{ "metadata": { "ParamSize": 869, "ParamBytes": 1873364948.0, "BitsPerParam": 3.7021917590823534 }, "records": [ { "dataPath": "params_shard_0.bin", "format": "raw-shard", "nbytes": 39502848, "records": [ { "name": "lm_head.q_weight", "shape": [ 32064, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 39502848, "byteOffset": 0 } ], "md5sum": "8a35ddbbc2ab05256c6c2d54d6094fda" }, { "dataPath": "params_shard_1.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.18.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "5e643ec4c20f60ffaeae2a841566cd75" }, { "dataPath": "params_shard_2.bin", "format": "raw-shard", "nbytes": 31588480, "records": [ { "name": "lm_head.q_scale", "shape": [ 32064, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 4937856, "byteOffset": 0 }, { "name": "model.h.18.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 4937856 }, { "name": "model.h.18.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 4944000 }, { "name": "model.h.18.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 15020160 }, { "name": "model.h.18.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 16279680 }, { "name": "model.h.18.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 18802816 }, { "name": "model.h.18.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 18808960 }, { "name": "model.h.18.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 30163072 }, { "name": "model.h.19.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 31582336 } ], "md5sum": "24fab578168aeadbc9559a578c7183e7" }, { "dataPath": "params_shard_3.bin", "format": "raw-shard", "nbytes": 31520768, "records": [ { "name": "model.h.19.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 0 }, { "name": "model.h.19.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 10076160 }, { "name": "model.h.19.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 11335680 } ], "md5sum": "419b98c8f9ad7d166c3979d8d9c7559b" }, { "dataPath": "params_shard_4.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.20.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "1664118860d52dd98f030f503f04ad24" }, { "dataPath": "params_shard_5.bin", "format": "raw-shard", "nbytes": 33431552, "records": [ { "name": "model.h.19.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 0 }, { "name": "model.h.19.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 2523136 }, { "name": "model.h.19.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 2529280 }, { "name": "model.h.19.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 6313984 }, { "name": "model.h.19.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 6787072 }, { "name": "model.h.19.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 18141184 }, { "name": "model.h.20.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 19560448 }, { "name": "model.h.20.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 19566592 }, { "name": "model.h.20.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 29642752 }, { "name": "model.h.20.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 30902272 }, { "name": "model.h.20.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 33425408 } ], "md5sum": "c2a46d96f42d25691193feaf4dfdb7f2" }, { "dataPath": "params_shard_6.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.21.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "5a0644c2a7b96ecc7f44db61c76980b1" }, { "dataPath": "params_shard_7.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.20.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.20.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.20.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.20.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.21.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.21.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.21.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.21.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.21.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "e207ed95b32cf79cd2023cf0ec11d8c7" }, { "dataPath": "params_shard_8.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.22.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "f4302a88dfb0c1dd8f39023f158eecb1" }, { "dataPath": "params_shard_9.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.21.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.21.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.21.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.21.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.22.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.22.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.22.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.22.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.22.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "526c77bf2a28ad9f696f99d2f36324b5" }, { "dataPath": "params_shard_10.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.23.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "7d502dc51f0b4ba087038840e2e3fa2a" }, { "dataPath": "params_shard_11.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.22.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.22.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.22.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.22.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.23.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.23.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.23.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.23.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.23.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "40b08c6e4a04c85dcc8d20763a131aaa" }, { "dataPath": "params_shard_12.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.24.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "ad0c16081945fd54ca9d5bf88322cf6e" }, { "dataPath": "params_shard_13.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.23.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.23.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.23.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.23.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.24.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.24.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.24.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.24.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.24.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "ff97580cc8b519c9a0dafbec95ce9076" }, { "dataPath": "params_shard_14.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.25.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "900c18766113eff094153343f44eae68" }, { "dataPath": "params_shard_15.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.24.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.24.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.24.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.24.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.25.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.25.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.25.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.25.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.25.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "354d4710d313309d24b60d9dc0d7de86" }, { "dataPath": "params_shard_16.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.26.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "375e40a711c06214bcf155207c27c8df" }, { "dataPath": "params_shard_17.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.25.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.25.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.25.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.25.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.26.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.26.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.26.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.26.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.26.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "2a08f7bdb8837cf2659d8d2d62b47cfb" }, { "dataPath": "params_shard_18.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.27.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "02e503ff2aa0900aa47e73d0b85ef80d" }, { "dataPath": "params_shard_19.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.26.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.26.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.26.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.26.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.27.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.27.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.27.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.27.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.27.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "089a8975a68756edba36ba294c67a8ef" }, { "dataPath": "params_shard_20.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.28.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "e55ee3c00cc9b2fc5db22c815ed1e35f" }, { "dataPath": "params_shard_21.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.27.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.27.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.27.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.27.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.28.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.28.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.28.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.28.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.28.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "f3478b8fbcd86f16cec0803918d5e3ec" }, { "dataPath": "params_shard_22.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.29.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "2d47274575a56cc29842d9f9413d801c" }, { "dataPath": "params_shard_23.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.28.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.28.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.28.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.28.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.29.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.29.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.29.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.29.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.29.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "6871519612a4b73c0d24e2b32bcfb785" }, { "dataPath": "params_shard_24.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.30.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "b90a3c96a5a1207a3574ae02a48669ef" }, { "dataPath": "params_shard_25.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.29.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.29.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.29.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.29.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.30.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.30.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.30.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.30.