{"name": "layers.3.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00005-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.3.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00005-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.3.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00005-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.3.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00005-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.3.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00005-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.3.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00005-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.3.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00005-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.3.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00005-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.3.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00005-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.3.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00005-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.3.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00005-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.3.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00005-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.3.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00005-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.3.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00005-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.3.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00005-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.3.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00005-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.3.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00005-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.3.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00005-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.3.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00005-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.3.hc_attn_base", "shape": [24], "pack_file": "misc_model-00005-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.3.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00005-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.3.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00005-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.3.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00005-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.3.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00005-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.3.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00005-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "embed.weight", "shape": [129280, 4096], "pack_file": "misc_model-00001-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [1059061760], "nbytes": 1059061760}
{"name": "layers.2.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00004-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.2.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00004-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.2.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00004-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.2.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00004-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.2.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00004-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.2.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00004-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.2.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00004-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.2.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00004-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.2.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00004-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.2.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00004-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.2.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00004-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.2.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00004-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.2.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00004-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.2.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00004-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.2.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00004-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.2.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00004-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.2.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00004-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.2.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00004-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.2.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00004-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.2.ffn.gate.tid2eid", "shape": [129280, 6], "pack_file": "misc_model-00004-of-00046.pack", "offset": 136869120, "kind": "torch.int64", "sections": [6205440], "nbytes": 6205440}
{"name": "layers.2.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00004-of-00046.pack", "offset": 143074560, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.2.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00004-of-00046.pack", "offset": 145171712, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.2.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00004-of-00046.pack", "offset": 153560832, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.2.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00004-of-00046.pack", "offset": 161949952, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.2.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00004-of-00046.pack", "offset": 170339072, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.2.hc_attn_base", "shape": [24], "pack_file": "misc_model-00004-of-00046.pack", "offset": 170347264, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.2.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00004-of-00046.pack", "offset": 170347312, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.2.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00004-of-00046.pack", "offset": 171133744, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.2.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00004-of-00046.pack", "offset": 171133750, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.2.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00004-of-00046.pack", "offset": 171133798, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.2.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00004-of-00046.pack", "offset": 171920230, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.4.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00006-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.4.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00006-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.4.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00006-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.4.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00006-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.4.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00006-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.4.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00006-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.4.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00006-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.4.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00006-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.4.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00006-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.4.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00006-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.4.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00006-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.4.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00006-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.4.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00006-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.4.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00006-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.4.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00006-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.4.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00006-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.4.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00006-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.4.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00006-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.4.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00006-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.4.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00006-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.4.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00006-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.4.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00006-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.4.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00006-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.4.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00006-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.4.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00006-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.4.hc_attn_base", "shape": [24], "pack_file": "misc_model-00006-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.4.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00006-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.4.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00006-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.4.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00006-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.4.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00006-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.4.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00006-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.5.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00007-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.5.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00007-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.5.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00007-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.5.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00007-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.5.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00007-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.5.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00007-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.5.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00007-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.5.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00007-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.5.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00007-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.5.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00007-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.5.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00007-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.5.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00007-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.5.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00007-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.5.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00007-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.5.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00007-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.5.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00007-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.5.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00007-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.5.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00007-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.5.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00007-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.5.hc_attn_base", "shape": [24], "pack_file": "misc_model-00007-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.5.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00007-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.5.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00007-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.5.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00007-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.5.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00007-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.5.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00007-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.6.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00008-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.6.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00008-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.6.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00008-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.6.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00008-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.6.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00008-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.6.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00008-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.6.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00008-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.6.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00008-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.6.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00008-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.6.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00008-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.6.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00008-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.6.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00008-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.6.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00008-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.6.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00008-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.6.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00008-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.6.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00008-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.6.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00008-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.6.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00008-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.6.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00008-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.6.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00008-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.6.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00008-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.6.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00008-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.6.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00008-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.6.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00008-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.6.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00008-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.6.hc_attn_base", "shape": [24], "pack_file": "misc_model-00008-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.6.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00008-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.6.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00008-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.6.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00008-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.6.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00008-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.6.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00008-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.7.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00009-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.7.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00009-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.7.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00009-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.7.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00009-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.7.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00009-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.7.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00009-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.7.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00009-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.7.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00009-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.7.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00009-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.7.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00009-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.7.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00009-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.7.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00009-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.7.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00009-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.7.