diff --git "a/ndarray-cache.json" "b/ndarray-cache.json" new file mode 100644--- /dev/null +++ "b/ndarray-cache.json" @@ -0,0 +1,5227 @@ +{ + "metadata": { + "ParamSize": 455, + "ParamBytes": 1564948480.0, + "BitsPerParam": 4.503961083574167 + }, + "records": [ + { + "dataPath": "params_shard_0.bin", + "format": "raw-shard", + "nbytes": 65536000, + "records": [ + { + "name": "lm_head.linear.q_weight", + "shape": [ + 51200, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 65536000, + "byteOffset": 0 + } + ], + "md5sum": "cfb4f0e8a243b7dcaf2965bccbfff2a9" + }, + { + "dataPath": "params_shard_1.bin", + "format": "raw-shard", + "nbytes": 23086080, + "records": [ + { + "name": "lm_head.linear.bias", + "shape": [ + 51200 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 102400, + "byteOffset": 0 + }, + { + "name": "lm_head.linear.q_scale", + "shape": [ + 51200, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 8192000, + "byteOffset": 102400 + }, + { + "name": "lm_head.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 8294400 + }, + { + "name": "lm_head.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 8299520 + }, + { + "name": "transformer.h.30.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 8304640 + }, + { + "name": "transformer.h.30.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 8309760 + }, + { + "name": "transformer.h.30.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 8314880 + }, + { + "name": "transformer.h.30.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 8335360 + }, + { + "name": "transformer.h.30.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 21442560 + }, + { + "name": "transformer.h.30.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 23080960 + } + ], + "md5sum": "dc0f02c7ad5ef542cc435dc786a2bcdc" + }, + { + "dataPath": "params_shard_2.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.30.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.30.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.30.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.30.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.30.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 18027520 + }, + { + "name": "transformer.h.30.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 18437120 + }, + { + "name": "transformer.h.30.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.30.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 28282880 + }, + { + "name": "transformer.h.31.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "transformer.h.31.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "transformer.h.31.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "206b913f5b72fb164b4aafaff9a6771f" + }, + { + "dataPath": "params_shard_3.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.31.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.31.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.31.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.31.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.31.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.31.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.31.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.31.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 32778240 + }, + { + "name": "transformer.h.31.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 33187840 + } + ], + "md5sum": "5167e6ab22c0e1f8a3beae472dd85f43" + }, + { + "dataPath": "params_shard_4.bin", + "format": "raw-shard", + "nbytes": 65536000, + "records": [ + { + "name": "transformer.embd.q_weight", + "shape": [ + 51200, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 65536000, + "byteOffset": 0 + } + ], + "md5sum": "478ad7ae97b8be71515834fd36010ef4" + }, + { + "dataPath": "params_shard_5.bin", + "format": "raw-shard", + "nbytes": 32389120, + "records": [ + { + "name": "transformer.h.31.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.31.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.embd.q_scale", + "shape": [ + 51200, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 8192000, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.0.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 19251200 + }, + { + "name": "transformer.h.0.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 19256320 + }, + { + "name": "transformer.h.0.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 19261440 + }, + { + "name": "transformer.h.0.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 19281920 + } + ], + "md5sum": "2b0938bf4a77fc4b4639e8476eb67e45" + }, + { + "dataPath": "params_shard_6.bin", + "format": "raw-shard", + "nbytes": 31185920, + "records": [ + { + "name": "transformer.h.0.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 0 + }, + { + "name": "transformer.h.0.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 1638400 + }, + { + "name": "transformer.h.0.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 1643520 + }, + { + "name": "transformer.h.0.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.0.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 16389120 + }, + { + "name": "transformer.h.0.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 16394240 + }, + { + "name": "transformer.h.0.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 19671040 + }, + { + "name": "transformer.h.0.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 20080640 + }, + { + "name": "transformer.h.0.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 20096000 + }, + { + "name": "transformer.h.0.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 29926400 + }, + { + "name": "transformer.h.1.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 31155200 + }, + { + "name": "transformer.h.1.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 31160320 + }, + { + "name": "transformer.h.1.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 31165440 + } + ], + "md5sum": "88c2ff4564c8097d4c0153e2fe7a04cd" + }, + { + "dataPath": "params_shard_7.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.1.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.1.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.1.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.1.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.1.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.1.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.1.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.1.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 32778240 + }, + { + "name": "transformer.h.1.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 33187840 + } + ], + "md5sum": "976241350ad6ca09fffb26358a6c001f" + }, + { + "dataPath": "params_shard_8.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.1.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.1.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.10.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.10.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.10.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 11069440 + }, + { + "name": "transformer.h.10.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.10.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 24197120 + }, + { + "name": "transformer.h.10.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25835520 + } + ], + "md5sum": "b60eb4c7247778e2583791979ab695c3" + }, + { + "dataPath": "params_shard_9.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.10.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.10.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.10.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.10.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.10.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 18027520 + }, + { + "name": "transformer.h.10.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 18437120 + }, + { + "name": "transformer.h.10.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.10.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 28282880 + }, + { + "name": "transformer.h.11.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "transformer.h.11.