{ "architectures": [ "Qwen4ExpForConditionalGeneration" ], "image_token_id": 248056, "language_model_only": false, "model_type": "qwen4_exp", "text_config": { "attention_bias": false, "attention_dropout": 0.0, "bos_token_id": 248044, "dtype": "bfloat16", "eos_token_id": 248044, "full_attention_interval": 4, "hc_count": 4, "hc_lowrank": 320, "head_dim": 256, "heads_per_ngram": 8, "hidden_act": "silu", "hidden_size": 2560, "indexer_budget": 2048, "indexer_compress_ratio": 4, "indexer_head_dim": 128, "indexer_kv_heads": 1, "indexer_n_heads": 4, "initializer_range": 0.02, "layer_types": [ "linear_attention", "linear_attention", "linear_attention", "full_attention", "linear_attention", "linear_attention", "linear_attention", "full_attention", "linear_attention", "linear_attention", "linear_attention", "full_attention", "linear_attention", "linear_attention", "linear_attention", "full_attention", "linear_attention", "linear_attention", "linear_attention", "full_attention", "linear_attention", "linear_attention", "linear_attention", "full_attention", "linear_attention", "linear_attention", "linear_attention", "full_attention", "linear_attention", "linear_attention", "linear_attention", "full_attention", "linear_attention", "linear_attention", "linear_attention", "full_attention", "linear_attention", "linear_attention", "linear_attention", "full_attention", "linear_attention", "linear_attention", "linear_attention", "full_attention", "linear_attention", "linear_attention", "linear_attention", "full_attention" ], "linear_conv_kernel_dim": 4, "linear_key_head_dim": 128, "linear_num_key_heads": 16, "linear_num_value_heads": 48, "linear_value_head_dim": 128, "make_ngram_vocab_size_divisible_by": 128, "mamba_ssm_dtype": "float32", "max_position_embeddings": 262144, "model_type": "qwen4_exp_text", "moe_intermediate_size": 640, "mtp": { "hybrid": true, "layer_types": [ "full_attention" ], "mtp_use_hidden_state_from_layer": null, "num_hidden_layers": 1, "rope_theta": 10000000 }, "mtp_num_hidden_layers": 1, "mtp_use_dedicated_embeddings": false, "ngram_size": 3, "ngram_vocab_size_base": 20000000, "num_attention_heads": 24, "num_experts": 512, "num_experts_per_tok": 10, "num_hidden_layers": 48, "num_key_value_heads": 2, "output_gate_type": "sigmoid", "output_router_logits": false, "pad_token_id": null, "partial_rotary_factor": 0.25, "ple_conv_kernel_size": 4, "ple_embed_dim": 2560, "ple_layer_ids": [ 2 ], "rms_norm_eps": 1e-06, "rope_parameters": { "mrope_interleaved": true, "mrope_section": [ 11, 11, 10 ], "partial_rotary_factor": 0.25, "rope_theta": 10000000, "rope_type": "default" }, "router_aux_loss_coef": 0.001, "shared_expert_intermediate_size": 640, "split_ngram_parts": 128, "tie_word_embeddings": false, "use_cache": true, "vocab_size": 248320 }, "tie_word_embeddings": false, "transformers_version": "5.8.0.dev0", "video_token_id": 248057, "vision_config": { "deepstack_visual_indexes": [], "depth": 27, "hidden_act": "gelu_pytorch_tanh", "hidden_size": 1152, "in_channels": 3, "initializer_range": 0.02, "intermediate_size": 4304, "model_type": "qwen4_exp", "num_heads": 16, "num_position_embeddings": 2304, "out_hidden_size": 2560, "patch_size": 16, "spatial_merge_size": 2, "temporal_patch_size": 2 }, "vision_end_token_id": 248054, "vision_start_token_id": 248053, "jang_config": { "format": "jang_v2", "family": "qwen4_exp", "norm_convention": "runtime_plus1_applied", "bit_map": { "default": { "bits": 8, "group_size": 64 }, "language_model.layers.0.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.1.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.2.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.3.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.4.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.5.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.6.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.7.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.8.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.9.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.10.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.11.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.12.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.13.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.14.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.15.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.16.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.17.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.18.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.19.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.20.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.21.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.22.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.23.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.24.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.25.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.26.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.27.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.28.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.29.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.30.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.31.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.32.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.33.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.34.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.35.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.36.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.37.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.38.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.39.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.40.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.41.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.42.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.43.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.44.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.45.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.46.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.47.mlp.switch_mlp": { "bits": 4, "group_size": 64 }, "language_model.layers.*.ple.ngram_embedding.shards.0.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.1.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.2.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.3.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.4.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.5.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.6.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.7.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.8.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.9.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.10.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.11.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.12.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.13.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.14.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.15.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.16.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.17.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.18.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.19.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.20.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.21.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.22.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.23.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.24.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.25.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.26.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.27.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.28.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.29.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.30.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.31.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.32.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.33.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.34.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.35.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.36.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.37.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.38.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.39.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.40.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.41.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.42.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.43.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.44.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.45.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.46.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.47.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.48.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.49.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.50.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.51.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.52.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.53.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.54.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.55.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.56.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.57.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.58.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.59.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.60.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.61.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.62.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.63.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.64.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.65.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.66.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.67.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.68.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.69.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.70.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.71.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.72.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.73.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.74.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.75.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.76.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.77.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.78.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.79.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.80.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.81.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.82.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.83.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.84.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.85.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.86.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.87.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.88.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.89.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.90.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.91.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.92.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.93.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.94.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.95.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.96.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.97.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.98.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.99.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.100.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.101.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.102.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.103.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.104.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.105.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.106.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.107.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.108.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.109.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.110.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.111.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.112.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.113.