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.30.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "bd8f516641f37b8ab8fdfbf5341150d7" }, { "dataPath": "params_shard_26.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.31.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "5cad5fb8c628578200af5174a42c330d" }, { "dataPath": "params_shard_27.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.30.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.30.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.30.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.30.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.31.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.31.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.31.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.31.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.31.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "3aabdc1f95147e394d5de5df9697ef2b" }, { "dataPath": "params_shard_28.bin", "format": "raw-shard", "nbytes": 39502848, "records": [ { "name": "model.embd.q_weight", "shape": [ 32064, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 39502848, "byteOffset": 0 } ], "md5sum": "381023fb23574049257f2a0eec9b554f" }, { "dataPath": "params_shard_29.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.0.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "b05a8053cb19739cf76c4d5e303b95ed" }, { "dataPath": "params_shard_30.bin", "format": "raw-shard", "nbytes": 33316992, "records": [ { "name": "model.h.31.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.31.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.31.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.31.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.norm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.embd.q_scale", "shape": [ 32064, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 4937856, "byteOffset": 17037312 }, { "name": "model.h.0.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 21975168 }, { "name": "model.h.0.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 21981312 }, { "name": "model.h.0.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 32057472 } ], "md5sum": "55427a0d100d982af219140f2f1b3b19" }, { "dataPath": "params_shard_31.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.1.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "655b2041eb3bc532b2f76f8a3044a80e" }, { "dataPath": "params_shard_32.bin", "format": "raw-shard", "nbytes": 33431552, "records": [ { "name": "model.h.0.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 0 }, { "name": "model.h.0.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 2523136 }, { "name": "model.h.0.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 2529280 }, { "name": "model.h.0.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 6313984 }, { "name": "model.h.0.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 6787072 }, { "name": "model.h.0.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 18141184 }, { "name": "model.h.1.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 19560448 }, { "name": "model.h.1.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 19566592 }, { "name": "model.h.1.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 29642752 }, { "name": "model.h.1.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 30902272 }, { "name": "model.h.1.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 33425408 } ], "md5sum": "60597c76e7ae98d8687f3ef6d24ef84d" }, { "dataPath": "params_shard_33.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.10.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "3beb4e5859d4d77a395a9cfcda94fd21" }, { "dataPath": "params_shard_34.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.1.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.1.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.1.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.1.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.10.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.10.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.10.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.10.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.10.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "a5a2e49eb5e992f4a4da8593a0e32c46" }, { "dataPath": "params_shard_35.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.11.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "a5ee83fb741c7a33c7e86784d9be28d6" }, { "dataPath": "params_shard_36.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.10.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.10.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.10.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.10.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.11.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.11.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.11.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.11.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.11.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "8b09bdf75eba6d273336c78223a3fbac" }, { "dataPath": "params_shard_37.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.12.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "a3b6820d889f86491595e7c64a292346" }, { "dataPath": "params_shard_38.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.11.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.11.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.11.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.11.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.12.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.12.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.12.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.12.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.12.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "2cf21a5ab798e9a8424ba6c8d4475414" }, { "dataPath": "params_shard_39.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.13.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "f4ab43131aa9a0ffee3815b901dc8591" }, { "dataPath": "params_shard_40.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.12.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.12.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.12.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.12.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.13.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.13.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.13.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.13.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.13.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "c0cb92677fedc090b87887989db52c5b" }, { "dataPath": "params_shard_41.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.14.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "0756b15ae5e75580f8a9f33f62914dda" }, { "dataPath": "params_shard_42.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.13.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.13.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.13.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.13.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.14.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.14.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.14.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.14.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.14.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "f7ec2b75978886fca92ba13a6a511bf0" }, { "dataPath": "params_shard_43.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.15.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "67c0eba6afa7bead6674cbc479f6b5a7" }, { "dataPath": "params_shard_44.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.14.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.14.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.14.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.14.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.15.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.15.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.15.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.15.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.15.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "f67124710df92edb87263cc95ca70806" }, { "dataPath": "params_shard_45.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.16.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "db05762038bee40f53680fbf656b91da" }, { "dataPath": "params_shard_46.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.15.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.15.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.15.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.15.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.16.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.16.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.16.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.16.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.16.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "939795fea6dc100035f9fc0a49d711b0" }, { "dataPath": "params_shard_47.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.17.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "f9fb91d66e9fca6030f1ae038d840f00" }, { "dataPath": "params_shard_48.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.16.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.16.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.16.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.16.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.17.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.17.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.17.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.17.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.17.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "8ab8d33f317b9a6c5c60140793ff0fd3" }, { "dataPath": "params_shard_49.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.2.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "d02c9a6a3e469a42b7727dcbe8da5372" }, { "dataPath": "params_shard_50.bin", "format": "raw-shard", "nbytes": 32630784, "records": [ { "name": "model.h.17.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.17.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.17.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.17.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.18.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 17031168 }, { "name": "model.h.18.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 20815872 }, { "name": "model.h.2.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 21288960 }, { "name": "model.h.2.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 21295104 }, { "name": "model.h.2.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 31371264 } ], "md5sum": "34bb78bf6f0142eae8fc499be2eae4a1" }, { "dataPath": "params_shard_51.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.3.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "1b55a46a61b4c4c44d98239d787cf521" }, { "dataPath": "params_shard_52.bin", "format": "raw-shard", "nbytes": 33431552, "records": [ { "name": "model.h.2.