diff --git "a/ndarray-cache.json" "b/ndarray-cache.json" new file mode 100755--- /dev/null +++ "b/ndarray-cache.json" @@ -0,0 +1,12104 @@ +{ + "metadata": { + "ParamSize": 807 + }, + "records": [ + { + "dataPath": "params_shard_0.bin", + "format": "raw-shard", + "nbytes": 131072000, + "records": [ + { + "name": "param_0", + "shape": [ + 32000, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 131072000, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_1.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_2", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_2.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_4", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_3.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_6", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_4.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_7", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_5.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_8", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_6.bin", + "format": "raw-shard", + "nbytes": 25821184, + "records": [ + { + "name": "param_1", + "shape": [ + 32000, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384000, + "byteOffset": 0 + }, + { + "name": "param_3", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 16384000 + }, + { + "name": "param_5", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 21626880 + } + ] + }, + { + "dataPath": "params_shard_7.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_12", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_8.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_14", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_9.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_16", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_10.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_17", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_11.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_18", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_12.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_9", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_10", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_11", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_13", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_15", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_13.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_22", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_14.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_24", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_15.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_26", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_16.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_27", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_17.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_28", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_18.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_19", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_20", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_21", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_23", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_25", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_19.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_32", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_20.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_34", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_21.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_36", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_22.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_37", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_23.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_38", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_24.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_29", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_30", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_31", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_33", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_35", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_25.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_42", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_26.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_44", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_27.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_46", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_28.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_47", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_29.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_48", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_30.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_39", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_40", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_41", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_43", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_45", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_31.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_52", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_32.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_54", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_33.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_56", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_34.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_57", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_35.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_58", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_36.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_49", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_50", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_51", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_53", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_55", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_37.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_62", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_38.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_64", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_39.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_66", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_40.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_67", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_41.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_68", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_42.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_59", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_60", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_61", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_63", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_65", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_43.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_72", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_44.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_74", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_45.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_76", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_46.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_77", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_47.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_78", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_48.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_69", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_70", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_71", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_73", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_75", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_49.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_82", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_50.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_84", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_51.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_86", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_52.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_87", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_53.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_88", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_54.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_79", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_80", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_81", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_83", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_85", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_55.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_92", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_56.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_94", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_57.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_96", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_58.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_97", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_59.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_98", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_60.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_89", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_90", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_91", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_93", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_95", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_61.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_102", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_62.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_104", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_63.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_106", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_64.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_107", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_65.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_108", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_66.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_99", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_100", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_101", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_103", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_105", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_67.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_112", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_68.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_114", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_69.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_116", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_70.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_117", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_71.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_118", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_72.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_109", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_110", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_111", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_113", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_115", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_73.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_122", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_74.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_124", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_75.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_126", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_76.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_127", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_77.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_128", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_78.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_119", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_120", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_121", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_123", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_125", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_79.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_132", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_80.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_134", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_81.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_136", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_82.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_137", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_83.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_138", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_84.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_129", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_130", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_131", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_133", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_135", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_85.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_142", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_86.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_144", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_87.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_146", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_88.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_147", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_89.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_148", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_90.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_139", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_140", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_141", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_143", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_145", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_91.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_152", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_92.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_154", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_93.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_156", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_94.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_157", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_95.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_158", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_96.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_149", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_150", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_151", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_153", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_155", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_97.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_162", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_98.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_164", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_99.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_166", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_100.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_167", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_101.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_168", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_102.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_159", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_160", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_161", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_163", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_165", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_103.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_172", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_104.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_174", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_105.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_176", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_106.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_177", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_107.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_178", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_108.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_169", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_170", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_171", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_173", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_175", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_109.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_182", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_110.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_184", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_111.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_186", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_112.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_187", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_113.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_188", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_114.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_179", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_180", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_181", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_183", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_185", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_115.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_192", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_116.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_194", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_117.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_196", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_118.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_197", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_119.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_198", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_120.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_189", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_190", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_191", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_193", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_195", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_121.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_202", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_122.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_204", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_123.