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": [ + 320, + 51200 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 65536000, + "byteOffset": 0 + } + ], + "md5sum": "a90a14eeed995852350db563d1cc3c8e" + }, + { + "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": [ + 80, + 51200 + ], + "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": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 8335360 + }, + { + "name": "transformer.h.30.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": "0c2f4106181323f4c1becf3f2e10e202" + }, + { + "dataPath": "params_shard_2.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.30.mlp.fc2.q_weight", + "shape": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.30.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.30.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.30.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": "e3dd026b18e921490b3e92662940ac04" + }, + { + "dataPath": "params_shard_3.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.31.mlp.fc1.q_weight", + "shape": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.31.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.31.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.31.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": "89ba5eb531a685f95e14ce123b97e6e0" + }, + { + "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": "c30425bd555bd4c0899a07df9f6784ad" + }, + { + "dataPath": "params_shard_5.bin", + "format": "raw-shard", + "nbytes": 32389120, + "records": [ + { + "name": "transformer.h.31.mixer.Wqkv.q_weight", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.31.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 19281920 + } + ], + "md5sum": "dc7512dcf68e82fd361d1329f44ffe0d" + }, + { + "dataPath": "params_shard_6.bin", + "format": "raw-shard", + "nbytes": 31185920, + "records": [ + { + "name": "transformer.h.0.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 1643520 + }, + { + "name": "transformer.h.0.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 16394240 + }, + { + "name": "transformer.h.0.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 20096000 + }, + { + "name": "transformer.h.0.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": "183a357d72428b0b30be42cbcb2e2244" + }, + { + "dataPath": "params_shard_7.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.1.mlp.fc1.q_weight", + "shape": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.1.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.1.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.1.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": "0eeff02e97c0e4b0d923ad520cdf70ef" + }, + { + "dataPath": "params_shard_8.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.1.mixer.Wqkv.q_weight", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.1.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.10.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": "732e6ed61235f0f949c1d429570466e3" + }, + { + "dataPath": "params_shard_9.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.10.mlp.fc2.q_weight", + "shape": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.10.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.10.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.10.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": "d798291cc9f79e0ac9d83d3f2f22547f" + }, + { + "dataPath": "params_shard_10.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.11.mlp.fc1.q_weight", + "shape": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.11.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.11.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.11.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": "240f746a11dbc5a652ddfaa8dfed90b4" + }, + { + "dataPath": "params_shard_11.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.11.mixer.Wqkv.q_weight", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.11.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.12.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": "d83e9c9f37502da9b2e8b21c747b2aea" + }, + { + "dataPath": "params_shard_12.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.12.mlp.fc2.q_weight", + "shape": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.12.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.12.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.12.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": "eb4c42d5eee3c2f9bac743ea82d8e23c" + }, + { + "dataPath": "params_shard_13.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.13.mlp.fc1.q_weight", + "shape": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.13.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.13.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.13.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": "4daf034ae4f4d26e721a9f87b67ff883" + }, + { + "dataPath": "params_shard_14.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.13.mixer.Wqkv.q_weight", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.13.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.14.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": "b46754c8b86788eefdb62c0921d003f4" + }, + { + "dataPath": "params_shard_15.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.14.mlp.fc2.q_weight", + "shape": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.14.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.14.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.14.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": "f4fcf6ca836b7769999afa6768048bad" + }, + { + "dataPath": "params_shard_16.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.15.mlp.fc1.q_weight", + "shape": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.15.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.15.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.15.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": "00ecc91b50d3fbf52d8d00ed3bdb10dd" + }, + { + "dataPath": "params_shard_17.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.15.mixer.Wqkv.q_weight", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.15.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.16.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": "530ff9207eeb6953fc4697c9f8363ac7" + }, + { + "dataPath": "params_shard_18.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.16.mlp.fc2.q_weight", + "shape": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.16.