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 0 }, { "name": "model.h.2.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 2523136 }, { "name": "model.h.2.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 2529280 }, { "name": "model.h.2.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 6313984 }, { "name": "model.h.2.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 6787072 }, { "name": "model.h.2.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 18141184 }, { "name": "model.h.3.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 19560448 }, { "name": "model.h.3.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 19566592 }, { "name": "model.h.3.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 29642752 }, { "name": "model.h.3.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 30902272 }, { "name": "model.h.3.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 33425408 } ], "md5sum": "8cdb1371166272e1a9084901fc4be2f2" }, { "dataPath": "params_shard_53.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.4.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "8d9bb581a610c0f8253fb86edb63b3e7" }, { "dataPath": "params_shard_54.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.3.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.3.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.3.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.3.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.4.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.4.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.4.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.4.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.4.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "35b6c52919b02d23fcc4eb6c96fa3939" }, { "dataPath": "params_shard_55.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.5.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "4fdd2e740eb529f99aa42517aff952e1" }, { "dataPath": "params_shard_56.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.4.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.4.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.4.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.4.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.5.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.5.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.5.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.5.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.5.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "34d8e5a154849d6cd0e80c3cbfeb415a" }, { "dataPath": "params_shard_57.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.6.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "5424057c52fefadd949acaf659270b4f" }, { "dataPath": "params_shard_58.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.5.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.5.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.5.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.5.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.6.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.6.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.6.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.6.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.6.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "d9a4476ee91dcd2b8b4e1fc21e62e3ae" }, { "dataPath": "params_shard_59.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.7.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "3c7dfad2b08f7f7286e527bb245d0c22" }, { "dataPath": "params_shard_60.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.6.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.6.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.6.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.6.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.7.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.7.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.7.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.7.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.7.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "021ced36e57a5c9c9c4799aaa95aa1b3" }, { "dataPath": "params_shard_61.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.8.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "15b1fe414b809a835331cb0688bd903f" }, { "dataPath": "params_shard_62.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.7.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.7.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.7.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.7.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.8.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.8.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.8.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.8.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.8.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "7b85b21a4f10e21933aa76f10079d7c6" }, { "dataPath": "params_shard_63.bin", "format": "raw-shard", "nbytes": 20185088, "records": [ { "name": "model.h.9.mlp.gate_up_proj.q_weight", "shape": [ 16384, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 20185088, "byteOffset": 0 } ], "md5sum": "80ffcfa1ab4cb04357f75e1fcd6d2018" }, { "dataPath": "params_shard_64.bin", "format": "raw-shard", "nbytes": 30902272, "records": [ { "name": "model.h.8.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.8.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.8.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.8.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "model.h.9.ln.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 17031168 }, { "name": "model.h.9.mlp.down_proj.q_weight", "shape": [ 3072, 820 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 10076160, "byteOffset": 17037312 }, { "name": "model.h.9.mlp.down_proj.q_scale", "shape": [ 3072, 205 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1259520, "byteOffset": 27113472 }, { "name": "model.h.9.mlp.gate_up_proj.q_scale", "shape": [ 16384, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2523136, "byteOffset": 28372992 }, { "name": "model.h.9.post_attention_layernorm.weight", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30896128 } ], "md5sum": "1a6d6b9b372685fc903858f709c330d3" }, { "dataPath": "params_shard_65.bin", "format": "raw-shard", "nbytes": 31968980, "records": [ { "name": "model.h.9.mixer.out_proj.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "model.h.9.mixer.out_proj.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "model.h.9.mixer.qkv_proj.q_weight", "shape": [ 9216, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 11354112, "byteOffset": 4257792 }, { "name": "model.h.9.mixer.qkv_proj.q_scale", "shape": [ 9216, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1419264, "byteOffset": 15611904 }, { "name": "vision_embed_tokens.glb_GN", "shape": [ 1, 1, 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 17031168 }, { "name": "vision_embed_tokens.img_processor.vision_model.embeddings.class_embedding", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 17039360 }, { "name": "vision_embed_tokens.img_processor.vision_model.embeddings.patch_embedding.weight", "shape": [ 1024, 3, 14, 14 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 1204224, "byteOffset": 17041408 }, { "name": "vision_embed_tokens.img_processor.vision_model.embeddings.position_embedding.q_weight", "shape": [ 577, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 240032, "byteOffset": 18245632 }, { "name": "vision_embed_tokens.img_processor.vision_model.embeddings.position_embedding.q_scale", "shape": [ 577, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 30004, "byteOffset": 18485664 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 18515668 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 18517716 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 18519764 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 18521812 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 18523860 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 18532052 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 20235988 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 20448980 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 20451028 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 22138580 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 22349524 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 22351572 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 22777556 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 22830804 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 22832852 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 23258836 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 23312084 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 23314132 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 23740116 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 23793364 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 23795412 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 24221396 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 24274644 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 24276692 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 24278740 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 24280788 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 24282836 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 24291028 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 25994964 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 26207956 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 26210004 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 27897556 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 28108500 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 28110548 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 28536532 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 28589780 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 28591828 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 29017812 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 29071060 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 29073108 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 29499092 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 29552340 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 29554388 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 29980372 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 30033620 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 30035668 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 30037716 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 30039764 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 30041812 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 30050004 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 31753940 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 31966932 } ], "md5sum": "50ec01b6812197d7e6a45c6f47b0dfe5" }, { "dataPath": "params_shard_66.bin", "format": "raw-shard", "nbytes": 