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_206", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_124.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_207", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_125.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_208", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_126.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_199", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_200", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_201", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_203", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_205", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_127.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_212", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_128.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_214", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_129.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_216", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_130.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_217", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_131.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_218", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_132.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_209", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_210", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_211", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_213", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_215", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_133.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_222", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_134.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_224", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_135.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_226", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_136.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_227", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_137.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_228", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_138.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_219", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_220", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_221", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_223", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_225", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_139.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_232", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_140.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_234", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_141.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_236", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_142.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_237", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_143.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_238", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_144.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_229", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_230", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_231", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_233", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_235", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_145.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_242", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_146.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_244", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_147.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_246", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_148.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_247", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_149.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_248", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_150.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_239", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_240", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_241", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_243", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_245", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_151.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_252", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_152.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_254", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_153.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_256", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_154.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_257", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_155.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_258", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_156.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_249", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_250", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_251", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_253", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_255", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_157.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_262", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_158.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_264", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_159.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_266", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_160.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_267", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_161.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_268", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_162.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_259", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_260", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_261", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_263", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_265", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_163.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_272", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_164.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_274", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_165.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_276", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_166.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_277", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_167.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_278", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_168.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_269", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_270", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_271", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_273", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_275", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_169.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_282", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_170.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_284", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_171.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_286", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_172.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_287", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_173.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_288", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_174.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_279", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_280", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_281", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_283", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_285", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_175.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_292", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_176.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_294", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_177.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_296", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_178.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_297", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_179.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_298", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_180.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_289", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_290", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_291", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_293", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_295", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_181.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_302", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_182.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_304", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_183.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_306", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_184.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_307", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_185.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_308", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_186.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_299", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_300", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_301", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_303", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_305", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_187.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_312", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_188.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_314", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_189.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_316", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_190.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_317", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_191.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_318", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_192.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_309", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_310", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_311", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_313", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_315", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_193.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_322", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_194.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_324", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_195.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_326", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_196.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_327", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_197.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_328", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_198.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_319", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_320", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_321", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_323", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_325", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_199.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_332", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_200.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_334", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_201.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_336", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_202.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_337", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_203.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_338", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_204.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_329", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_330", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_331", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_333", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_335", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_205.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_342", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_206.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_344", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_207.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_346", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_208.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_347", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_209.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_348", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_210.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_339", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_340", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_341", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_343", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_345", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_211.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_352", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_212.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_354", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_213.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_356", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_214.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_357", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_215.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_358", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_216.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_349", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_350", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_351", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_353", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_355", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_217.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_362", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_218.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_364", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_219.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_366", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_220.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_367", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_221.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_368", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_222.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_359", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_360", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_361", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_363", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_365", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_223.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_372", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_224.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_374", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_225.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_376", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_226.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_377", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_227.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_378", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_228.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_369", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_370", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_371", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_373", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_375", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_229.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_382", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_230.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_384", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_231.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_386", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_232.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_387", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_233.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_388", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_234.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_379", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_380", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_381", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_383", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_385", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_235.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_392", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_236.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_394", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_237.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_396", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_238.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_397", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_239.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_398", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_240.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_389", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_390", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_391", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_393", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_395", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_241.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_402", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_242.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_404", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_243.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_406", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_244.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_407", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_245.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_408", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_246.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_399", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_400", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_401", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_403", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_405", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_247.