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 +1,5857 @@ +{ + "metadata": { + "ParamSize": 518, + "ParamBytes": 1560094720.0, + "BitsPerParam": 4.506810185604597 + }, + "records": [ + { + "dataPath": "params_shard_0.bin", + "format": "raw-shard", + "nbytes": 62914560, + "records": [ + { + "name": "gpt_neox.embed_in.q_weight", + "shape": [ + 49152, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 62914560, + "byteOffset": 0 + } + ], + "md5sum": "c021cc289146c0a59e5b5b5f7ee23e0a" + }, + { + "dataPath": "params_shard_1.bin", + "format": "raw-shard", + "nbytes": 22650880, + "records": [ + { + "name": "gpt_neox.embed_in.q_scale", + "shape": [ + 49152, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.0.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 7864320 + }, + { + "name": "gpt_neox.layers.0.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 7869440 + }, + { + "name": "gpt_neox.layers.0.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 7874560 + }, + { + "name": "gpt_neox.layers.0.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 7879680 + }, + { + "name": "gpt_neox.layers.0.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 7884800 + }, + { + "name": "gpt_neox.layers.0.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 17715200 + }, + { + "name": "gpt_neox.layers.0.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 18944000 + }, + { + "name": "gpt_neox.layers.0.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 18959360 + }, + { + "name": "gpt_neox.layers.0.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 22236160 + }, + { + "name": "gpt_neox.layers.0.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 22645760 + } + ], + "md5sum": "af21d15168499845c2e785eb4f85076e" + }, + { + "dataPath": "params_shard_2.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.0.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.0.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.0.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.0.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.0.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.0.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "gpt_neox.layers.1.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "gpt_neox.layers.1.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29521920 + }, + { + "name": "gpt_neox.layers.1.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29527040 + }, + { + "name": "gpt_neox.layers.1.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "0f829710ac9c8f2466761bf29afc67c8" + }, + { + "dataPath": "params_shard_3.bin", + "format": "raw-shard", + "nbytes": 29532160, + "records": [ + { + "name": "gpt_neox.layers.1.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.1.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "gpt_neox.layers.1.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 11059200 + }, + { + "name": "gpt_neox.layers.1.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11074560 + }, + { + "name": "gpt_neox.layers.1.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14351360 + }, + { + "name": "gpt_neox.layers.1.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.1.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.1.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.1.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 29511680 + } + ], + "md5sum": "a55ca060ffe65c8b84b6670b5fc003a5" + }, + { + "dataPath": "params_shard_4.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.1.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.1.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.1.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.2.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.2.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "gpt_neox.layers.2.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.2.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.2.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.2.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "gpt_neox.layers.2.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 25830400 + }, + { + "name": "gpt_neox.layers.2.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25845760 + }, + { + "name": "gpt_neox.layers.2.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29122560 + }, + { + "name": "gpt_neox.layers.2.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "2fe734cfa91236a0e6d0c3c5b90b5528" + }, + { + "dataPath": "params_shard_5.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.2.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.2.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.2.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.2.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.2.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.2.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "gpt_neox.layers.3.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "gpt_neox.layers.3.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29521920 + }, + { + "name": "gpt_neox.layers.3.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29527040 + }, + { + "name": "gpt_neox.layers.3.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "0c667de01ca2caabd3091631d0da2e45" + }, + { + "dataPath": "params_shard_6.bin", + "format": "raw-shard", + "nbytes": 29532160, + "records": [ + { + "name": "gpt_neox.layers.3.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.3.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "gpt_neox.layers.3.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 11059200 + }, + { + "name": "gpt_neox.layers.3.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11074560 + }, + { + "name": "gpt_neox.layers.3.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14351360 + }, + { + "name": "gpt_neox.layers.3.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.3.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.3.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.3.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 29511680 + } + ], + "md5sum": "1a801e5c816b294eb7d109d9361927b8" + }, + { + "dataPath": "params_shard_7.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.3.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.3.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.3.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.4.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.4.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "gpt_neox.layers.4.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.4.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.4.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.4.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "gpt_neox.layers.4.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 25830400 + }, + { + "name": "gpt_neox.layers.4.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25845760 + }, + { + "name": "gpt_neox.layers.4.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29122560 + }, + { + "name": "gpt_neox.layers.4.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "bd918970ae42484f06eaac6d603c90b3" + }, + { + "dataPath": "params_shard_8.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.4.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.4.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.4.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.4.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.4.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.4.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "gpt_neox.layers.5.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "gpt_neox.layers.5.