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "transformer.h.11.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "ed68e1a164928e220b0e9f043c72a27a" + }, + { + "dataPath": "params_shard_10.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.11.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.11.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.11.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.11.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.11.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.11.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.11.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.11.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 32778240 + }, + { + "name": "transformer.h.11.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 33187840 + } + ], + "md5sum": "db41b4360a536a2d6093da6a09344a03" + }, + { + "dataPath": "params_shard_11.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.11.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.11.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.12.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.12.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.12.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 11069440 + }, + { + "name": "transformer.h.12.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.12.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 24197120 + }, + { + "name": "transformer.h.12.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25835520 + } + ], + "md5sum": "44ff722e76c0843f14c2518317420d61" + }, + { + "dataPath": "params_shard_12.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.12.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.12.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.12.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.12.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.12.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 18027520 + }, + { + "name": "transformer.h.12.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 18437120 + }, + { + "name": "transformer.h.12.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.12.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 28282880 + }, + { + "name": "transformer.h.13.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "transformer.h.13.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "transformer.h.13.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "fd3731c3748eb7520e9836437a86ba4b" + }, + { + "dataPath": "params_shard_13.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.13.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.13.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.13.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.13.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.13.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.13.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.13.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.13.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 32778240 + }, + { + "name": "transformer.h.13.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 33187840 + } + ], + "md5sum": "d738c53e2bbdd0167f777a99d404e24a" + }, + { + "dataPath": "params_shard_14.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.13.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.13.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.14.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.14.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.14.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 11069440 + }, + { + "name": "transformer.h.14.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.14.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 24197120 + }, + { + "name": "transformer.h.14.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25835520 + } + ], + "md5sum": "e06611f20973c1c533017e2c8f616875" + }, + { + "dataPath": "params_shard_15.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.14.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.14.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.14.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.14.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.14.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 18027520 + }, + { + "name": "transformer.h.14.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 18437120 + }, + { + "name": "transformer.h.14.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.14.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 28282880 + }, + { + "name": "transformer.h.15.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "transformer.h.15.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "transformer.h.15.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "2a22f9a3ede0a79ca7d2b10c5cf833c9" + }, + { + "dataPath": "params_shard_16.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.15.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.15.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.15.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.15.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.15.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.15.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.15.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.15.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 32778240 + }, + { + "name": "transformer.h.15.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 33187840 + } + ], + "md5sum": "2f5b886c037bed43b3df1d3a44516420" + }, + { + "dataPath": "params_shard_17.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.15.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.15.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.16.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.16.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.16.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 11069440 + }, + { + "name": "transformer.h.16.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.16.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 24197120 + }, + { + "name": "transformer.h.16.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25835520 + } + ], + "md5sum": "6020296d7ad3a8f0ccb6093409ce9876" + }, + { + "dataPath": "params_shard_18.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.16.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.16.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.16.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.16.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.16.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 18027520 + }, + { + "name": "transformer.h.16.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 18437120 + }, + { + "name": "transformer.h.16.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.16.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 28282880 + }, + { + "name": "transformer.h.17.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "transformer.h.17.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "transformer.h.17.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "a71078501aacec1834323b133c185144" + }, + { + "dataPath": "params_shard_19.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.17.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.17.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.17.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.17.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.17.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.17.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.17.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.17.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 32778240 + }, + { + "name": "transformer.h.17.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 33187840 + } + ], + "md5sum": "2981ebcbb002cdafa2c6e7d4d3d97029" + }, + { + "dataPath": "params_shard_20.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.17.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.17.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.18.