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.114.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.115.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.116.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.117.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.118.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.119.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.120.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.121.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.122.weight": { "bits": 4, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.123.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.124.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.125.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.126.weight": { "bits": 3, "group_size": 32 }, "language_model.layers.*.ple.ngram_embedding.shards.127.weight": { "bits": 3, "group_size": 32 }, "language_model.embed_tokens": { "bits": 8, "group_size": 64 }, "lm_head": { "bits": 8, "group_size": 64 }, "visual": { "bits": 8, "group_size": 64 }, "language_model.layers.*.mlp.gate.weight": { "bits": 8, "group_size": 64 }, "language_model.layers.*.self_attn": { "bits": 8, "group_size": 64 }, "language_model.layers.*.mlp.shared_expert.": { "bits": 8, "group_size": 64 }, "language_model.layers.*.linear_attn": { "bits": 8, "group_size": 64 }, "language_model.layers.*.ple.key_proj": { "bits": 8, "group_size": 64 }, "language_model.layers.*.ple.value_proj": { "bits": 8, "group_size": 64 }, "mtp.": { "bits": 4, "group_size": 64 } }, "created": "2026-08-26", "quantization": { "calibrated": true, "imatrix": "diag.safetensors", "imatrix_refit": true, "awq_folded": true, "awq_scales": "awq_scales.safetensors", "hessian_allocation": "jang_4m.json", "gptq": false }, "chat": { "sampling_defaults": { "temperature": 1.0, "top_p": 0.95, "top_k": 20, "min_p": 0.0, "presence_penalty": 0.0, "repetition_penalty": 1.0 }, "sampling_modes": { "thinking": { "temperature": 1.0, "top_p": 0.95, "top_k": 20, "min_p": 0.0, "presence_penalty": 0.0, "repetition_penalty": 1.0 }, "instruct": { "temperature": 0.7, "top_p": 0.8, "top_k": 20, "min_p": 0.0, "presence_penalty": 1.5, "repetition_penalty": 1.0 } }, "default_sampling_mode": "thinking", "recommended_max_tokens": { "reasoning": 262144, "final_response": 131072 } }, "reasoning": { "supported": true, "default": "on", "supported_reasoning_efforts": [ "low", "medium", "xhigh" ], "default_reasoning_effort": "xhigh", "reasoning_effort_transport": "chat_template_kwarg", "preserve_thinking_supported": true, "preserve_thinking_default": true, "preserve_thinking_transport": "chat_template_kwarg", "notes": "Thinking ON by default ( block). enable_thinking=False prefills a closed think block. Efforts verified 2026-08-26: low/medium inject system directives, xhigh==default render." }, "capabilities": { "has_vision": true, "has_video": true, "has_audio": false, "supports_thinking": true, "default_reasoning": "on", "think_in_template": true, "reasoning_parser": "qwen3", "tool_parser": "hermes" }, "mtp": { "mtp_mode": "preserved_enabled", "num_layers": 1, "trained_multi_step": true, "notes": "MTP head preserved (4-bit gs64 class). best_depth requires a measured acceptance+wall-clock sweep on the target runtime before any recommendation (unvalidated until then)." }, "context": { "native": 262144, "extensible_yarn": 1000000 } }, "capabilities": { "has_vision": true, "has_video": true, "has_audio": false, "supports_thinking": true, "default_reasoning": "on", "think_in_template": true, "reasoning_parser": "qwen3", "tool_parser": "hermes" }, "quantization": { "group_size": 64, "bits": 8, "language_model.embed_tokens": { "group_size": 64, "bits": 8 }, "language_model.layers.0.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.0.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.0.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.0.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.0.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.0.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.0.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.0.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.0.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.0.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.0.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.0.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.1.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.1.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.1.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.1.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.1.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.1.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.1.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.1.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.1.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.1.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.1.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.1.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.1.ple.key_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.1.ple.ngram_embedding.shards.0": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.1": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.10": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.100": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.101": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.102": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.103": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.104": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.105": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.106": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.107": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.108": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.109": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.11": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.110": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.111": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.112": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.113": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.114": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.115": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.116": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.117": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.118": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.119": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.12": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.120": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.121": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.122": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.123": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.124": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.125": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.126": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.127": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.13": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.14": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.15": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.16": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.17": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.18": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.19": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.2": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.20": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.21": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.22": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.23": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.24": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.25": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.26": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.27": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.28": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.29": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.3": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.30": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.31": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.32": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.33": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.34": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.35": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.36": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.37": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.38": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.39": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.4": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.40": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.41": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.42": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.43": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.44": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.45": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.46": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.47": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.48": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.49": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.5": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.50": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.51": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.52": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.53": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.54": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.55": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.56": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.57": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.58": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.59": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.6": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.60": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.61": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.62": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.63": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.64": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.65": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.66": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.67": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.68": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.69": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.7": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.70": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.71": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.72": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.73": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.74": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.75": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.76": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.77": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.78": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.79": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.8": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.80": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.81": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.82": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.83": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.84": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.85": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.86": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.87": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.88": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.89": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.9": { "group_size": 32, "bits": 4 }, "language_model.layers.1.ple.ngram_embedding.shards.90": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.91": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.92": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.93": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.94": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.95": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.96": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.97": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.98": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.ngram_embedding.shards.99": { "group_size": 32, "bits": 3 }, "language_model.layers.1.ple.value_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.10.