diff --git "a/ndarray-cache.json" "b/ndarray-cache.json" new file mode 100644--- /dev/null +++ "b/ndarray-cache.json" @@ -0,0 +1,9427 @@ +{ + "metadata": { + "ParamSize": 806 + }, + "records": [ + { + "dataPath": "params_shard_0.bin", + "format": "raw-shard", + "nbytes": 411869184, + "records": [ + { + "name": "param_0", + "shape": [ + 50277, + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 411869184, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_1.bin", + "format": "raw-shard", + "nbytes": 28418048, + "records": [ + { + "name": "param_1", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 0 + }, + { + "name": "param_2", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 8192 + }, + { + "name": "param_3", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 16384 + }, + { + "name": "param_4", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 24576 + }, + { + "name": "param_5", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 32768 + }, + { + "name": "param_6", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 40960 + }, + { + "name": "param_7", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 49152 + }, + { + "name": "param_8", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 65536 + }, + { + "name": "param_9", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 81920 + }, + { + "name": "param_10", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 90112 + }, + { + "name": "param_11", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 98304 + }, + { + "name": "param_12", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 106496 + }, + { + "name": "param_13", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8495104 + }, + { + "name": "param_14", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9543680 + }, + { + "name": "param_15", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17932288 + }, + { + "name": "param_16", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18980864 + }, + { + "name": "param_17", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27369472 + } + ] + }, + { + "dataPath": "params_shard_2.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_22", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_3.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_26", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_4.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_18", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_19", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_20", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_21", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_23", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_24", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_25", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_27", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_28", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_29", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_30", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_31", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_32", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_33", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_34", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_35", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_36", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_5.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_37", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_38", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_39", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_40", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_41", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_42", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_6.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_47", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_7.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_51", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_8.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_43", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_44", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_45", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_46", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_48", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_49", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_50", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_52", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_53", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_54", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_55", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_56", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_57", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_58", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_59", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_60", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_61", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_9.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_62", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_63", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_64", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_65", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_66", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_67", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_10.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_72", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_11.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_76", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_12.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_68", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_69", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_70", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_71", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_73", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_74", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_75", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_77", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_78", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_79", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_80", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_81", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_82", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_83", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_84", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_85", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_86", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_13.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_87", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_88", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_89", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_90", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_91", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_92", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_14.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_97", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_15.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_101", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_16.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_93", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_94", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_95", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_96", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_98", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_99", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_100", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_102", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_103", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_104", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_105", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_106", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_107", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_108", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_109", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_110", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_111", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_17.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_112", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_113", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_114", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_115", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_116", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_117", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_18.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_122", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_19.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_126", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_20.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_118", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_119", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_120", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_121", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_123", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_124", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_125", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_127", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_128", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_129", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_130", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_131", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_132", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_133", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_134", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_135", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_136", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_21.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_137", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_138", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_139", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_140", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_141", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_142", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_22.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_147", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_23.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_151", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_24.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_143", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_144", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_145", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_146", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_148", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_149", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_150", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_152", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_153", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_154", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_155", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_156", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_157", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_158", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_159", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_160", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_161", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_25.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_162", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_163", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_164", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_165", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_166", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_167", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_26.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_172", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_27.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_176", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_28.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_168", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_169", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_170", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_171", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_173", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_174", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_175", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_177", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_178", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_179", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_180", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_181", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_182", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_183", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_184", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_185", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_186", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_29.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_187", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_188", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_189", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_190", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_191", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_192", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_30.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_197", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_31.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_201", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_32.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_193", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_194", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_195", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_196", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_198", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_199", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_200", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_202", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_203", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_204", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_205", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_206", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_207", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_208", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_209", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_210", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_211", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_33.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_212", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_213", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_214", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_215", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_216", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_217", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_34.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_222", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_35.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_226", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_36.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_218", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_219", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_220", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_221", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_223", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_224", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_225", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_227", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_228", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_229", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_230", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_231", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_232", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_233", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_234", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_235", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_236", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_37.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_237", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_238", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_239", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_240", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_241", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_242", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_38.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_247", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_39.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_251", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_40.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_243", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_244", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_245", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_246", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_248", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_249", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_250", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_252", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_253", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_254", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_255", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_256", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_257", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_258", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_259", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_260", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_261", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_41.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_262", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_263", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_264", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_265", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_266", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_267", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_42.