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29521920 + }, + { + "name": "gpt_neox.layers.5.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29527040 + }, + { + "name": "gpt_neox.layers.5.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "4ab33f4a3866c2a14d128e23cea84dae" + }, + { + "dataPath": "params_shard_9.bin", + "format": "raw-shard", + "nbytes": 29532160, + "records": [ + { + "name": "gpt_neox.layers.5.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.5.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "gpt_neox.layers.5.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 11059200 + }, + { + "name": "gpt_neox.layers.5.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11074560 + }, + { + "name": "gpt_neox.layers.5.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14351360 + }, + { + "name": "gpt_neox.layers.5.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.5.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.5.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.5.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 29511680 + } + ], + "md5sum": "d1a59506f378cef31c90afb49b48f969" + }, + { + "dataPath": "params_shard_10.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.5.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.5.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.5.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.6.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.6.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "gpt_neox.layers.6.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.6.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.6.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.6.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "gpt_neox.layers.6.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 25830400 + }, + { + "name": "gpt_neox.layers.6.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25845760 + }, + { + "name": "gpt_neox.layers.6.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29122560 + }, + { + "name": "gpt_neox.layers.6.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "b4150e00f2b4973204faa71a96417544" + }, + { + "dataPath": "params_shard_11.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.6.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.6.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.6.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.6.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.6.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.6.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "gpt_neox.layers.7.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "gpt_neox.layers.7.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29521920 + }, + { + "name": "gpt_neox.layers.7.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29527040 + }, + { + "name": "gpt_neox.layers.7.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "985657dccfdf8dea0adf2bcb6eb65678" + }, + { + "dataPath": "params_shard_12.bin", + "format": "raw-shard", + "nbytes": 29532160, + "records": [ + { + "name": "gpt_neox.layers.7.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.7.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "gpt_neox.layers.7.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 11059200 + }, + { + "name": "gpt_neox.layers.7.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11074560 + }, + { + "name": "gpt_neox.layers.7.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14351360 + }, + { + "name": "gpt_neox.layers.7.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.7.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.7.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.7.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 29511680 + } + ], + "md5sum": "033481d2f2f563832c65c84f5dd882cd" + }, + { + "dataPath": "params_shard_13.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.7.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.7.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.7.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.8.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.8.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "gpt_neox.layers.8.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.8.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.8.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.8.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "gpt_neox.layers.8.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 25830400 + }, + { + "name": "gpt_neox.layers.8.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25845760 + }, + { + "name": "gpt_neox.layers.8.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29122560 + }, + { + "name": "gpt_neox.layers.8.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "9cee9a545f63071c1f830b7945838151" + }, + { + "dataPath": "params_shard_14.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.8.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.8.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.8.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.8.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.8.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.8.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "gpt_neox.layers.9.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "gpt_neox.layers.9.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29521920 + }, + { + "name": "gpt_neox.layers.9.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29527040 + }, + { + "name": "gpt_neox.layers.9.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "b35958c3aad1f8163c039b93dc3c74f8" + }, + { + "dataPath": "params_shard_15.bin", + "format": "raw-shard", + "nbytes": 29532160, + "records": [ + { + "name": "gpt_neox.layers.9.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.9.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "gpt_neox.layers.9.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 11059200 + }, + { + "name": "gpt_neox.layers.9.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11074560 + }, + { + "name": "gpt_neox.layers.9.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14351360 + }, + { + "name": "gpt_neox.layers.9.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.9.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.9.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.9.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 29511680 + } + ], + "md5sum": "69d5aaf05cd920939ff714cf730009ef" + }, + { + "dataPath": "params_shard_16.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.9.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.9.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.9.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.10.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.10.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "gpt_neox.layers.10.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.10.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.10.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.10.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "gpt_neox.layers.10.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 25830400 + }, + { + "name": "gpt_neox.layers.10.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25845760 + }, + { + "name": "gpt_neox.layers.10.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29122560 + }, + { + "name": "gpt_neox.layers.10.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "10e72dd7bf1bd57876e4ddce52015c53" + }, + { + "dataPath": "params_shard_17.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.10.