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.10.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.10.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.10.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.10.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.10.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.10.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.10.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.10.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.10.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.10.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.10.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.11.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.11.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.11.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.11.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.11.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.11.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.11.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.11.self_attn.indexer.index_qk_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.11.self_attn.k_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.11.self_attn.o_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.11.self_attn.q_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.11.self_attn.v_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.12.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.12.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.12.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.12.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.12.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.12.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.12.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.12.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.12.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.12.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.12.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.12.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.13.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.13.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.13.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.13.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.13.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.13.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.13.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.13.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.13.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.13.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.13.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.13.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.14.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.14.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.14.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.14.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.14.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.14.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.14.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.14.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.14.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.14.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.14.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.14.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.15.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.15.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.15.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.15.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.15.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.15.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.15.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.15.self_attn.indexer.index_qk_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.15.self_attn.k_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.15.self_attn.o_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.15.self_attn.q_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.15.self_attn.v_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.16.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.16.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.16.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.16.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.16.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.16.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.16.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.16.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.16.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.16.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.16.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.16.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.17.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.17.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.17.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.17.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.17.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.17.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.17.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.17.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.17.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.17.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.17.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.17.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.18.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.18.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.18.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.18.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.18.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.18.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.18.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.18.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.18.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.18.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.18.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.18.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.19.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.19.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.19.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.19.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.19.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.19.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.19.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.19.self_attn.indexer.index_qk_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.19.self_attn.k_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.19.self_attn.o_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.19.self_attn.q_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.19.self_attn.v_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.2.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.2.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.2.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.2.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.2.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.2.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.2.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.2.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.2.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.2.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.2.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.2.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.20.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.20.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.20.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.20.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.20.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.20.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.20.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.20.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.20.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.20.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.20.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.20.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.21.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.21.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.21.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.21.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.21.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.21.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.21.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.21.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.21.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.21.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.21.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.21.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.22.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.22.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.22.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.22.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.22.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.22.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.22.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.22.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.22.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.22.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.22.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.22.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.23.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.23.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.23.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.23.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.23.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.23.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.23.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.23.self_attn.indexer.index_qk_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.23.self_attn.k_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.23.self_attn.o_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.23.self_attn.q_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.23.self_attn.v_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.24.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.24.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.24.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.24.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.24.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.24.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.24.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.24.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.24.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.24.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.24.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.24.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.25.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.25.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.25.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.25.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.25.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.25.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.25.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.25.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.25.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.25.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.25.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.25.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.26.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.26.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.26.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.26.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.26.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.26.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.26.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.26.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.26.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.26.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.26.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.26.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.27.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.27.