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00009-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.7.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00009-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.7.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00009-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.7.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00009-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.7.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00009-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.7.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00009-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.7.hc_attn_base", "shape": [24], "pack_file": "misc_model-00009-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.7.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00009-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.7.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00009-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.7.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00009-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.7.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00009-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.7.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00009-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.8.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00010-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.8.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00010-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.8.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00010-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.8.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00010-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.8.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00010-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.8.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00010-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.8.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00010-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.8.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00010-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.8.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00010-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.8.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00010-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.8.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00010-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.8.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00010-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.8.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00010-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.8.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00010-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.8.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00010-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.8.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00010-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.8.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00010-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.8.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00010-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.8.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00010-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.8.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00010-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.8.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00010-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.8.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00010-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.8.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00010-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.8.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00010-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.8.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00010-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.8.hc_attn_base", "shape": [24], "pack_file": "misc_model-00010-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.8.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00010-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.8.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00010-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.8.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00010-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.8.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00010-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.8.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00010-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.9.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00011-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.9.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00011-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.9.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00011-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.9.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00011-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.9.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00011-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.9.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00011-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.9.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00011-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.9.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00011-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.9.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00011-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.9.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00011-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.9.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00011-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.9.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00011-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.9.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00011-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.9.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00011-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.9.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00011-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.9.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00011-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.9.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00011-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.9.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00011-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.9.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00011-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.9.hc_attn_base", "shape": [24], "pack_file": "misc_model-00011-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.9.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00011-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.9.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00011-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.9.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00011-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.9.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00011-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.9.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00011-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.10.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00012-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.10.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00012-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.10.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00012-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.10.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00012-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.10.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00012-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.10.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00012-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.10.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00012-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.10.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00012-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.10.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00012-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.10.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00012-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.10.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00012-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.10.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00012-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.10.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00012-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.10.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00012-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.10.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00012-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.10.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00012-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.10.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00012-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.10.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00012-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.10.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00012-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.10.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00012-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.10.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00012-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.10.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00012-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.10.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00012-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.10.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00012-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.10.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00012-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.10.hc_attn_base", "shape": [24], "pack_file": "misc_model-00012-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.10.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00012-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.10.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00012-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.10.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00012-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.10.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00012-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.10.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00012-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.11.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00013-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.11.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00013-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.11.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00013-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.11.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00013-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.11.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00013-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.11.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00013-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.11.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00013-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.11.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00013-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.11.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00013-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.11.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00013-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.11.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00013-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.11.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00013-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.11.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00013-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.11.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00013-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.11.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00013-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.11.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00013-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.11.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00013-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.11.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00013-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.11.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00013-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.11.hc_attn_base", "shape": [24], "pack_file": "misc_model-00013-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.11.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00013-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.11.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00013-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.11.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00013-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.11.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00013-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.11.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00013-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.12.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00014-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.12.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00014-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.12.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00014-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.12.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00014-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.12.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00014-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.12.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00014-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.12.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00014-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.12.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00014-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.12.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00014-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.12.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00014-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.12.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00014-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.12.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00014-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.12.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00014-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.12.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00014-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.12.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00014-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.12.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00014-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.12.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00014-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.12.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00014-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.12.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00014-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.12.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00014-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.12.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00014-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.12.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00014-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.12.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00014-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.12.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00014-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.12.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00014-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.12.hc_attn_base", "shape": [24], "pack_file": "misc_model-00014-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.12.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00014-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.12.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00014-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.12.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00014-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.12.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00014-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.12.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00014-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.13.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00015-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.13.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00015-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.13.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00015-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.13.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00015-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.13.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00015-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.13.