32634880, "records": [ { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 1687552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 1898496 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 1900544 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 2326528 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 2379776 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 2381824 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 2807808 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 2861056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 2863104 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 3289088 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 3342336 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 3344384 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 3770368 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 3823616 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 3825664 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 3827712 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 3829760 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 3831808 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 3840000 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 5543936 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 5756928 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 5758976 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 7446528 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 7657472 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 7659520 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 8085504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 8138752 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 8140800 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 8566784 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 8620032 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 8622080 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 9048064 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 9101312 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 9103360 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 9529344 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 9582592 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 9584640 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 9586688 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 9588736 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 9590784 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 9598976 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 11302912 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 11515904 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 11517952 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 13205504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 13416448 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 13418496 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 13844480 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 13897728 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 13899776 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 14325760 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 14379008 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 14381056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 14807040 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 14860288 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 14862336 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 15288320 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 15341568 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 15343616 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 15345664 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 15347712 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 15349760 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 15357952 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 17061888 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 17274880 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 17276928 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 18964480 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 19175424 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 19177472 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 19603456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 19656704 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 19658752 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 20084736 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 20137984 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 20140032 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 20566016 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 20619264 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 20621312 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 21047296 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 21100544 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 21102592 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 21104640 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 21106688 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 21108736 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 21116928 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 22820864 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 23033856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 23035904 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 24723456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 24934400 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 24936448 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 25362432 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 25415680 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 25417728 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 25843712 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 25896960 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 25899008 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 26324992 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 26378240 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 26380288 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 26806272 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 26859520 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 26861568 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 26863616 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 26865664 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 26867712 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 26875904 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 28579840 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 28792832 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 28794880 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 30482432 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 30693376 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 30695424 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 31121408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 31174656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 31176704 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 31602688 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 31655936 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 31657984 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 32083968 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 32137216 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 32139264 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 32565248 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 32618496 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 32620544 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 32622592 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 32624640 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 32626688 } ], "md5sum": "81c15ab7a906dcc39b7bf83864d4b44f" }, { "dataPath": "params_shard_67.bin", "format": "raw-shard", "nbytes": 33521664, "records": [ { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 1703936 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 1916928 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 1918976 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 3606528 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 3817472 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 3819520 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 4245504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 4298752 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 4300800 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 4726784 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 4780032 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 4782080 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 5208064 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 5261312 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 5263360 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 5689344 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 5742592 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 5744640 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 5746688 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 5748736 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 5750784 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 5758976 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 7462912 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 7675904 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 7677952 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 9365504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 9576448 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 9578496 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 10004480 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 10057728 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 10059776 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 10485760 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 10539008 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 10541056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 10967040 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 11020288 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 11022336 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 11448320 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 11501568 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 11503616 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 11505664 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 11507712 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 11509760 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 11517952 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 13221888 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 13434880 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 13436928 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 