.29.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 32486400 + } + ], + "md5sum": "8f41a5547067d9f3cc4d20c4a5f9ea29" + }, + { + "dataPath": "params_shard_40.bin", + "format": "raw-shard", + "nbytes": 26593280, + "records": [ + { + "name": "gpt_neox.layers.29.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.29.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 2621440 + }, + { + "name": "gpt_neox.layers.29.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 2949120 + }, + { + "name": "gpt_neox.layers.29.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 2954240 + }, + { + "name": "gpt_neox.layers.29.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 13440000 + }, + { + "name": "gpt_neox.layers.29.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.29.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.29.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 25256960 + }, + { + "name": "gpt_neox.layers.29.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 26567680 + }, + { + "name": "gpt_neox.layers.30.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26572800 + }, + { + "name": "gpt_neox.layers.30.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26577920 + }, + { + "name": "gpt_neox.layers.30.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26583040 + }, + { + "name": "gpt_neox.layers.30.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 26588160 + } + ], + "md5sum": "49610c4bc96a7eb9dcd4b5b427cc4b45" + }, + { + "dataPath": "params_shard_41.bin", + "format": "raw-shard", + "nbytes": 23633920, + "records": [ + { + "name": "gpt_neox.layers.30.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.30.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 7864320 + }, + { + "name": "gpt_neox.layers.30.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 8847360 + }, + { + "name": "gpt_neox.layers.30.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 8862720 + }, + { + "name": "gpt_neox.layers.30.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 11484160 + }, + { + "name": "gpt_neox.layers.30.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.30.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.30.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.30.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 23613440 + } + ], + "md5sum": "f7f83e15035c52b5c0a57b86ba981dfd" + }, + { + "dataPath": "params_shard_42.bin", + "format": "raw-shard", + "nbytes": 23639040, + "records": [ + { + "name": "gpt_neox.layers.30.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.30.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.30.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.31.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11801600 + }, + { + "name": "gpt_neox.layers.31.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11806720 + }, + { + "name": "gpt_neox.layers.31.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11811840 + }, + { + "name": "gpt_neox.layers.31.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.31.attention.query_key_value.q_weight", + "shape": [ + 7680, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 11822080 + }, + { + "name": "gpt_neox.layers.31.attention.query_key_value.q_scale", + "shape": [ + 7680, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 983040, + "byteOffset": 19686400 + }, + { + "name": "gpt_neox.layers.31.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 20669440 + }, + { + "name": "gpt_neox.layers.31.attention.dense.q_weight", + "shape": [ + 2560, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 2621440, + "byteOffset": 20684800 + }, + { + "name": "gpt_neox.layers.31.attention.dense.q_scale", + "shape": [ + 2560, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 327680, + "byteOffset": 23306240 + }, + { + "name": "gpt_neox.layers.31.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23633920 + } + ], + "md5sum": "3da31b16c599d71e8b0f02c1206d4fc2" + }, + { + "dataPath": "params_shard_43.bin", + "format": "raw-shard", + "nbytes": 50331648, + "records": [ + { + "name": "embed_out.q_weight", + "shape": [ + 49152, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 50331648, + "byteOffset": 0 + } + ], + "md5sum": "6b3517f783e1489814775d738e3a9975" + }, + { + "dataPath": "params_shard_44.bin", + "format": "raw-shard", + "nbytes": 29920256, + "records": [ + { + "name": "gpt_neox.layers.31.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 256 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.31.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 10485760 + }, + { + "name": "gpt_neox.layers.31.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 11796480 + }, + { + "name": "gpt_neox.layers.31.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1024 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 10485760, + "byteOffset": 11816960 + }, + { + "name": "gpt_neox.layers.31.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 256 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1310720, + "byteOffset": 22302720 + }, + { + "name": "gpt_neox.layers.31.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 23613440 + }, + { + "name": "gpt_neox.final_layer_norm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23618560 + }, + { + "name": "gpt_neox.final_layer_norm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23623680 + }, + { + "name": "embed_out.q_scale", + "shape": [ + 49152, + 64 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 6291456, + "byteOffset": 23628800 + } + ], + "md5sum": "20ae36bfc162a22884bdf8b22ec13cce" + } + ] +} \ No newline at end of file