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00015-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.13.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00015-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.13.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00015-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.13.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00015-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.13.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00015-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.13.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00015-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.13.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00015-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.13.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00015-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.13.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00015-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.13.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00015-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.13.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00015-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.13.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00015-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.13.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00015-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.13.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00015-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.13.hc_attn_base", "shape": [24], "pack_file": "misc_model-00015-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.13.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00015-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.13.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00015-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.13.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00015-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.13.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00015-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.13.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00015-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.14.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00016-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.14.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00016-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.14.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00016-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.14.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00016-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.14.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00016-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.14.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00016-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.14.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00016-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.14.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00016-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.14.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00016-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.14.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00016-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.14.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00016-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.14.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00016-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.14.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00016-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.14.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00016-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.14.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00016-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.14.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00016-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.14.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00016-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.14.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00016-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.14.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00016-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.14.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00016-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.14.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00016-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.14.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00016-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.14.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00016-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.14.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00016-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.14.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00016-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.14.hc_attn_base", "shape": [24], "pack_file": "misc_model-00016-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.14.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00016-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.14.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00016-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.14.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00016-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.14.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00016-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.14.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00016-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.15.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00017-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.15.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00017-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.15.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00017-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.15.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00017-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.15.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00017-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.15.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00017-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.15.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00017-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.15.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00017-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.15.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00017-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.15.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00017-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.15.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00017-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.15.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00017-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.15.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00017-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.15.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00017-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.15.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00017-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.15.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00017-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.15.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00017-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.15.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00017-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.15.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00017-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.15.hc_attn_base", "shape": [24], "pack_file": "misc_model-00017-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.15.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00017-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.15.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00017-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.15.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00017-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.15.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00017-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.15.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00017-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.16.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00018-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.16.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00018-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.16.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00018-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.16.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00018-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.16.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00018-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.16.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00018-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.16.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00018-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.16.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00018-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.16.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00018-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.16.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00018-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.16.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00018-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.16.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00018-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.16.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00018-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.16.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00018-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.16.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00018-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.16.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00018-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.16.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00018-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.16.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00018-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.16.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00018-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.16.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00018-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.16.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00018-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.16.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00018-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.16.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00018-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.16.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00018-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.16.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00018-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.16.hc_attn_base", "shape": [24], "pack_file": "misc_model-00018-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.16.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00018-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.16.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00018-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.16.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00018-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.16.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00018-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.16.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00018-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.17.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00019-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.17.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00019-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.17.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00019-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.17.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00019-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.17.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00019-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.17.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00019-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.17.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00019-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.17.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00019-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.17.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00019-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.17.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00019-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.17.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00019-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.17.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00019-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.17.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00019-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.17.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00019-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.17.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00019-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.17.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00019-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.17.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00019-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.17.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00019-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.17.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00019-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.17.hc_attn_base", "shape": [24], "pack_file": "misc_model-00019-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.17.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00019-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.17.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00019-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.17.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00019-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.17.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00019-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.17.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00019-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.18.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00020-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.18.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00020-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.18.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00020-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.18.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00020-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.18.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00020-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.18.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00020-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.18.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00020-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.18.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00020-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.18.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00020-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.18.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00020-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.18.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00020-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.18.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00020-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.18.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00020-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.18.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00020-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.18.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00020-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.18.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00020-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.18.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00020-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.18.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00020-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.18.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00020-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.18.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00020-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.18.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00020-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.18.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00020-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.18.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00020-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.18.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00020-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.18.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00020-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.18.hc_attn_base", "shape": [24], "pack_file": "misc_model-00020-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.18.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00020-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.18.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00020-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.18.