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_272", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_43.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_276", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_44.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_268", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_269", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_270", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_271", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_273", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_274", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_275", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_277", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_278", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_279", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_280", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_281", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_282", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_283", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_284", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_285", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_286", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_45.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_287", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_288", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_289", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_290", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_291", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_292", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_46.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_297", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_47.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_301", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_48.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_293", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_294", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_295", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_296", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_298", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_299", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_300", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_302", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_303", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_304", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_305", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_306", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_307", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_308", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_309", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_310", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_311", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_49.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_312", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_313", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_314", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_315", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_316", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_317", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_50.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_322", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_51.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_326", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_52.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_318", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_319", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_320", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_321", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_323", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_324", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_325", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_327", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_328", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_329", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_330", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_331", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_332", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_333", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_334", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_335", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_336", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_53.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_337", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_338", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_339", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_340", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_341", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_342", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_54.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_347", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_55.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_351", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_56.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_343", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_344", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_345", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_346", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_348", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_349", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_350", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_352", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_353", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_354", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_355", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_356", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_357", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_358", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_359", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_360", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_361", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_57.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_362", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_363", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_364", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_365", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_366", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_367", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_58.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_372", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_59.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_376", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_60.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_368", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_369", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_370", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_371", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_373", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_374", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_375", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_377", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_378", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_379", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_380", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_381", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_382", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_383", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_384", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_385", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_386", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_61.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_387", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_388", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_389", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_390", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_391", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_392", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_62.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_397", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_63.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_401", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_64.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_393", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_394", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_395", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_396", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_398", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_399", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_400", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_402", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_403", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_404", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_405", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_406", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_407", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_408", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_409", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_410", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_411", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_65.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_412", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_413", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_414", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_415", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_416", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_417", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_66.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_422", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_67.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_426", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_68.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_418", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_419", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_420", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_421", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_423", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_424", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_425", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_427", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_428", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_429", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_430", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_431", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_432", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_433", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_434", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_435", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_436", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_69.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_437", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_438", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_439", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_440", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_441", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_442", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_70.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_447", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_71.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_451", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_72.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_443", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_444", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_445", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_446", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_448", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_449", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_450", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_452", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_453", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_454", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_455", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_456", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_457", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_458", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_459", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_460", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_461", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_73.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_462", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_463", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_464", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_465", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_466", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_467", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_74.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_472", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_75.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_476", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_76.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_468", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_469", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_470", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_471", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_473", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_474", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_475", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_477", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_478", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_479", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_480", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_481", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_482", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_483", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_484", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_485", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_486", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_77.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_487", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_488", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_489", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_490", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_491", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_492", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_78.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_497", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_79.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_501", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_80.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_493", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_494", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_495", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_496", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_498", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_499", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_500", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_502", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_503", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_504", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_505", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_506", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_507", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_508", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_509", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_510", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_511", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_81.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_512", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_513", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_514", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_515", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_516", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_517", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_82.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_522", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_83.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_526", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_84.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_518", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_519", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_520", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_521", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_523", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_524", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_525", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_527", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_528", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_529", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_530", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_531", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_532", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_533", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_534", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_535", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_536", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_85.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_537", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_538", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_539", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_540", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_541", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_542", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_86.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_547", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_87.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_551", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_88.