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.27.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.27.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.27.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.27.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.27.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.27.self_attn.indexer.index_qk_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.27.self_attn.k_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.27.self_attn.o_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.27.self_attn.q_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.27.self_attn.v_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.28.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.28.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.28.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.28.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.28.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.28.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.28.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.28.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.28.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.28.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.28.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.28.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.29.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.29.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.29.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.29.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.29.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.29.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.29.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.29.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.29.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.29.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.29.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.29.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.3.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.3.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.3.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.3.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.3.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.3.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.3.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.3.self_attn.indexer.index_qk_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.3.self_attn.k_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.3.self_attn.o_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.3.self_attn.q_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.3.self_attn.v_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.30.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.30.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.30.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.30.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.30.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.30.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.30.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.30.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.30.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.30.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.30.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.30.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.31.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.31.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.31.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.31.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.31.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.31.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.31.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.31.self_attn.indexer.index_qk_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.31.self_attn.k_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.31.self_attn.o_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.31.self_attn.q_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.31.self_attn.v_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.32.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.32.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.32.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.32.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.32.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.32.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.32.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.32.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.32.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.32.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.32.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.32.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.33.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.33.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.33.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.33.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.33.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.33.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.33.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.33.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.33.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.33.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.33.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.33.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.34.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.34.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.34.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.34.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.34.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.34.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.34.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.34.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.34.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.34.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.34.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.34.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.35.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.35.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.35.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.35.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.35.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.35.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.35.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.35.self_attn.indexer.index_qk_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.35.self_attn.k_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.35.self_attn.o_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.35.self_attn.q_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.35.self_attn.v_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.36.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.36.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.36.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.36.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.36.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.36.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.36.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.36.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.36.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.36.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.36.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.36.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.37.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.37.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.37.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.37.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.37.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.37.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.37.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.37.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.37.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.37.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.37.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.37.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.38.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.38.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.38.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.38.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.38.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.38.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.38.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.38.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.38.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.38.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.38.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.38.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.39.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.39.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.39.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.39.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.39.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.39.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.39.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.39.self_attn.indexer.index_qk_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.39.self_attn.k_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.39.self_attn.o_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.39.self_attn.q_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.39.self_attn.v_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.4.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.4.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.4.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.4.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.4.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.4.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.4.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.4.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.4.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.4.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.4.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.4.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.40.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.40.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.40.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.40.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.40.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.40.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.40.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.40.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.40.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.40.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.40.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.40.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.41.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.41.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.41.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.41.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.41.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.41.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.41.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.41.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.41.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.41.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.41.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.41.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.42.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.42.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.42.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.42.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.42.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.42.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.42.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.42.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.42.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.42.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.42.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.42.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.43.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.43.