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_412", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_248.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_414", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_249.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_416", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_250.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_417", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_251.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_418", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_252.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_409", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_410", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_411", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_413", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_415", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_253.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_422", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_254.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_424", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_255.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_426", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_256.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_427", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_257.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_428", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_258.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_419", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_420", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_421", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_423", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_425", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_259.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_432", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_260.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_434", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_261.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_436", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_262.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_437", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_263.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_438", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_264.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_429", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_430", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_431", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_433", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_435", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_265.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_442", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_266.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_444", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_267.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_446", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_268.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_447", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_269.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_448", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_270.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_439", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_440", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_441", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_443", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_445", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_271.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_452", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_272.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_454", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_273.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_456", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_274.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_457", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_275.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_458", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_276.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_449", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_450", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_451", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_453", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_455", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_277.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_462", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_278.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_464", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_279.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_466", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_280.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_467", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_281.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_468", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_282.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_459", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_460", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_461", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_463", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_465", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_283.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_472", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_284.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_474", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_285.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_476", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_286.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_477", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_287.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_478", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_288.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_469", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_470", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_471", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_473", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_475", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_289.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_482", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_290.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_484", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_291.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_486", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_292.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_487", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_293.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_488", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_294.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_479", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_480", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_481", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_483", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_485", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_295.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_492", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_296.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_494", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_297.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_496", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_298.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_497", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_299.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_498", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_300.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_489", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_490", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_491", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_493", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_495", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_301.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_502", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_302.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_504", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_303.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_506", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_304.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_507", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_305.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_508", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_306.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_499", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_500", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_501", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_503", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_505", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_307.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_512", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_308.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_514", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_309.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_516", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_310.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_517", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_311.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_518", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_312.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_509", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_510", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_511", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_513", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_515", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_313.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_522", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_314.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_524", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_315.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_526", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_316.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_527", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_317.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_528", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_318.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_519", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_520", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_521", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_523", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_525", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_319.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_532", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_320.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_534", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_321.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_536", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_322.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_537", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_323.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_538", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_324.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_529", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_530", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_531", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_533", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_535", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_325.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_542", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_326.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_544", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_327.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_546", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_328.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_547", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_329.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_548", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_330.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_539", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_540", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_541", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_543", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_545", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_331.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_552", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_332.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_554", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_333.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_556", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_334.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_557", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_335.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_558", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_336.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_549", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_550", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_551", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_553", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_555", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_337.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_562", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_338.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_564", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_339.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_566", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_340.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_567", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_341.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_568", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_342.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_559", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_560", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_561", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_563", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_565", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_343.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_572", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_344.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_574", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_345.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_576", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_346.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_577", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_347.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_578", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_348.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_569", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_570", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_571", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_573", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_575", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_349.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_582", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_350.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_584", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_351.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_586", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_352.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_587", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_353.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_588", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_354.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_579", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_580", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_581", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_583", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_585", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_355.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_592", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_356.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_594", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_357.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_596", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_358.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_597", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_359.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_598", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_360.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_589", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_590", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_591", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_593", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_595", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_361.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_602", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_362.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_604", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_363.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_606", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_364.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_607", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_365.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_608", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_366.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_599", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_600", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_601", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_603", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_605", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_367.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_612", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_368.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_614", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_369.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_616", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_370.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_617", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_371.