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.10.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.10.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.10.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.10.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.10.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "gpt_neox.layers.11.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "gpt_neox.layers.11.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29521920 + }, + { + "name": "gpt_neox.layers.11.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29527040 + }, + { + "name": "gpt_neox.layers.11.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "c363b6cd7e4765aa8648f5ae6287aa1f" + }, + { + "dataPath": "params_shard_18.bin", + "format": "raw-shard", + "nbytes": 29532160, + "records": [ + { + "name": "gpt_neox.layers.11.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.11.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "gpt_neox.layers.11.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 11059200 + }, + { + "name": "gpt_neox.layers.11.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11074560 + }, + { + "name": "gpt_neox.layers.11.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14351360 + }, + { + "name": "gpt_neox.layers.11.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.11.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.11.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.11.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 29511680 + } + ], + "md5sum": "e1fab8a0a561bdff6a84bbe414a1b59f" + }, + { + "dataPath": "params_shard_19.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.11.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.11.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.11.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.12.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.12.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "gpt_neox.layers.12.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.12.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.12.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.12.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "gpt_neox.layers.12.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 25830400 + }, + { + "name": "gpt_neox.layers.12.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25845760 + }, + { + "name": "gpt_neox.layers.12.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29122560 + }, + { + "name": "gpt_neox.layers.12.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "a6ca638a935a2f759150acb294402bbf" + }, + { + "dataPath": "params_shard_20.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.12.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.12.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.12.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.12.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.12.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.12.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "gpt_neox.layers.13.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "gpt_neox.layers.13.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29521920 + }, + { + "name": "gpt_neox.layers.13.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29527040 + }, + { + "name": "gpt_neox.layers.13.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "b447131d6537c0d1658e6f3f75afba06" + }, + { + "dataPath": "params_shard_21.bin", + "format": "raw-shard", + "nbytes": 29532160, + "records": [ + { + "name": "gpt_neox.layers.13.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.13.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "gpt_neox.layers.13.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 11059200 + }, + { + "name": "gpt_neox.layers.13.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11074560 + }, + { + "name": "gpt_neox.layers.13.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14351360 + }, + { + "name": "gpt_neox.layers.13.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.13.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.13.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.13.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 29511680 + } + ], + "md5sum": "0a27856efce38f4c5a42f400d0077f60" + }, + { + "dataPath": "params_shard_22.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.13.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.13.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.13.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.14.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.14.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "gpt_neox.layers.14.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.14.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.14.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.14.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "gpt_neox.layers.14.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 25830400 + }, + { + "name": "gpt_neox.layers.14.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25845760 + }, + { + "name": "gpt_neox.layers.14.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29122560 + }, + { + "name": "gpt_neox.layers.14.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "f7a09660c99cb414a026b28b10de4576" + }, + { + "dataPath": "params_shard_23.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.14.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.14.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.14.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.14.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.14.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.14.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "gpt_neox.layers.15.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "gpt_neox.layers.15.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29521920 + }, + { + "name": "gpt_neox.layers.15.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29527040 + }, + { + "name": "gpt_neox.layers.15.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "87ca8d002596dc696a4896876f7857cf" + }, + { + "dataPath": "params_shard_24.bin", + "format": "raw-shard", + "nbytes": 29532160, + "records": [ + { + "name": "gpt_neox.layers.15.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.15.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "gpt_neox.layers.15.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 11059200 + }, + { + "name": "gpt_neox.layers.15.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11074560 + }, + { + "name": "gpt_neox.layers.15.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14351360 + }, + { + "name": "gpt_neox.layers.15.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.15.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.15.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.15.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 29511680 + } + ], + "md5sum": "b5d77b51eb35c5020d54515dbf96b5ba" + }, + { + "dataPath": "params_shard_25.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.15.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.15.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.15.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.16.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.16.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "gpt_neox.layers.16.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.16.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.16.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.16.