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.16.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.16.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": "72d578a33a0289e6d98eba7de22ae2b4" + }, + { + "dataPath": "params_shard_19.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.17.mlp.fc1.q_weight", + "shape": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.17.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.17.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.17.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": "263b272c21771f55cebca21bc0c22364" + }, + { + "dataPath": "params_shard_20.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.17.mixer.Wqkv.q_weight", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.17.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.18.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": "ad8c2e16af11e8b6e91ecc06517ad68e" + }, + { + "dataPath": "params_shard_21.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.18.mlp.fc2.q_weight", + "shape": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.18.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.18.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.18.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": "e8c5066e23a2040c61af6e319bfa3c26" + }, + { + "dataPath": "params_shard_22.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.19.mlp.fc1.q_weight", + "shape": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.19.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.19.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.19.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": "2ee327c66751552e1dd2d05431fb1df8" + }, + { + "dataPath": "params_shard_23.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.19.mixer.Wqkv.q_weight", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.19.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.2.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": "17ddc36d2f53bb428d5786a660f477f8" + }, + { + "dataPath": "params_shard_24.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.2.mlp.fc2.q_weight", + "shape": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.2.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.2.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.2.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": "aa2dcbed05c1949b78d09443725850ae" + }, + { + "dataPath": "params_shard_25.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.20.mlp.fc1.q_weight", + "shape": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.20.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.20.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.20.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": "d418fe03a7983e4e09e70572facde505" + }, + { + "dataPath": "params_shard_26.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.20.mixer.Wqkv.q_weight", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.20.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.21.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": "ead95b8b17349dec6347dc96397d4c1b" + }, + { + "dataPath": "params_shard_27.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.21.mlp.fc2.q_weight", + "shape": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.21.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.21.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.21.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": "977d00cd9bc6c6e7c8e09027accdddc2" + }, + { + "dataPath": "params_shard_28.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.22.mlp.fc1.q_weight", + "shape": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.22.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.22.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.22.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": "ef567712237115e8d51b81d8e2fe90df" + }, + { + "dataPath": "params_shard_29.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.22.mixer.Wqkv.q_weight", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.22.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.23.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": "b4390127af984eccfb9f95c55092a6bb" + }, + { + "dataPath": "params_shard_30.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.23.mlp.fc2.q_weight", + "shape": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.23.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.23.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.23.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": "dd7ea87d7b5e367c302875cbf5057165" + }, + { + "dataPath": "params_shard_31.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.24.mlp.fc1.q_weight", + "shape": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.24.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.24.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.24.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": "771bba4598b6c39e94b958ca6188ac2d" + }, + { + "dataPath": "params_shard_32.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.24.mixer.Wqkv.q_weight", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.24.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.25.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": "25c0fe9adb0852022367fba0cc7c3fbd" + }, + { + "dataPath": "params_shard_33.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.25.mlp.fc2.q_weight", + "shape": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.25.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.25.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.25.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": "373a63ede28ed0035269a75b0da6c1d8" + }, + { + "dataPath": "params_shard_34.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.26.mlp.fc1.q_weight", + "shape": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.26.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.26.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.26.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": "19b6688bc1774ace5fd59c3e66fc5c5e" + }, + { + "dataPath": "params_shard_35.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.26.mixer.Wqkv.q_weight", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.26.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.27.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": "c5c02c821a00cf4a19696eae035cf5c5" + }, + { + "dataPath": "params_shard_36.