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_618", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_372.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_609", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_610", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_611", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_613", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_615", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_373.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_622", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_374.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_624", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_375.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_626", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_376.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_627", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_377.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_628", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_378.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_619", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_620", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_621", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_623", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_625", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_379.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_632", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_380.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_634", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_381.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_636", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_382.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_637", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_383.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_638", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_384.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_629", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_630", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_631", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_633", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_635", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_385.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_642", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_386.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_644", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_387.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_646", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_388.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_647", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_389.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_648", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_390.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_639", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_640", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_641", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_643", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_645", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_391.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_652", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_392.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_654", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_393.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_656", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_394.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_657", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_395.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_658", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_396.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_649", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_650", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_651", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_653", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_655", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_397.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_662", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_398.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_664", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_399.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_666", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_400.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_667", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_401.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_668", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_402.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_659", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_660", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_661", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_663", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_665", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_403.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_672", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_404.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_674", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_405.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_676", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_406.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_677", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_407.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_678", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_408.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_669", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_670", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_671", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_673", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_675", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_409.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_682", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_410.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_684", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_411.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_686", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_412.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_687", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_413.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_688", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_414.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_679", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_680", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_681", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_683", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_685", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_415.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_692", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_416.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_694", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_417.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_696", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_418.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_697", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_419.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_698", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_420.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_689", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_690", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_691", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_693", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_695", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_421.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_702", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_422.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_704", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_423.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_706", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_424.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_707", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_425.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_708", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_426.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_699", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_700", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_701", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_703", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_705", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_427.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_712", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_428.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_714", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_429.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_716", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_430.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_717", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_431.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_718", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_432.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_709", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_710", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_711", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_713", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_715", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_433.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_722", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_434.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_724", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_435.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_726", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_436.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_727", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_437.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_728", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_438.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_719", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_720", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_721", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_723", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_725", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_439.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_732", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_440.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_734", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_441.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_736", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_442.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_737", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_443.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_738", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_444.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_729", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_730", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_731", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_733", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_735", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_445.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_742", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_446.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_744", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_447.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_746", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_448.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_747", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_449.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_748", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_450.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_739", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_740", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_741", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_743", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_745", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_451.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_752", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_452.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_754", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_453.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_756", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_454.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_757", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_455.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_758", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_456.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_749", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_750", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_751", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_753", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_755", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_457.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_762", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_458.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_764", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_459.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_766", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_460.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_767", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_461.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_768", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_462.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_759", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_760", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_761", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_763", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_765", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_463.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_772", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_464.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_774", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_465.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_776", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_466.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_777", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_467.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_778", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_468.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_769", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_770", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_771", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_773", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_775", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_469.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_782", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_470.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_784", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_471.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_786", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_472.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_787", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_473.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_788", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_474.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_779", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_780", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_781", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_783", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_785", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_475.bin", + "format": "raw-shard", + "nbytes": 41943040, + "records": [ + { + "name": "param_792", + "shape": [ + 10240, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 41943040, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_476.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_794", + "shape": [ + 8192, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_477.bin", + "format": "raw-shard", + "nbytes": 234881024, + "records": [ + { + "name": "param_796", + "shape": [ + 57344, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 234881024, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_478.bin", + "format": "raw-shard", + "nbytes": 29360128, + "records": [ + { + "name": "param_797", + "shape": [ + 57344, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 29360128, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_479.bin", + "format": "raw-shard", + "nbytes": 117440512, + "records": [ + { + "name": "param_798", + "shape": [ + 8192, + 3584 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 117440512, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_480.bin", + "format": "raw-shard", + "nbytes": 24150016, + "records": [ + { + "name": "param_789", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_790", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_791", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_793", + "shape": [ + 10240, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5242880, + "byteOffset": 14712832 + }, + { + "name": "param_795", + "shape": [ + 8192, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 19955712 + } + ] + }, + { + "dataPath": "params_shard_481.bin", + "format": "raw-shard", + "nbytes": 131072000, + "records": [ + { + "name": "param_803", + "shape": [ + 32000, + 1024 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 131072000, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_482.bin", + "format": "raw-shard", + "nbytes": 32161792, + "records": [ + { + "name": "param_799", + "shape": [ + 8192, + 896 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 14680064, + "byteOffset": 0 + }, + { + "name": "param_800", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14680064 + }, + { + "name": "param_801", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14696448 + }, + { + "name": "param_802", + "shape": [ + 8192 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384, + "byteOffset": 14712832 + }, + { + "name": "param_804", + "shape": [ + 32000, + 256 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 16384000, + "byteOffset": 14729216 + }, + { + "name": "param_805", + "shape": [ + 2048, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 524288, + "byteOffset": 31113216 + }, + { + "name": "param_806", + "shape": [ + 2048, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 524288, + "byteOffset": 31637504 + } + ] + } + ] +} \ No newline at end of file