diff --git "a/ndarray-cache.json" "b/ndarray-cache.json" new file mode 100644--- /dev/null +++ "b/ndarray-cache.json" @@ -0,0 +1,4636 @@ +{ + "metadata": { + "ParamSize": 390 + }, + "records": [ + { + "dataPath": "params_shard_0.bin", + "format": "raw-shard", + "nbytes": 64389120, + "records": [ + { + "name": "param_0", + "shape": [ + 50304, + 320 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 64389120, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_1.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_6", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_2.bin", + "format": "raw-shard", + "nbytes": 25006080, + "records": [ + { + "name": "param_1", + "shape": [ + 50304, + 80 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8048640, + "byteOffset": 0 + }, + { + "name": "param_2", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 8048640 + }, + { + "name": "param_3", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 17879040 + }, + { + "name": "param_4", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 19107840 + }, + { + "name": "param_5", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 22384640 + }, + { + "name": "param_7", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 22794240 + } + ] + }, + { + "dataPath": "params_shard_3.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_18", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_4.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_8", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_9", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_10", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_11", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_12", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_13", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_14", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_15", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_16", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_17", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_19", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_5.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_30", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_6.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_20", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_21", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_22", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_23", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_24", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_25", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_26", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_27", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_28", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_29", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_31", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_7.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_42", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_8.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_32", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_33", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_34", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_35", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_36", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_37", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_38", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_39", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_40", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_41", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_43", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_9.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_54", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_10.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_44", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_45", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_46", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_47", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_48", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_49", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_50", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_51", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_52", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_53", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_55", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_11.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_66", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_12.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_56", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_57", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_58", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_59", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_60", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_61", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_62", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_63", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_64", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_65", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_67", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_13.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_78", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_14.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_68", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_69", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_70", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_71", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_72", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_73", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_74", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_75", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_76", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_77", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_79", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_15.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_90", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_16.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_80", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_81", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_82", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_83", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_84", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_85", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_86", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_87", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_88", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_89", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_91", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_17.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_102", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_18.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_92", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_93", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_94", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_95", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_96", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_97", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_98", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_99", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_100", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_101", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_103", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_19.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_114", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_20.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_104", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_105", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_106", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_107", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_108", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_109", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_110", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_111", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_112", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_113", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_115", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_21.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_126", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_22.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_116", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_117", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_118", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_119", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_120", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_121", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_122", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_123", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_124", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_125", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_127", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_23.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_138", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_24.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_128", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_129", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_130", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_131", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_132", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_133", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_134", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_135", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_136", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_137", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_139", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_25.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_150", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_26.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_140", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_141", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_142", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_143", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_144", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_145", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_146", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_147", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_148", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_149", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_151", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_27.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_162", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_28.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_152", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_153", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_154", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_155", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_156", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_157", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_158", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_159", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_160", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_161", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_163", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_29.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_174", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_30.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_164", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_165", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_166", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_167", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_168", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_169", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_170", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_171", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_172", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_173", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_175", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_31.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_186", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_32.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_176", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_177", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_178", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_179", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_180", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_181", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_182", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_183", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_184", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_185", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_187", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_33.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_198", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_34.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_188", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_189", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_190", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_191", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_192", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_193", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_194", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_195", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_196", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_197", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_199", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_35.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_210", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_36.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_200", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_201", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_202", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_203", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_204", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_205", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_206", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_207", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_208", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_209", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_211", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_37.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_222", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_38.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_212", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_213", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_214", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_215", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_216", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_217", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_218", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_219", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_220", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_221", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_223", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_39.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_234", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_40.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_224", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_225", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_226", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_227", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_228", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_229", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_230", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_231", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_232", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_233", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_235", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_41.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_246", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_42.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_236", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_237", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_238", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_239", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_240", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_241", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_242", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_243", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_244", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_245", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_247", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_43.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_258", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_44.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_248", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_249", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_250", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_251", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_252", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_253", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_254", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_255", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_256", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_257", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_259", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_45.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_270", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_46.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_260", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_261", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_262", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_263", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_264", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_265", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_266", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_267", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_268", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_269", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_271", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_47.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_282", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_48.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_272", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_273", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_274", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_275", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_276", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_277", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_278", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_279", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_280", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_281", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_283", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_49.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_294", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_50.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_284", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_285", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_286", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_287", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_288", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_289", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_290", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_291", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_292", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_293", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_295", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_51.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_306", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_52.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_296", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_297", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_298", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_299", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_300", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_301", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_302", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_303", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_304", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_305", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_307", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_53.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_318", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_54.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_308", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_309", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_310", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_311", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_312", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_313", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_314", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_315", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_316", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_317", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_319", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_55.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_330", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_56.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_320", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_321", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_322", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_323", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_324", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_325", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_326", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_327", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_328", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_329", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_331", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_57.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_342", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_58.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_332", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_333", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_334", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_335", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_336", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_337", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_338", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_339", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_340", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_341", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_343", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_59.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_354", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_60.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_344", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_345", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_346", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_347", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_348", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_349", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_350", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_351", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_352", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_353", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_355", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_61.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_366", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_62.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_356", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_357", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_358", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_359", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_360", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_361", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_362", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_363", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_364", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_365", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_367", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_63.bin", + "format": "raw-shard", + "nbytes": 17694720, + "records": [ + { + "name": "param_378", + "shape": [ + 320, + 13824 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 17694720, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_64.bin", + "format": "raw-shard", + "nbytes": 26931200, + "records": [ + { + "name": "param_368", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_369", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_370", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_371", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_372", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_373", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_374", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 9830400, + "byteOffset": 9973760 + }, + { + "name": "param_375", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1228800, + "byteOffset": 19804160 + }, + { + "name": "param_376", + "shape": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 3276800, + "byteOffset": 21032960 + }, + { + "name": "param_377", + "shape": [ + 80, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 409600, + "byteOffset": 24309760 + }, + { + "name": "param_379", + "shape": [ + 80, + 13824 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 2211840, + "byteOffset": 24719360 + } + ] + }, + { + "dataPath": "params_shard_65.bin", + "format": "raw-shard", + "nbytes": 64389120, + "records": [ + { + "name": "param_388", + "shape": [ + 320, + 50304 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 64389120, + "byteOffset": 0 + } + ] + }, + { + "dataPath": "params_shard_66.bin", + "format": "raw-shard", + "nbytes": 18032640, + "records": [ + { + "name": "param_380", + "shape": [ + 864, + 2560 + ], + "dtype": "uint32", + "format": "raw", + "nbytes": 8847360, + "byteOffset": 0 + }, + { + "name": "param_381", + "shape": [ + 216, + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 1105920, + "byteOffset": 8847360 + }, + { + "name": "param_382", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9953280 + }, + { + "name": "param_383", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9958400 + }, + { + "name": "param_384", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9963520 + }, + { + "name": "param_385", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9968640 + }, + { + "name": "param_386", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9973760 + }, + { + "name": "param_387", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 5120, + "byteOffset": 9978880 + }, + { + "name": "param_389", + "shape": [ + 80, + 50304 + ], + "dtype": "float16", + "format": "raw", + "nbytes": 8048640, + "byteOffset": 9984000 + } + ] + } + ] +} \ No newline at end of file