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "gpt_neox.layers.16.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 25830400 + }, + { + "name": "gpt_neox.layers.16.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25845760 + }, + { + "name": "gpt_neox.layers.16.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29122560 + }, + { + "name": "gpt_neox.layers.16.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "23e48de12f99e8e1c625185b69d1f211" + }, + { + "dataPath": "params_shard_26.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.16.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.16.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.16.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.16.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.16.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.16.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "gpt_neox.layers.17.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "gpt_neox.layers.17.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29521920 + }, + { + "name": "gpt_neox.layers.17.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29527040 + }, + { + "name": "gpt_neox.layers.17.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "c172559a127295149b7368a2a964462e" + }, + { + "dataPath": "params_shard_27.bin", + "format": "raw-shard", + "nbytes": 29532160, + "records": [ + { + "name": "gpt_neox.layers.17.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.17.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "gpt_neox.layers.17.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 11059200 + }, + { + "name": "gpt_neox.layers.17.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11074560 + }, + { + "name": "gpt_neox.layers.17.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14351360 + }, + { + "name": "gpt_neox.layers.17.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.17.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.17.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.17.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 29511680 + } + ], + "md5sum": "d24872ff2f88c6fe6e7c9b3c20434288" + }, + { + "dataPath": "params_shard_28.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.17.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.17.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.17.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.18.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.18.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "gpt_neox.layers.18.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.18.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.18.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.18.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "gpt_neox.layers.18.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 25830400 + }, + { + "name": "gpt_neox.layers.18.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25845760 + }, + { + "name": "gpt_neox.layers.18.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29122560 + }, + { + "name": "gpt_neox.layers.18.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "fdb99f7ededcd486c77ebe05f86e012a" + }, + { + "dataPath": "params_shard_29.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.18.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.18.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.18.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.18.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.18.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.18.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "gpt_neox.layers.19.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "gpt_neox.layers.19.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29521920 + }, + { + "name": "gpt_neox.layers.19.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29527040 + }, + { + "name": "gpt_neox.layers.19.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "e84ce4842361a8c4cdf5aa2d80d09832" + }, + { + "dataPath": "params_shard_30.bin", + "format": "raw-shard", + "nbytes": 29532160, + "records": [ + { + "name": "gpt_neox.layers.19.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.19.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "gpt_neox.layers.19.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 11059200 + }, + { + "name": "gpt_neox.layers.19.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11074560 + }, + { + "name": "gpt_neox.layers.19.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14351360 + }, + { + "name": "gpt_neox.layers.19.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.19.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.19.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.19.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 29511680 + } + ], + "md5sum": "e532a9334f3c9a7b17112738545505e3" + }, + { + "dataPath": "params_shard_31.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.19.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.19.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.19.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.20.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.20.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "gpt_neox.layers.20.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.20.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.20.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.20.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "gpt_neox.layers.20.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 25830400 + }, + { + "name": "gpt_neox.layers.20.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25845760 + }, + { + "name": "gpt_neox.layers.20.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29122560 + }, + { + "name": "gpt_neox.layers.20.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "8b84c1b8861ab134756d37a5f21b16aa" + }, + { + "dataPath": "params_shard_32.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.20.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.20.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.20.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.20.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.20.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.20.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "gpt_neox.layers.21.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "gpt_neox.layers.21.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29521920 + }, + { + "name": "gpt_neox.layers.21.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29527040 + }, + { + "name": "gpt_neox.layers.21.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "0b15abf459a0ad2f7d9e89f58c1ba05e" + }, + { + "dataPath": "params_shard_33.bin", + "format": "raw-shard", + "nbytes": 29532160, + "records": [ + { + "name": "gpt_neox.layers.21.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.21.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "gpt_neox.layers.21.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 11059200 + }, + { + "name": "gpt_neox.layers.21.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11074560 + }, + { + "name": "gpt_neox.layers.21.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14351360 + }, + { + "name": "gpt_neox.layers.21.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.21.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.21.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.21.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 29511680 + } + ], + "md5sum": "731d088ab42e0f1bb70c495b15fe43e6" + }, + { + "dataPath": "params_shard_34.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.21.