15124480 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 15335424 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 15337472 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 15763456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 15816704 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 15818752 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 16244736 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 16297984 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 16300032 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 16726016 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 16779264 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 16781312 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 17207296 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 17260544 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 17262592 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 17264640 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 17266688 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 17268736 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 17276928 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 18980864 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 19193856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 19195904 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 20883456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 21094400 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 21096448 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 21522432 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 21575680 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 21577728 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 22003712 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 22056960 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 22059008 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 22484992 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 22538240 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 22540288 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 22966272 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 23019520 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 23021568 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 23023616 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 23025664 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 23027712 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 23035904 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 24739840 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 24952832 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 24954880 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 26642432 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 26853376 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 26855424 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 27281408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 27334656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 27336704 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 27762688 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 27815936 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 27817984 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 28243968 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 28297216 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 28299264 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 28725248 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 28778496 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 28780544 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 28782592 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 28784640 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 28786688 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 28794880 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 30498816 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 30711808 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 30713856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 32401408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 32612352 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 32614400 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 33040384 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 33093632 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 33095680 } ], "md5sum": "c24f2e57d8e79878e359cf6ab34d1bee" }, { "dataPath": "params_shard_68.bin", "format": "raw-shard", "nbytes": 33433600, "records": [ { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 53248 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 55296 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 481280 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 534528 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 536576 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 962560 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 1015808 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 1017856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 1019904 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 1021952 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 1024000 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 1032192 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 2736128 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 2949120 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 2951168 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 4638720 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 4849664 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 4851712 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 5277696 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 5330944 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 5332992 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 5758976 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 5812224 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 5814272 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 6240256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 6293504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 6295552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 6721536 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 6774784 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 6776832 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 6778880 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 6780928 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 6782976 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 6791168 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 8495104 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 8708096 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 8710144 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 10397696 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 10608640 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 10610688 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 11036672 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 11089920 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 11091968 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 11517952 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 11571200 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 11573248 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 11999232 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 12052480 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 12054528 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 12480512 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 12533760 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 12535808 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 12537856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 12539904 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 12541952 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 12550144 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 14254080 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 14467072 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 14469120 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 16156672 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 16367616 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 16369664 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 16795648 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 16848896 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 16850944 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 17276928 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 17330176 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 17332224 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 17758208 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 17811456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 17813504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 18239488 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 18292736 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 18294784 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 18296832 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 18298880 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 18300928 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 18309120 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 20013056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 20226048 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 20228096 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 21915648 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 22126592 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 22128640 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 22554624 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 22607872 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 22609920 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 23035904 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 23089152 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 23091200 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 23517184 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 23570432 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 23572480 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 23998464 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 24051712 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 24053760 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 24055808 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 24057856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 24059904 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 24068096 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 25772032 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 25985024 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 25987072 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 27674624 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 27885568 