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.18.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.18.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 11069440 + }, + { + "name": "transformer.h.18.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.18.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 24197120 + }, + { + "name": "transformer.h.18.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25835520 + } + ], + "md5sum": "fc72c2fec7ee1a6b3cf4ee634b70aa9f" + }, + { + "dataPath": "params_shard_21.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.18.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.18.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.18.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.18.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.18.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 18027520 + }, + { + "name": "transformer.h.18.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 18437120 + }, + { + "name": "transformer.h.18.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.18.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 28282880 + }, + { + "name": "transformer.h.19.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "transformer.h.19.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "transformer.h.19.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "3b3cc3921712a8792c74e9ad40c7e084" + }, + { + "dataPath": "params_shard_22.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.19.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.19.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.19.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.19.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.19.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.19.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.19.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.19.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 32778240 + }, + { + "name": "transformer.h.19.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 33187840 + } + ], + "md5sum": "f609bed9ef1b25c3bd7435f3c0314023" + }, + { + "dataPath": "params_shard_23.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.19.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.19.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.2.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.2.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.2.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 11069440 + }, + { + "name": "transformer.h.2.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.2.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 24197120 + }, + { + "name": "transformer.h.2.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25835520 + } + ], + "md5sum": "8ff6b6cded337c2fe73f33064841e9fd" + }, + { + "dataPath": "params_shard_24.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.2.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.2.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.2.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.2.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.2.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 18027520 + }, + { + "name": "transformer.h.2.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 18437120 + }, + { + "name": "transformer.h.2.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.2.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 28282880 + }, + { + "name": "transformer.h.20.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "transformer.h.20.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "transformer.h.20.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "21742135736ef3b9d6bc9d77641758d9" + }, + { + "dataPath": "params_shard_25.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.20.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.20.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.20.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.20.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.20.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.20.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.20.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.20.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 32778240 + }, + { + "name": "transformer.h.20.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 33187840 + } + ], + "md5sum": "8321a07cfb23d7deeba3942f5a957bd8" + }, + { + "dataPath": "params_shard_26.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.20.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.20.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.21.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.21.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.21.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 11069440 + }, + { + "name": "transformer.h.21.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.21.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 24197120 + }, + { + "name": "transformer.h.21.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25835520 + } + ], + "md5sum": "6b432ca0dc5cd1db54b75cf18e7f28de" + }, + { + "dataPath": "params_shard_27.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.21.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.21.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.21.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.21.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.21.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 18027520 + }, + { + "name": "transformer.h.21.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 18437120 + }, + { + "name": "transformer.h.21.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.21.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 28282880 + }, + { + "name": "transformer.h.22.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "transformer.h.22.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "transformer.h.22.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "4dbfe14e54e88a49084f3d3961cff5e4" + }, + { + "dataPath": "params_shard_28.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.22.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.22.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.22.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.22.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.22.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.22.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.22.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.22.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 32778240 + }, + { + "name": "transformer.h.22.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 33187840 + } + ], + "md5sum": "eaeca88a88516bd732e0d197b5dd97bc" + }, + { + "dataPath": "params_shard_29.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.22.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.22.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.23.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.23.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.23.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 11069440 + }, + { + "name": "transformer.h.23.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.23.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 24197120 + }, + { + "name": "transformer.h.23.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25835520 + } + ], + "md5sum": "094ea3730cb226ad5f57f7e6e8780ad6" + }, + { + "dataPath": "params_shard_30.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.23.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.23.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.23.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.23.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.23.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 18027520 + }, + { + "name": "transformer.h.23.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 18437120 + }, + { + "name": "transformer.h.23.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.23.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 28282880 + }, + { + "name": "transformer.h.24.