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_543", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_544", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_545", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_546", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_548", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_549", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_550", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_552", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_553", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_554", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_555", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_556", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_557", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_558", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_559", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_560", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_561", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_89.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_562", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_563", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_564", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_565", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_566", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_567", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_90.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_572", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_91.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_576", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_92.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_568", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_569", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_570", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_571", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_573", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_574", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_575", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_577", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_578", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_579", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_580", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_581", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_582", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_583", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_584", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_585", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_586", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_93.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_587", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_588", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_589", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_590", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_591", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_592", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_94.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_597", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_95.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_601", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_96.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_593", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_594", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_595", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_596", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_598", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_599", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_600", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_602", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_603", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_604", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_605", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_606", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_607", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_608", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_609", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_610", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_611", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_97.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_612", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_613", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_614", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_615", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_616", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_617", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_98.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_622", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_99.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_626", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_100.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_618", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_619", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_620", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_621", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_623", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_624", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_625", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_627", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_628", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_629", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_630", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_631", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_632", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_633", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_634", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_635", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_636", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_101.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_637", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_638", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_639", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_640", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_641", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_642", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_102.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_647", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_103.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_651", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_104.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_643", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_644", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_645", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_646", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_648", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_649", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_650", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_652", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_653", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_654", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_655", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_656", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_657", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_658", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_659", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_660", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_661", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_105.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_662", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_663", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_664", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_665", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_666", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_667", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_106.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_672", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_107.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_676", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_108.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_668", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_669", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_670", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_671", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_673", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_674", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_675", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_677", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_678", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_679", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_680", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_681", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_682", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_683", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_684", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_685", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_686", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_109.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_687", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_688", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_689", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_690", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_691", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_692", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_110.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_697", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_111.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_701", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_112.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_693", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_694", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_695", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_696", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_698", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_699", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_700", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_702", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_703", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_704", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_705", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_706", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_707", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_708", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_709", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_710", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_711", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_113.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_712", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_713", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_714", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_715", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_716", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_717", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_114.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_722", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_115.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_726", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_116.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_718", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_719", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_720", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_721", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_723", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_724", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_725", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_727", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_728", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_729", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_730", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_731", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_732", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_733", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_734", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_735", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_736", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_117.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_737", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_738", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_739", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_740", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_741", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_742", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_118.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_747", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_119.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_751", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_120.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_743", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_744", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_745", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_746", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_748", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_749", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_750", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_752", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_753", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_754", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_755", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_756", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_757", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_758", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_759", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_760", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_761", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_121.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_762", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_763", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_764", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_765", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_766", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_767", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_122.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_772", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_123.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_776", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_124.bin", + "format": "raw-shard", + "nbytes": 27369472, + "records": [ + { + "name": "param_768", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_769", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_770", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_771", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_773", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_774", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_775", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_777", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_778", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_779", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + }, + { + "name": "param_780", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27295744 + }, + { + "name": "param_781", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27303936 + }, + { + "name": "param_782", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27312128 + }, + { + "name": "param_783", + "shape": [ + 4096 + ], + "dtype": "float32", + "format": "raw", + "nbytes": 16384, + "byteOffset": 27328512 + }, + { + "name": "param_784", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27344896 + }, + { + "name": "param_785", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27353088 + }, + { + "name": "param_786", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27361280 + } + ] + }, + { + "dataPath": "params_shard_125.bin", + "format": "raw-shard", + "nbytes": 28311552, + "records": [ + { + "name": "param_787", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_788", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_789", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 9437184 + }, + { + "name": "param_790", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 17825792 + }, + { + "name": "param_791", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 18874368 + }, + { + "name": "param_792", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 27262976 + } + ] + }, + { + "dataPath": "params_shard_126.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_797", + "shape": [ + 16384, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_127.bin", + "format": "raw-shard", + "nbytes": 33554432, + "records": [ + { + "name": "param_801", + "shape": [ + 4096, + 2048 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 33554432, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_128.bin", + "format": "raw-shard", + "nbytes": 411869184, + "records": [ + { + "name": "param_805", + "shape": [ + 50277, + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 411869184, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_129.bin", + "format": "raw-shard", + "nbytes": 27295744, + "records": [ + { + "name": "param_793", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 0 + }, + { + "name": "param_794", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 8388608 + }, + { + "name": "param_795", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9437184 + }, + { + "name": "param_796", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 9445376 + }, + { + "name": "param_798", + "shape": [ + 16384, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 9453568 + }, + { + "name": "param_799", + "shape": [ + 4096, + 512 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8388608, + "byteOffset": 13647872 + }, + { + "name": "param_800", + "shape": [ + 4096, + 128 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1048576, + "byteOffset": 22036480 + }, + { + "name": "param_802", + "shape": [ + 4096, + 512 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 4194304, + "byteOffset": 23085056 + }, + { + "name": "param_803", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27279360 + }, + { + "name": "param_804", + "shape": [ + 4096 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 27287552 + } + ] + } + ] +} \ No newline at end of file