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 27887616 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 28313600 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 28366848 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 28368896 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 28794880 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 28848128 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 28850176 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 29276160 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 29329408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 29331456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 29757440 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 29810688 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 29812736 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 29814784 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 29816832 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 29818880 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 29827072 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 31531008 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 31744000 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 31746048 } ], "md5sum": "a2dadce37a87f77e5bc03f6942019c7b" }, { "dataPath": "params_shard_69.bin", "format": "raw-shard", "nbytes": 30887936, "records": [ { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 210944 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 212992 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 638976 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 692224 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 694272 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 1120256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 1173504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 1175552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 1601536 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 1654784 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 1656832 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 2082816 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 2136064 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 2138112 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 2140160 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 2142208 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 2144256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 2152448 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 3856384 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 4069376 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 4071424 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 5758976 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 5969920 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 5971968 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 6397952 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 6451200 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 6453248 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 6879232 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 6932480 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 6934528 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 7360512 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 7413760 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 7415808 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 7841792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 7895040 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 7897088 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 7899136 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 7901184 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 7903232 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 7911424 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 9615360 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 9828352 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 9830400 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 11517952 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 11728896 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 11730944 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 12156928 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 12210176 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 12212224 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 12638208 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 12691456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 12693504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 13119488 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 13172736 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 13174784 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 13600768 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 13654016 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 13656064 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 13658112 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 13660160 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 13662208 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 13670400 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 15374336 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 15587328 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 15589376 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 17276928 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 17487872 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 17489920 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 17915904 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 17969152 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 17971200 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 18397184 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 18450432 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 18452480 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 18878464 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 18931712 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 18933760 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 19359744 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.layer_norm1.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 19412992 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.layer_norm1.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 19415040 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.layer_norm2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 19417088 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.layer_norm2.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 19419136 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.mlp.fc1.bias", "shape": [ 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 19421184 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.mlp.fc1.q_weight", "shape": [ 4096, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1703936, "byteOffset": 19429376 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.mlp.fc1.q_scale", "shape": [ 4096, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 212992, "byteOffset": 21133312 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.mlp.fc2.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 21346304 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.mlp.fc2.q_weight", "shape": [ 1024, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 1687552, "byteOffset": 21348352 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.mlp.fc2.q_scale", "shape": [ 1024, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 210944, "byteOffset": 23035904 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.k_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 23246848 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.k_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 23248896 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.k_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 23674880 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.out_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 23728128 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.out_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 23730176 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.out_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 24156160 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.q_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 24209408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.q_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 24211456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.q_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 24637440 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.v_proj.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 24690688 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.v_proj.q_weight", "shape": [ 1024, 104 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 425984, "byteOffset": 24692736 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.v_proj.q_scale", "shape": [ 1024, 26 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 53248, "byteOffset": 25118720 }, { "name": "vision_embed_tokens.img_processor.vision_model.post_layernorm.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 25171968 }, { "name": "vision_embed_tokens.img_processor.vision_model.post_layernorm.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 25174016 }, { "name": "vision_embed_tokens.img_processor.vision_model.pre_layrnorm.bias", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 25176064 }, { "name": "vision_embed_tokens.img_processor.vision_model.pre_layrnorm.weight", "shape": [ 1024 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 2048, "byteOffset": 25178112 }, { "name": "vision_embed_tokens.img_projection.linear_1.bias", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 25180160 }, { "name": "vision_embed_tokens.img_projection.linear_1.q_weight", "shape": [ 3072, 412 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 5062656, "byteOffset": 25186304 }, { "name": "vision_embed_tokens.img_projection.linear_1.q_scale", "shape": [ 3072, 103 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 632832, "byteOffset": 30248960 }, { "name": "vision_embed_tokens.img_projection.linear_2.bias", "shape": [ 3072 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 6144, "byteOffset": 30881792 } ], "md5sum": "96daef3567f9842251bb425924ac75a3" }, { "dataPath": "params_shard_70.bin", "format": "raw-shard", "nbytes": 4265984, "records": [ { "name": "vision_embed_tokens.img_projection.linear_2.q_weight", "shape": [ 3072, 308 ], "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 3784704, "byteOffset": 0 }, { "name": "vision_embed_tokens.img_projection.linear_2.q_scale", "shape": [ 3072, 77 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 473088, "byteOffset": 3784704 }, { "name": "vision_embed_tokens.sub_GN", "shape": [ 1, 1, 1, 4096 ], "dtype": "float16", "format": "f32-to-bf16", "nbytes": 8192, "byteOffset": 4257792 } ], "md5sum": "ebe2947681e943d49c3e4fa08218992a" } ] }