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "transformer.h.24.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "transformer.h.24.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "c36b5db5e427911c3cb42ea63c9303b4" + }, + { + "dataPath": "params_shard_31.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.24.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.24.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.24.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.24.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.24.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.24.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.24.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.24.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 32778240 + }, + { + "name": "transformer.h.24.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 33187840 + } + ], + "md5sum": "ad136c0afde304c21d95fe7433552daf" + }, + { + "dataPath": "params_shard_32.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.24.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.24.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.25.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.25.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.25.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 11069440 + }, + { + "name": "transformer.h.25.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.25.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 24197120 + }, + { + "name": "transformer.h.25.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25835520 + } + ], + "md5sum": "7057c5b708813c0456474267429615ea" + }, + { + "dataPath": "params_shard_33.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.25.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.25.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.25.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.25.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.25.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 18027520 + }, + { + "name": "transformer.h.25.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 18437120 + }, + { + "name": "transformer.h.25.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.25.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 28282880 + }, + { + "name": "transformer.h.26.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "transformer.h.26.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "transformer.h.26.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "706934073d90c50faaa2ad4f94ab82c0" + }, + { + "dataPath": "params_shard_34.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.26.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.26.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.26.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.26.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.26.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.26.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.26.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.26.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 32778240 + }, + { + "name": "transformer.h.26.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 33187840 + } + ], + "md5sum": "7ea60991548cc508976bb7529e1eb481" + }, + { + "dataPath": "params_shard_35.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.26.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.26.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.27.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.27.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.27.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 11069440 + }, + { + "name": "transformer.h.27.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.27.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 24197120 + }, + { + "name": "transformer.h.27.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25835520 + } + ], + "md5sum": "c6b062d51a8d8774efea63b2f46c64e1" + }, + { + "dataPath": "params_shard_36.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.27.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.27.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.27.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.27.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.27.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 18027520 + }, + { + "name": "transformer.h.27.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 18437120 + }, + { + "name": "transformer.h.27.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.27.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 28282880 + }, + { + "name": "transformer.h.28.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "transformer.h.28.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "transformer.h.28.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "529d29f15e665ebb125b99da737c0efd" + }, + { + "dataPath": "params_shard_37.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.28.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.28.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.28.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.28.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.28.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.28.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.28.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.28.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 32778240 + }, + { + "name": "transformer.h.28.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 33187840 + } + ], + "md5sum": "98b7756758f80d630b5e9ccadcb7556c" + }, + { + "dataPath": "params_shard_38.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.28.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.28.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.29.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.29.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.29.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 11069440 + }, + { + "name": "transformer.h.29.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.29.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 24197120 + }, + { + "name": "transformer.h.29.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25835520 + } + ], + "md5sum": "ec47a916cc30f05f328ff81b4cde4e2c" + }, + { + "dataPath": "params_shard_39.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.29.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.29.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.29.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.29.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.29.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 18027520 + }, + { + "name": "transformer.h.29.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 18437120 + }, + { + "name": "transformer.h.29.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.29.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 28282880 + }, + { + "name": "transformer.h.3.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "transformer.h.3.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "transformer.h.3.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "16e5984ee0c0fba65cc10bd1f3922b56" + }, + { + "dataPath": "params_shard_40.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.3.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.3.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.3.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.3.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.3.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.3.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.3.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.3.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 32778240 + }, + { + "name": "transformer.h.3.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 33187840 + } + ], + "md5sum": "22df66cb2f9dd146e877fff27c4fde17" + }, + { + "dataPath": "params_shard_41.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.3.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.3.