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00020-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.18.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00020-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.18.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00020-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.19.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00021-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.19.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00021-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.19.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00021-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.19.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00021-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.19.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00021-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.19.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00021-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.19.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00021-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.19.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00021-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.19.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00021-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.19.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00021-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.19.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00021-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.19.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00021-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.19.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00021-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.19.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00021-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.19.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00021-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.19.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00021-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.19.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00021-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.19.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00021-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.19.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00021-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.19.hc_attn_base", "shape": [24], "pack_file": "misc_model-00021-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.19.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00021-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.19.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00021-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.19.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00021-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.19.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00021-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.19.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00021-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.20.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00022-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.20.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00022-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.20.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00022-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.20.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00022-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.20.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00022-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.20.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00022-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.20.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00022-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.20.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00022-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.20.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00022-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.20.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00022-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.20.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00022-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.20.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00022-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.20.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00022-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.20.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00022-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.20.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00022-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.20.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00022-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.20.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00022-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.20.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00022-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.20.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00022-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.20.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00022-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.20.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00022-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.20.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00022-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.20.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00022-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.20.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00022-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.20.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00022-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.20.hc_attn_base", "shape": [24], "pack_file": "misc_model-00022-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.20.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00022-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.20.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00022-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.20.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00022-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.20.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00022-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.20.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00022-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.21.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00023-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.21.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00023-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.21.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00023-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.21.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00023-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.21.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00023-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.21.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00023-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.21.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00023-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.21.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00023-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.21.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00023-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.21.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00023-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.21.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00023-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.21.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00023-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.21.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00023-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.21.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00023-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.21.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00023-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.21.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00023-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.21.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00023-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.21.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00023-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.21.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00023-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.21.hc_attn_base", "shape": [24], "pack_file": "misc_model-00023-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.21.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00023-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.21.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00023-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.21.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00023-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.21.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00023-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.21.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00023-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.22.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00024-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.22.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00024-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.22.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00024-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.22.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00024-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.22.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00024-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.22.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00024-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.22.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00024-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.22.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00024-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.22.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00024-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.22.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00024-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.22.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00024-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.22.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00024-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.22.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00024-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.22.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00024-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.22.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00024-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.22.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00024-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.22.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00024-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.22.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00024-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.22.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00024-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.22.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00024-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.22.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00024-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.22.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00024-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.22.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00024-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.22.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00024-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.22.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00024-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.22.hc_attn_base", "shape": [24], "pack_file": "misc_model-00024-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.22.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00024-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.22.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00024-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.22.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00024-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.22.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00024-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.22.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00024-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.23.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00025-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.23.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00025-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.23.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00025-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.23.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00025-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.23.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00025-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.23.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00025-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.23.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00025-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.23.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00025-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.23.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00025-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.23.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00025-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.23.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00025-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.23.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00025-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.23.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00025-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.23.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00025-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.23.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00025-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.23.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00025-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.23.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00025-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.23.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00025-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.23.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00025-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.23.hc_attn_base", "shape": [24], "pack_file": "misc_model-00025-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.23.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00025-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.23.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00025-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.23.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00025-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.23.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00025-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.23.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00025-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.24.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00026-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.24.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00026-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.24.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00026-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.24.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00026-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.24.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00026-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.24.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00026-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.24.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00026-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.24.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00026-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.24.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00026-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.24.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00026-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.24.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00026-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.24.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00026-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.24.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00026-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.24.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00026-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.24.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00026-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.24.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00026-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.24.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00026-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.24.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00026-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.24.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00026-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.24.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00026-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.24.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00026-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.24.