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.21.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.21.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.22.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.22.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "gpt_neox.layers.22.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.22.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.22.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.22.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "gpt_neox.layers.22.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 25830400 + }, + { + "name": "gpt_neox.layers.22.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25845760 + }, + { + "name": "gpt_neox.layers.22.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29122560 + }, + { + "name": "gpt_neox.layers.22.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "2d44a2fde063cf9f315a6df2dc6b8208" + }, + { + "dataPath": "params_shard_35.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.22.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.22.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.22.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.22.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.22.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.22.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "gpt_neox.layers.23.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "gpt_neox.layers.23.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29521920 + }, + { + "name": "gpt_neox.layers.23.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29527040 + }, + { + "name": "gpt_neox.layers.23.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "fd26b12198a0035b0ca8766fa5d28d07" + }, + { + "dataPath": "params_shard_36.bin", + "format": "raw-shard", + "nbytes": 29532160, + "records": [ + { + "name": "gpt_neox.layers.23.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.23.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "gpt_neox.layers.23.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 11059200 + }, + { + "name": "gpt_neox.layers.23.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11074560 + }, + { + "name": "gpt_neox.layers.23.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14351360 + }, + { + "name": "gpt_neox.layers.23.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.23.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.23.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.23.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 29511680 + } + ], + "md5sum": "7dc7c0ce49591209ebdb3dcf83997d2f" + }, + { + "dataPath": "params_shard_37.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.23.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.23.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.23.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.24.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.24.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "gpt_neox.layers.24.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.24.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.24.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.24.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "gpt_neox.layers.24.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 25830400 + }, + { + "name": "gpt_neox.layers.24.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25845760 + }, + { + "name": "gpt_neox.layers.24.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29122560 + }, + { + "name": "gpt_neox.layers.24.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "216e321cbc7523dffb0ca82a2329028b" + }, + { + "dataPath": "params_shard_38.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.24.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.24.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.24.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.24.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.24.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.24.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "gpt_neox.layers.25.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "gpt_neox.layers.25.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29521920 + }, + { + "name": "gpt_neox.layers.25.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29527040 + }, + { + "name": "gpt_neox.layers.25.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "05126cc45972fa07e105278eb8d0cc27" + }, + { + "dataPath": "params_shard_39.bin", + "format": "raw-shard", + "nbytes": 29532160, + "records": [ + { + "name": "gpt_neox.layers.25.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.25.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "gpt_neox.layers.25.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 11059200 + }, + { + "name": "gpt_neox.layers.25.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11074560 + }, + { + "name": "gpt_neox.layers.25.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14351360 + }, + { + "name": "gpt_neox.layers.25.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.25.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.25.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.25.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 29511680 + } + ], + "md5sum": "9b12e6d0f01f441a75056b61017b6df8" + }, + { + "dataPath": "params_shard_40.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.25.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.25.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.25.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.26.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.26.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "gpt_neox.layers.26.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.26.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.26.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.26.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "gpt_neox.layers.26.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 25830400 + }, + { + "name": "gpt_neox.layers.26.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25845760 + }, + { + "name": "gpt_neox.layers.26.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29122560 + }, + { + "name": "gpt_neox.layers.26.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "f8a3dd9afc148e2daa5449b960508192" + }, + { + "dataPath": "params_shard_41.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.26.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.26.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.26.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.26.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.26.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.26.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "gpt_neox.layers.27.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "gpt_neox.layers.27.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29521920 + }, + { + "name": "gpt_neox.layers.27.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29527040 + }, + { + "name": "gpt_neox.layers.27.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "aa49daa8e20aa7459d22ab83a69fc503" + }, + { + "dataPath": "params_shard_42.bin", + "format": "raw-shard", + "nbytes": 29532160, + "records": [ + { + "name": "gpt_neox.layers.27.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.27.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "gpt_neox.layers.27.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 11059200 + }, + { + "name": "gpt_neox.layers.27.