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.27.mlp.fc2.q_weight", + "shape": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.27.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.27.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.27.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": "bc313d99e2b13e3b93286393bb8ce11d" + }, + { + "dataPath": "params_shard_37.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.28.mlp.fc1.q_weight", + "shape": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.28.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.28.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.28.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": "51766f9d5eeb56da866b5941bc8d82aa" + }, + { + "dataPath": "params_shard_38.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.28.mixer.Wqkv.q_weight", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.28.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.29.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": "57863bb6165dc6e8adc620794d0a2b03" + }, + { + "dataPath": "params_shard_39.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.29.mlp.fc2.q_weight", + "shape": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.29.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.29.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.29.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": "e10f78449bfc7c2a6b35da2b4a742496" + }, + { + "dataPath": "params_shard_40.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.3.mlp.fc1.q_weight", + "shape": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.3.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.3.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.3.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": "6ddfefec81906ed568be8ba6d1f2c8a0" + }, + { + "dataPath": "params_shard_41.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.3.mixer.Wqkv.q_weight", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.3.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.4.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": "0545d74cd03b225843f782cb9a6d107a" + }, + { + "dataPath": "params_shard_42.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.4.mlp.fc2.q_weight", + "shape": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.4.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.4.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.4.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": "52de41fa048453b7b178e7fd7942e437" + }, + { + "dataPath": "params_shard_43.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.5.mlp.fc1.q_weight", + "shape": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.5.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.5.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.5.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": "931057f410e882c8493b2087a2245139" + }, + { + "dataPath": "params_shard_44.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.5.mixer.Wqkv.q_weight", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.5.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.6.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": "62702d886aa93eeaada17f63bc441b4b" + }, + { + "dataPath": "params_shard_45.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.6.mlp.fc2.q_weight", + "shape": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.6.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.6.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.6.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": "69129fca8d5c2775055bf0e69083fa47" + }, + { + "dataPath": "params_shard_46.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.7.mlp.fc1.q_weight", + "shape": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.7.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.7.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.7.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": "02d8fc127126107eb0d3ffb4c74a20be" + }, + { + "dataPath": "params_shard_47.bin", + "format": "raw-shard", + "nbytes": 25840640, + "records": [ + { + "name": "transformer.h.7.mixer.Wqkv.q_weight", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.7.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 11089920 + }, + { + "name": "transformer.h.8.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": "74646e0bbe35017a2dd158413de5f5f7" + }, + { + "dataPath": "params_shard_48.bin", + "format": "raw-shard", + "nbytes": 29542400, + "records": [ + { + "name": "transformer.h.8.mlp.fc2.q_weight", + "shape": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.8.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.8.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 18452480 + }, + { + "name": "transformer.h.8.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "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": "ccb161a260fa51db70c5ba78e66b5f2e" + }, + { + "dataPath": "params_shard_49.bin", + "format": "raw-shard", + "nbytes": 33203200, + "records": [ + { + "name": "transformer.h.9.mlp.fc1.q_weight", + "shape": [ + 320, + 10240 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 0 + }, + { + "name": "transformer.h.9.mlp.fc1.q_scale", + "shape": [ + 80, + 10240 + ], + "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": [ + 1280, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 13107200, + "byteOffset": 14750720 + }, + { + "name": "transformer.h.9.mlp.fc2.q_scale", + "shape": [ + 320, + 2560 + ], + "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": [ + 320, + 2560 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 3276800, + "byteOffset": 29501440 + }, + { + "name": "transformer.h.9.mixer.out_proj.q_scale", + "shape": [ + 80, + 2560 + ], + "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": "680735df818c845e780c25c2053c2d52" + }, + { + "dataPath": "params_shard_50.bin", + "format": "raw-shard", + "nbytes": 11059200, + "records": [ + { + "name": "transformer.h.9.mixer.Wqkv.q_weight", + "shape": [ + 320, + 7680 + ], + "dtype": "uint32", + "format": "f32-to-bf16", + "nbytes": 9830400, + "byteOffset": 0 + }, + { + "name": "transformer.h.9.mixer.Wqkv.q_scale", + "shape": [ + 80, + 7680 + ], + "dtype": "float16", + "format": "f32-to-bf16", + "nbytes": 1228800, + "byteOffset": 9830400 + } + ], + "md5sum": "81d18cca7cef105286560633556c8f5c" + } + ] +} \ No newline at end of file