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00026-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.24.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00026-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.24.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00026-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.24.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00026-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.24.hc_attn_base", "shape": [24], "pack_file": "misc_model-00026-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.24.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00026-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.24.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00026-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.24.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00026-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.24.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00026-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.24.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00026-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.25.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00027-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.25.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00027-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.25.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00027-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.25.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00027-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.25.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00027-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.25.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00027-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.25.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00027-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.25.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00027-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.25.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00027-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.25.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00027-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.25.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00027-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.25.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00027-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.25.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00027-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.25.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00027-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.25.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00027-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.25.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00027-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.25.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00027-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.25.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00027-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.25.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00027-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.25.hc_attn_base", "shape": [24], "pack_file": "misc_model-00027-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.25.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00027-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.25.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00027-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.25.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00027-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.25.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00027-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.25.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00027-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.26.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00028-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.26.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00028-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.26.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00028-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.26.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00028-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.26.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00028-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.26.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00028-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.26.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00028-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.26.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00028-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.26.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00028-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.26.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00028-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.26.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00028-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.26.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00028-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.26.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00028-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.26.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00028-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.26.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00028-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.26.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00028-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.26.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00028-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.26.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00028-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.26.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00028-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.26.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00028-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.26.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00028-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.26.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00028-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.26.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00028-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.26.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00028-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.26.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00028-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.26.hc_attn_base", "shape": [24], "pack_file": "misc_model-00028-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.26.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00028-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.26.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00028-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.26.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00028-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.26.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00028-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.26.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00028-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.27.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00029-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.27.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00029-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.27.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00029-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.27.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00029-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.27.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00029-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.27.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00029-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.27.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00029-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.27.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00029-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.27.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00029-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.27.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00029-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.27.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00029-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.27.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00029-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.27.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00029-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.27.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00029-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.27.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00029-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.27.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00029-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.27.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00029-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.27.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00029-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.27.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00029-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.27.hc_attn_base", "shape": [24], "pack_file": "misc_model-00029-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.27.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00029-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.27.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00029-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.27.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00029-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.27.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00029-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.27.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00029-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.28.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00030-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.28.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00030-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.28.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00030-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.28.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00030-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.28.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00030-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.28.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00030-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.28.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00030-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.28.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00030-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.28.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00030-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.28.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00030-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.28.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00030-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.28.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00030-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.28.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00030-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.28.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00030-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.28.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00030-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.28.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00030-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.28.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00030-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.28.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00030-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.28.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00030-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.28.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00030-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.28.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00030-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.28.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00030-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.28.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00030-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.28.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00030-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.28.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00030-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.28.hc_attn_base", "shape": [24], "pack_file": "misc_model-00030-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.28.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00030-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.28.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00030-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.28.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00030-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.28.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00030-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.28.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00030-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.29.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00031-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.29.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00031-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.29.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00031-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.29.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00031-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.29.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00031-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.29.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00031-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.29.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00031-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.29.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00031-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.29.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00031-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.29.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00031-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.29.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00031-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.29.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00031-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.29.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00031-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.29.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00031-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.29.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00031-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.29.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00031-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.29.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00031-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.29.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00031-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.29.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00031-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.29.hc_attn_base", "shape": [24], "pack_file": "misc_model-00031-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.29.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00031-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.29.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00031-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.29.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00031-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.29.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00031-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.29.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00031-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.30.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00032-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.30.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00032-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.30.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00032-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.30.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00032-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.30.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00032-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.30.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00032-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.30.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00032-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.30.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00032-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.30.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00032-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.30.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00032-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.30.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00032-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.30.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00032-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.30.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00032-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.30.