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11074560 + }, + { + "name": "gpt_neox.layers.27.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14351360 + }, + { + "name": "gpt_neox.layers.27.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.27.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.27.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.27.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 29511680 + } + ], + "md5sum": "90e9ac2abd851feaf663da6feb7d0f37" + }, + { + "dataPath": "params_shard_43.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.27.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.27.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.27.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.28.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.28.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "gpt_neox.layers.28.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.28.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.28.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.28.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "gpt_neox.layers.28.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 25830400 + }, + { + "name": "gpt_neox.layers.28.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25845760 + }, + { + "name": "gpt_neox.layers.28.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29122560 + }, + { + "name": "gpt_neox.layers.28.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "67403ca262278abdf4daad45ed10074a" + }, + { + "dataPath": "params_shard_44.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.28.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.28.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.28.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.28.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.28.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.28.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "gpt_neox.layers.29.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "gpt_neox.layers.29.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29521920 + }, + { + "name": "gpt_neox.layers.29.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29527040 + }, + { + "name": "gpt_neox.layers.29.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "a4ca6638f285e7977dce4245dbf47232" + }, + { + "dataPath": "params_shard_45.bin", + "format": "raw-shard", + "nbytes": 29532160, + "records": [ + { + "name": "gpt_neox.layers.29.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.29.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "gpt_neox.layers.29.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 11059200 + }, + { + "name": "gpt_neox.layers.29.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11074560 + }, + { + "name": "gpt_neox.layers.29.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14351360 + }, + { + "name": "gpt_neox.layers.29.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.29.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.29.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.29.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 29511680 + } + ], + "md5sum": "20b135f55de8f071ecaeb1cac0187b86" + }, + { + "dataPath": "params_shard_46.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.29.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.29.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.29.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.30.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.layers.30.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "gpt_neox.layers.30.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.30.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.30.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 14771200 + }, + { + "name": "gpt_neox.layers.30.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 24601600 + }, + { + "name": "gpt_neox.layers.30.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 25830400 + }, + { + "name": "gpt_neox.layers.30.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 25845760 + }, + { + "name": "gpt_neox.layers.30.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 29122560 + }, + { + "name": "gpt_neox.layers.30.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "35eae9391ab9c54cbbd0115c09b72b08" + }, + { + "dataPath": "params_shard_47.bin", + "format": "raw-shard", + "nbytes": 29537280, + "records": [ + { + "name": "gpt_neox.layers.30.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.30.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.30.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.layers.30.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.30.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.30.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "gpt_neox.layers.31.input_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "gpt_neox.layers.31.input_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29521920 + }, + { + "name": "gpt_neox.layers.31.post_attention_layernorm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29527040 + }, + { + "name": "gpt_neox.layers.31.post_attention_layernorm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29532160 + } + ], + "md5sum": "8e9f9d9b9f797cb30a29c406a041a552" + }, + { + "dataPath": "params_shard_48.bin", + "format": "raw-shard", + "nbytes": 29532160, + "records": [ + { + "name": "gpt_neox.layers.31.attention.query_key_value.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.31.attention.query_key_value.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "gpt_neox.layers.31.attention.query_key_value.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 11059200 + }, + { + "name": "gpt_neox.layers.31.attention.dense.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 11074560 + }, + { + "name": "gpt_neox.layers.31.attention.dense.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 14351360 + }, + { + "name": "gpt_neox.layers.31.attention.dense.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14760960 + }, + { + "name": "gpt_neox.layers.31.mlp.dense_h_to_4h.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14766080 + }, + { + "name": "gpt_neox.layers.31.mlp.dense_h_to_4h.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27873280 + }, + { + "name": "gpt_neox.layers.31.mlp.dense_h_to_4h.bias", + "shape": [ + 10240 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 20480, + "byteOffset": 29511680 + } + ], + "md5sum": "d06a9164109647042761d6b83f18ba2d" + }, + { + "dataPath": "params_shard_49.bin", + "format": "raw-shard", + "nbytes": 62914560, + "records": [ + { + "name": "embed_out.q_weight", + "shape": [ + 49152, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 62914560, + "byteOffset": 0 + } + ], + "md5sum": "8c7c2ebd5f66480a760ce392b0bb68b7" + }, + { + "dataPath": "params_shard_50.bin", + "format": "raw-shard", + "nbytes": 22625280, + "records": [ + { + "name": "gpt_neox.layers.31.mlp.dense_4h_to_h.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "gpt_neox.layers.31.mlp.dense_4h_to_h.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "gpt_neox.layers.31.mlp.dense_4h_to_h.bias", + "shape": [ + 2560 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "gpt_neox.final_layer_norm.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14750720 + }, + { + "name": "gpt_neox.final_layer_norm.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14755840 + }, + { + "name": "embed_out.q_scale", + "shape": [ + 49152, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 7864320, + "byteOffset": 14760960 + } + ], + "md5sum": "3fc71222f4b51f72975a606d34f0735a" + } + ] +} \ No newline at end of file