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.4.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.4.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.4.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 11069440 + }, + { + "name": "transformer.h.4.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.4.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 24197120 + }, + { + "name": "transformer.h.4.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25835520 + } + ], + "md5sum": "d65383adadce43970614c4e231bbfed5" + }, + { + "dataPath": "params_shard_42.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.4.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.4.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.4.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.4.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.4.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 18027520 + }, + { + "name": "transformer.h.4.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 18437120 + }, + { + "name": "transformer.h.4.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.4.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 28282880 + }, + { + "name": "transformer.h.5.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "transformer.h.5.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "transformer.h.5.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "dc81f95d33b056bc723774aac9a4bdfd" + }, + { + "dataPath": "params_shard_43.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.5.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.5.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.5.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.5.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.5.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.5.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.5.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.5.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 32778240 + }, + { + "name": "transformer.h.5.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 33187840 + } + ], + "md5sum": "7b90ae0051917104cd953d6c9fa73634" + }, + { + "dataPath": "params_shard_44.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.5.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.5.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.6.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.6.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.6.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 11069440 + }, + { + "name": "transformer.h.6.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.6.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 24197120 + }, + { + "name": "transformer.h.6.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25835520 + } + ], + "md5sum": "f9eeb07ae4c16bb0765cb412048a1d70" + }, + { + "dataPath": "params_shard_45.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.6.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.6.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.6.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.6.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.6.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 18027520 + }, + { + "name": "transformer.h.6.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 18437120 + }, + { + "name": "transformer.h.6.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.6.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 28282880 + }, + { + "name": "transformer.h.7.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "transformer.h.7.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "transformer.h.7.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "4a83fd5df047638a1d5ca8c8d387964f" + }, + { + "dataPath": "params_shard_46.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.7.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.7.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.7.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.7.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.7.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.7.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.7.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.7.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 32778240 + }, + { + "name": "transformer.h.7.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 33187840 + } + ], + "md5sum": "0ddda3b2d1fa8ef941479444bdcf24d0" + }, + { + "dataPath": "params_shard_47.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.7.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.7.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + }, + { + "name": "transformer.h.8.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11059200 + }, + { + "name": "transformer.h.8.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 11064320 + }, + { + "name": "transformer.h.8.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 11069440 + }, + { + "name": "transformer.h.8.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.8.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 24197120 + }, + { + "name": "transformer.h.8.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 25835520 + } + ], + "md5sum": "fc20ebedd6143ccb40a5f7db0e45dc6f" + }, + { + "dataPath": "params_shard_48.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.8.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.8.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.8.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.8.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.8.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 18027520 + }, + { + "name": "transformer.h.8.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 18437120 + }, + { + "name": "transformer.h.8.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.8.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 28282880 + }, + { + "name": "transformer.h.9.ln.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29511680 + }, + { + "name": "transformer.h.9.ln.weight", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29516800 + }, + { + "name": "transformer.h.9.mlp.fc1.bias", + "shape": [ + 10240 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 20480, + "byteOffset": 29521920 + } + ], + "md5sum": "443207fb4bb5d041c49d5ac99c71393b" + }, + { + "dataPath": "params_shard_49.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.9.mlp.fc1.q_weight", + "shape": [ + 10240, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.9.mlp.fc1.q_scale", + "shape": [ + 10240, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 13107200 + }, + { + "name": "transformer.h.9.mlp.fc2.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 14745600 + }, + { + "name": "transformer.h.9.mlp.fc2.q_weight", + "shape": [ + 2560, + 1280 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.9.mlp.fc2.q_scale", + "shape": [ + 2560, + 320 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1638400, + "byteOffset": 27857920 + }, + { + "name": "transformer.h.9.mixer.out_proj.bias", + "shape": [ + 2560 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 5120, + "byteOffset": 29496320 + }, + { + "name": "transformer.h.9.mixer.out_proj.q_weight", + "shape": [ + 2560, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.9.mixer.out_proj.q_scale", + "shape": [ + 2560, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 409600, + "byteOffset": 32778240 + }, + { + "name": "transformer.h.9.mixer.Wqkv.bias", + "shape": [ + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 15360, + "byteOffset": 33187840 + } + ], + "md5sum": "7ecdbd5a49a79fa6cae4a5a68322c0b8" + }, + { + "dataPath": "params_shard_50.bin", + "format": "raw-shard", + "nbytes": 11059200, + "records": [ + { + "name": "transformer.h.9.mixer.Wqkv.q_weight", + "shape": [ + 7680, + 320 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.9.mixer.Wqkv.q_scale", + "shape": [ + 7680, + 80 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + } + ], + "md5sum": "0cdf9c368eecb107d3127d1799aa1b10" + } + ] +} \ No newline at end of file