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00032-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.30.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00032-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.30.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00032-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.30.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00032-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.30.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00032-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.30.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00032-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.30.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00032-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.30.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00032-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.30.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00032-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.30.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00032-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.30.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00032-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.30.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00032-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.30.hc_attn_base", "shape": [24], "pack_file": "misc_model-00032-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.30.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00032-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.30.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00032-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.30.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00032-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.30.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00032-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.30.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00032-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.31.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00033-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.31.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00033-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.31.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00033-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.31.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00033-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.31.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00033-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.31.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00033-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.31.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00033-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.31.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00033-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.31.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00033-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.31.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00033-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.31.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00033-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.31.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00033-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.31.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00033-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.31.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00033-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.31.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00033-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.31.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00033-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.31.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00033-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.31.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00033-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.31.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00033-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.31.hc_attn_base", "shape": [24], "pack_file": "misc_model-00033-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.31.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00033-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.31.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00033-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.31.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00033-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.31.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00033-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.31.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00033-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.32.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00034-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.32.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00034-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.32.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00034-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.32.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00034-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.32.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00034-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.32.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00034-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.32.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00034-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.32.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00034-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.32.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00034-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.32.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00034-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.32.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00034-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.32.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00034-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.32.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00034-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.32.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00034-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.32.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00034-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.32.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00034-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.32.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00034-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.32.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00034-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.32.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00034-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.32.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00034-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.32.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00034-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.32.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00034-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.32.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00034-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.32.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00034-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.32.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00034-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.32.hc_attn_base", "shape": [24], "pack_file": "misc_model-00034-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.32.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00034-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.32.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00034-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.32.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00034-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.32.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00034-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.32.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00034-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.33.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00035-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.33.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00035-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.33.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00035-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.33.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00035-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.33.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00035-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.33.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00035-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.33.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00035-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.33.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00035-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.33.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00035-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.33.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00035-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.33.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00035-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.33.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00035-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.33.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00035-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.33.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00035-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.33.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00035-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.33.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00035-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.33.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00035-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.33.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00035-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.33.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00035-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.33.hc_attn_base", "shape": [24], "pack_file": "misc_model-00035-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.33.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00035-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.33.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00035-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.33.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00035-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.33.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00035-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.33.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00035-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.34.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00036-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.34.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00036-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.34.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00036-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.34.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00036-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.34.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00036-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.34.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00036-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.34.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00036-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.34.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00036-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.34.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00036-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.34.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00036-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.34.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00036-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.34.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00036-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.34.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00036-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.34.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00036-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.34.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00036-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.34.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00036-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.34.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00036-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.34.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00036-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.34.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00036-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.34.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00036-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.34.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00036-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.34.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00036-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.34.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00036-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.34.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00036-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.34.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00036-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.34.hc_attn_base", "shape": [24], "pack_file": "misc_model-00036-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.34.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00036-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.34.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00036-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.34.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00036-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.34.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00036-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.34.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00036-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.35.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00037-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.35.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00037-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.35.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00037-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.35.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00037-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.35.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00037-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.35.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00037-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.35.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00037-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.35.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00037-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.35.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00037-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.35.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00037-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.35.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00037-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.35.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00037-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.35.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00037-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.35.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00037-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.35.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00037-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.35.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00037-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.35.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00037-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.35.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00037-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.35.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00037-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.35.hc_attn_base", "shape": [24], "pack_file": "misc_model-00037-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.35.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00037-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.35.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00037-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.35.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00037-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.35.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00037-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.35.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00037-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.36.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00038-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.36.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00038-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.36.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00038-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.36.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00038-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.36.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00038-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.36.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00038-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.36.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00038-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.36.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00038-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.36.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00038-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.36.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00038-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.36.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00038-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.36.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00038-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.36.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00038-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.36.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00038-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.36.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00038-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.36.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00038-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.36.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00038-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.36.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00038-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.36.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00038-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.36.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00038-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.36.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00038-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.36.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00038-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.36.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00038-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.36.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00038-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.36.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00038-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.36.hc_attn_base", "shape": [24], "pack_file": "misc_model-00038-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.36.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00038-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.36.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00038-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.36.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00038-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.36.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00038-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.36.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00038-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.37.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00039-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.37.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00039-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.37.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00039-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.37.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00039-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.37.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00039-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.37.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00039-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.37.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00039-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.37.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00039-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.37.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00039-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.37.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00039-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.37.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00039-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.37.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00039-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.37.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00039-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.37.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00039-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.37.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00039-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.37.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00039-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.37.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00039-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.37.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00039-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.37.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00039-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.37.hc_attn_base", "shape": [24], "pack_file": "misc_model-00039-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.37.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00039-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.37.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00039-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.37.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00039-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.37.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00039-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.37.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00039-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.38.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00040-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.38.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00040-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.38.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00040-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.38.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00040-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.38.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00040-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.38.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00040-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.38.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00040-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.38.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00040-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.38.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00040-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.38.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00040-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.38.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00040-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.38.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00040-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.38.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00040-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.38.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00040-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.38.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00040-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.38.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00040-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.38.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00040-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.38.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00040-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.38.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00040-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.38.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00040-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.38.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00040-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.38.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00040-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.38.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00040-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.38.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00040-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.38.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00040-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.38.hc_attn_base", "shape": [24], "pack_file": "misc_model-00040-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.38.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00040-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.38.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00040-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.38.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00040-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.38.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00040-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.38.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00040-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.39.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00041-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.39.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00041-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.39.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00041-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.39.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00041-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.39.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00041-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.39.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00041-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.39.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00041-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.39.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00041-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.39.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00041-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.39.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00041-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.39.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00041-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.39.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00041-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.39.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00041-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.39.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00041-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.39.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00041-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.39.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00041-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.39.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00041-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.39.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00041-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.39.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00041-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.39.hc_attn_base", "shape": [24], "pack_file": "misc_model-00041-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.39.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00041-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.39.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00041-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.39.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00041-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.39.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00041-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.39.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00041-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.40.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00042-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.40.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00042-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.40.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00042-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.40.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00042-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.40.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00042-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.40.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00042-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.40.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00042-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.40.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00042-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.40.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00042-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.40.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00042-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.40.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00042-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.40.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00042-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.40.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00042-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.40.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00042-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.40.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00042-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.40.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00042-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.40.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00042-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.40.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00042-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.40.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00042-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.40.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00042-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.40.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00042-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.40.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00042-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.40.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00042-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.40.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00042-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.40.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00042-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.40.hc_attn_base", "shape": [24], "pack_file": "misc_model-00042-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.40.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00042-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.40.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00042-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.40.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00042-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.40.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00042-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.40.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00042-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.41.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00043-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.41.attn.compressor.ape", "shape": [128, 512], "pack_file": "misc_model-00043-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "layers.41.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00043-of-00046.pack", "offset": 131200, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.41.attn.compressor.wgate.weight", "shape": [512, 4096], "pack_file": "misc_model-00043-of-00046.pack", "offset": 132224, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.41.attn.compressor.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00043-of-00046.pack", "offset": 4326528, "kind": "fp16", "sections": [4194304], "nbytes": 4194304}
{"name": "layers.41.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00043-of-00046.pack", "offset": 8520832, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.41.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00043-of-00046.pack", "offset": 8521856, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.41.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00043-of-00046.pack", "offset": 8523904, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.41.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00043-of-00046.pack", "offset": 10621184, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.41.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00043-of-00046.pack", "offset": 44177664, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.41.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00043-of-00046.pack", "offset": 77734144, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.41.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00043-of-00046.pack", "offset": 81928704, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.41.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00043-of-00046.pack", "offset": 115485184, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.41.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00043-of-00046.pack", "offset": 115493376, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.41.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00043-of-00046.pack", "offset": 115493888, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.41.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00043-of-00046.pack", "offset": 117591040, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.41.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00043-of-00046.pack", "offset": 125980160, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.41.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00043-of-00046.pack", "offset": 134369280, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.41.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00043-of-00046.pack", "offset": 142758400, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.41.hc_attn_base", "shape": [24], "pack_file": "misc_model-00043-of-00046.pack", "offset": 142766592, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.41.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00043-of-00046.pack", "offset": 142766640, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.41.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00043-of-00046.pack", "offset": 143553072, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.41.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00043-of-00046.pack", "offset": 143553078, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.41.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00043-of-00046.pack", "offset": 143553126, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.41.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00043-of-00046.pack", "offset": 144339558, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.42.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00044-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.42.attn.compressor.ape", "shape": [4, 1024], "pack_file": "misc_model-00044-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.42.attn.compressor.norm.weight", "shape": [512], "pack_file": "misc_model-00044-of-00046.pack", "offset": 8320, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.42.attn.compressor.wgate.weight", "shape": [1024, 4096], "pack_file": "misc_model-00044-of-00046.pack", "offset": 9344, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.42.attn.compressor.wkv.weight", "shape": [1024, 4096], "pack_file": "misc_model-00044-of-00046.pack", "offset": 8397952, "kind": "fp16", "sections": [8388608], "nbytes": 8388608}
{"name": "layers.42.attn.indexer.compressor.ape", "shape": [4, 256], "pack_file": "misc_model-00044-of-00046.pack", "offset": 16786560, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.42.attn.indexer.compressor.norm.weight", "shape": [128], "pack_file": "misc_model-00044-of-00046.pack", "offset": 16788608, "kind": "fp16", "sections": [256], "nbytes": 256}
{"name": "layers.42.attn.indexer.compressor.wgate.weight", "shape": [256, 4096], "pack_file": "misc_model-00044-of-00046.pack", "offset": 16788864, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.42.attn.indexer.compressor.wkv.weight", "shape": [256, 4096], "pack_file": "misc_model-00044-of-00046.pack", "offset": 18886016, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.42.attn.indexer.weights_proj.weight", "shape": [64, 4096], "pack_file": "misc_model-00044-of-00046.pack", "offset": 20983168, "kind": "fp16", "sections": [524288], "nbytes": 524288}
{"name": "layers.42.attn.indexer.wq_b.weight", "shape": [8192, 1024], "pack_file": "misc_model-00044-of-00046.pack", "offset": 21507456, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [64, 8], "nbytes": 8389120}
{"name": "layers.42.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00044-of-00046.pack", "offset": 29896576, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.42.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00044-of-00046.pack", "offset": 29897600, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.42.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00044-of-00046.pack", "offset": 29899648, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.42.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00044-of-00046.pack", "offset": 31996928, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.42.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00044-of-00046.pack", "offset": 65553408, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.42.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00044-of-00046.pack", "offset": 99109888, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.42.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00044-of-00046.pack", "offset": 103304448, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.42.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00044-of-00046.pack", "offset": 136860928, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.42.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00044-of-00046.pack", "offset": 136869120, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "layers.42.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00044-of-00046.pack", "offset": 136869632, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.42.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00044-of-00046.pack", "offset": 138966784, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.42.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00044-of-00046.pack", "offset": 147355904, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.42.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00044-of-00046.pack", "offset": 155745024, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.42.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00044-of-00046.pack", "offset": 164134144, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.42.hc_attn_base", "shape": [24], "pack_file": "misc_model-00044-of-00046.pack", "offset": 164142336, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.42.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00044-of-00046.pack", "offset": 164142384, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.42.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00044-of-00046.pack", "offset": 164928816, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.42.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00044-of-00046.pack", "offset": 164928822, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.42.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00044-of-00046.pack", "offset": 164928870, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.42.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00044-of-00046.pack", "offset": 165715302, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "hc_head_base", "shape": [4], "pack_file": "misc_model-00045-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [8], "nbytes": 8}
{"name": "hc_head_fn", "shape": [4, 16384], "pack_file": "misc_model-00045-of-00046.pack", "offset": 8, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "hc_head_scale", "shape": [1], "pack_file": "misc_model-00045-of-00046.pack", "offset": 131080, "kind": "fp16", "sections": [2], "nbytes": 2}
{"name": "head.weight", "shape": [129280, 4096], "pack_file": "misc_model-00045-of-00046.pack", "offset": 131082, "kind": "fp16", "sections": [1059061760], "nbytes": 1059061760}
{"name": "norm.weight", "shape": [4096], "pack_file": "misc_model-00045-of-00046.pack", "offset": 1059192842, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "mtp.0.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00046-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "mtp.0.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00046-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "mtp.0.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00046-of-00046.pack", "offset": 1152, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "mtp.0.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00046-of-00046.pack", "offset": 3200, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "mtp.0.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00046-of-00046.pack", "offset": 2100480, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "mtp.0.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00046-of-00046.pack", "offset": 35656960, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "mtp.0.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00046-of-00046.pack", "offset": 69213440, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "mtp.0.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00046-of-00046.pack", "offset": 73408000, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "mtp.0.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00046-of-00046.pack", "offset": 106964480, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "mtp.0.e_proj.weight", "shape": [4096, 4096], "pack_file": "misc_model-00046-of-00046.pack", "offset": 106972672, "kind": "fp8_block", "sections": [16777216, 1024], "scale_shape": [32, 32], "nbytes": 16778240}
{"name": "mtp.0.enorm.weight", "shape": [4096], "pack_file": "misc_model-00046-of-00046.pack", "offset": 123750912, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "mtp.0.ffn.gate.bias", "shape": [256], "pack_file": "misc_model-00046-of-00046.pack", "offset": 123759104, "kind": "fp16", "sections": [512], "nbytes": 512}
{"name": "mtp.0.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00046-of-00046.pack", "offset": 123759616, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "mtp.0.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00046-of-00046.pack", "offset": 125856768, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "mtp.0.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00046-of-00046.pack", "offset": 134245888, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "mtp.0.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00046-of-00046.pack", "offset": 142635008, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "mtp.0.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00046-of-00046.pack", "offset": 151024128, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "mtp.0.h_proj.weight", "shape": [4096, 4096], "pack_file": "misc_model-00046-of-00046.pack", "offset": 151032320, "kind": "fp8_block", "sections": [16777216, 1024], "scale_shape": [32, 32], "nbytes": 16778240}
{"name": "mtp.0.hc_attn_base", "shape": [24], "pack_file": "misc_model-00046-of-00046.pack", "offset": 167810560, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "mtp.0.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00046-of-00046.pack", "offset": 167810608, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "mtp.0.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00046-of-00046.pack", "offset": 168597040, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "mtp.0.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00046-of-00046.pack", "offset": 168597046, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "mtp.0.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00046-of-00046.pack", "offset": 168597094, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "mtp.0.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00046-of-00046.pack", "offset": 169383526, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "mtp.0.hc_head_base", "shape": [4], "pack_file": "misc_model-00046-of-00046.pack", "offset": 169383532, "kind": "fp16", "sections": [8], "nbytes": 8}
{"name": "mtp.0.hc_head_fn", "shape": [4, 16384], "pack_file": "misc_model-00046-of-00046.pack", "offset": 169383540, "kind": "fp16", "sections": [131072], "nbytes": 131072}
{"name": "mtp.0.hc_head_scale", "shape": [1], "pack_file": "misc_model-00046-of-00046.pack", "offset": 169514612, "kind": "fp16", "sections": [2], "nbytes": 2}
{"name": "mtp.0.hnorm.weight", "shape": [4096], "pack_file": "misc_model-00046-of-00046.pack", "offset": 169514614, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "mtp.0.norm.weight", "shape": [4096], "pack_file": "misc_model-00046-of-00046.pack", "offset": 169522806, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.0.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00002-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.0.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00002-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.0.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00002-of-00046.pack", "offset": 1152, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.0.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00002-of-00046.pack", "offset": 3200, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.0.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00002-of-00046.pack", "offset": 2100480, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.0.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00002-of-00046.pack", "offset": 35656960, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.0.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00002-of-00046.pack", "offset": 69213440, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.0.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00002-of-00046.pack", "offset": 73408000, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.0.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00002-of-00046.pack", "offset": 106964480, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.0.ffn.gate.tid2eid", "shape": [129280, 6], "pack_file": "misc_model-00002-of-00046.pack", "offset": 106972672, "kind": "torch.int64", "sections": [6205440], "nbytes": 6205440}
{"name": "layers.0.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00002-of-00046.pack", "offset": 113178112, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.0.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00002-of-00046.pack", "offset": 115275264, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.0.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00002-of-00046.pack", "offset": 123664384, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.0.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00002-of-00046.pack", "offset": 132053504, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.0.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00002-of-00046.pack", "offset": 140442624, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.0.hc_attn_base", "shape": [24], "pack_file": "misc_model-00002-of-00046.pack", "offset": 140450816, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.0.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00002-of-00046.pack", "offset": 140450864, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.0.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00002-of-00046.pack", "offset": 141237296, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.0.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00002-of-00046.pack", "offset": 141237302, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.0.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00002-of-00046.pack", "offset": 141237350, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.0.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00002-of-00046.pack", "offset": 142023782, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.1.attn.attn_sink", "shape": [64], "pack_file": "misc_model-00003-of-00046.pack", "offset": 0, "kind": "fp16", "sections": [128], "nbytes": 128}
{"name": "layers.1.attn.kv_norm.weight", "shape": [512], "pack_file": "misc_model-00003-of-00046.pack", "offset": 128, "kind": "fp16", "sections": [1024], "nbytes": 1024}
{"name": "layers.1.attn.q_norm.weight", "shape": [1024], "pack_file": "misc_model-00003-of-00046.pack", "offset": 1152, "kind": "fp16", "sections": [2048], "nbytes": 2048}
{"name": "layers.1.attn.wkv.weight", "shape": [512, 4096], "pack_file": "misc_model-00003-of-00046.pack", "offset": 3200, "kind": "fp8_block", "sections": [2097152, 128], "scale_shape": [4, 32], "nbytes": 2097280}
{"name": "layers.1.attn.wo_a.weight", "shape": [8192, 4096], "pack_file": "misc_model-00003-of-00046.pack", "offset": 2100480, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [64, 32], "nbytes": 33556480}
{"name": "layers.1.attn.wo_b.weight", "shape": [4096, 8192], "pack_file": "misc_model-00003-of-00046.pack", "offset": 35656960, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [32, 64], "nbytes": 33556480}
{"name": "layers.1.attn.wq_a.weight", "shape": [1024, 4096], "pack_file": "misc_model-00003-of-00046.pack", "offset": 69213440, "kind": "fp8_block", "sections": [4194304, 256], "scale_shape": [8, 32], "nbytes": 4194560}
{"name": "layers.1.attn.wq_b.weight", "shape": [32768, 1024], "pack_file": "misc_model-00003-of-00046.pack", "offset": 73408000, "kind": "fp8_block", "sections": [33554432, 2048], "scale_shape": [256, 8], "nbytes": 33556480}
{"name": "layers.1.attn_norm.weight", "shape": [4096], "pack_file": "misc_model-00003-of-00046.pack", "offset": 106964480, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.1.ffn.gate.tid2eid", "shape": [129280, 6], "pack_file": "misc_model-00003-of-00046.pack", "offset": 106972672, "kind": "torch.int64", "sections": [6205440], "nbytes": 6205440}
{"name": "layers.1.ffn.gate.weight", "shape": [256, 4096], "pack_file": "misc_model-00003-of-00046.pack", "offset": 113178112, "kind": "fp16", "sections": [2097152], "nbytes": 2097152}
{"name": "layers.1.ffn.shared_experts.w1.weight", "shape": [2048, 4096], "pack_file": "misc_model-00003-of-00046.pack", "offset": 115275264, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.1.ffn.shared_experts.w2.weight", "shape": [4096, 2048], "pack_file": "misc_model-00003-of-00046.pack", "offset": 123664384, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [32, 16], "nbytes": 8389120}
{"name": "layers.1.ffn.shared_experts.w3.weight", "shape": [2048, 4096], "pack_file": "misc_model-00003-of-00046.pack", "offset": 132053504, "kind": "fp8_block", "sections": [8388608, 512], "scale_shape": [16, 32], "nbytes": 8389120}
{"name": "layers.1.ffn_norm.weight", "shape": [4096], "pack_file": "misc_model-00003-of-00046.pack", "offset": 140442624, "kind": "fp16", "sections": [8192], "nbytes": 8192}
{"name": "layers.1.hc_attn_base", "shape": [24], "pack_file": "misc_model-00003-of-00046.pack", "offset": 140450816, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.1.hc_attn_fn", "shape": [24, 16384], "pack_file": "misc_model-00003-of-00046.pack", "offset": 140450864, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.1.hc_attn_scale", "shape": [3], "pack_file": "misc_model-00003-of-00046.pack", "offset": 141237296, "kind": "fp16", "sections": [6], "nbytes": 6}
{"name": "layers.1.hc_ffn_base", "shape": [24], "pack_file": "misc_model-00003-of-00046.pack", "offset": 141237302, "kind": "fp16", "sections": [48], "nbytes": 48}
{"name": "layers.1.hc_ffn_fn", "shape": [24, 16384], "pack_file": "misc_model-00003-of-00046.pack", "offset": 141237350, "kind": "fp16", "sections": [786432], "nbytes": 786432}
{"name": "layers.1.hc_ffn_scale", "shape": [3], "pack_file": "misc_model-00003-of-00046.pack", "offset": 142023782, "kind": "fp16", "sections": [6], "nbytes": 6}