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.43.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.43.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.43.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.43.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.43.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.43.self_attn.indexer.index_qk_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.43.self_attn.k_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.43.self_attn.o_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.43.self_attn.q_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.43.self_attn.v_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.44.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.44.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.44.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.44.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.44.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.44.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.44.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.44.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.44.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.44.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.44.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.44.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.45.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.45.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.45.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.45.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.45.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.45.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.45.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.45.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.45.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.45.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.45.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.45.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.46.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.46.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.46.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.46.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.46.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.46.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.46.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.46.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.46.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.46.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.46.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.46.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.47.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.47.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.47.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.47.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.47.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.47.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.47.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.47.self_attn.indexer.index_qk_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.47.self_attn.k_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.47.self_attn.o_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.47.self_attn.q_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.47.self_attn.v_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.5.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.5.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.5.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.5.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.5.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.5.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.5.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.5.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.5.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.5.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.5.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.5.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.6.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.6.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.6.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.6.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.6.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.6.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.6.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.6.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.6.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.6.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.6.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.6.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.7.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.7.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.7.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.7.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.7.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.7.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.7.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.7.self_attn.indexer.index_qk_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.7.self_attn.k_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.7.self_attn.o_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.7.self_attn.q_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.7.self_attn.v_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.8.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.8.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.8.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.8.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.8.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.8.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.8.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.8.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.8.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.8.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.8.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.8.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.9.linear_attn.in_proj_a": { "group_size": 64, "bits": 8 }, "language_model.layers.9.linear_attn.in_proj_b": { "group_size": 64, "bits": 8 }, "language_model.layers.9.linear_attn.in_proj_qkv": { "group_size": 64, "bits": 8 }, "language_model.layers.9.linear_attn.in_proj_z": { "group_size": 64, "bits": 8 }, "language_model.layers.9.linear_attn.out_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.9.mlp.gate": { "group_size": 64, "bits": 8 }, "language_model.layers.9.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.9.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.9.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 8 }, "language_model.layers.9.mlp.switch_mlp.down_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.9.mlp.switch_mlp.gate_proj": { "group_size": 64, "bits": 4 }, "language_model.layers.9.mlp.switch_mlp.up_proj": { "group_size": 64, "bits": 4 }, "lm_head": { "group_size": 64, "bits": 8 }, "mtp.fc_embedding": { "group_size": 64, "bits": 4 }, "mtp.fc_hidden": { "group_size": 64, "bits": 4 }, "mtp.layers.0.mlp.experts.down_proj": { "group_size": 64, "bits": 4 }, "mtp.layers.0.mlp.experts.gate_up_proj": { "group_size": 64, "bits": 4 }, "mtp.layers.0.mlp.gate": { "group_size": 64, "bits": 4 }, "mtp.layers.0.mlp.shared_expert.down_proj": { "group_size": 64, "bits": 4 }, "mtp.layers.0.mlp.shared_expert.gate_proj": { "group_size": 64, "bits": 4 }, "mtp.layers.0.mlp.shared_expert.up_proj": { "group_size": 64, "bits": 4 }, "mtp.layers.0.self_attn.indexer.index_qk_proj": { "group_size": 64, "bits": 4 }, "mtp.layers.0.self_attn.k_proj": { "group_size": 64, "bits": 4 }, "mtp.layers.0.self_attn.o_proj": { "group_size": 64, "bits": 4 }, "mtp.layers.0.self_attn.q_proj": { "group_size": 64, "bits": 4 }, "mtp.layers.0.self_attn.v_proj": { "group_size": 64, "bits": 4 }, "visual.blocks.0.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.0.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.0.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.1.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.1.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.1.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.10.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.10.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.10.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.11.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.11.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.11.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.12.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.12.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.12.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.13.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.13.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.13.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.14.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.14.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.14.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.15.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.15.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.15.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.16.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.16.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.16.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.17.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.17.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.17.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.18.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.18.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.18.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.19.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.19.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.19.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.2.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.2.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.2.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.20.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.20.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.20.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.21.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.21.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.21.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.22.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.22.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.22.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.23.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.23.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.23.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.24.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.24.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.24.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.25.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.25.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.25.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.26.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.26.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.26.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.3.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.3.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.3.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.4.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.4.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.4.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.5.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.5.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.5.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.6.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.6.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.6.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.7.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.7.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.7.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.8.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.8.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.8.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.blocks.9.attn.proj": { "group_size": 64, "bits": 8 }, "visual.blocks.9.attn.qkv": { "group_size": 64, "bits": 8 }, "visual.blocks.9.mlp.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.merger.linear_fc1": { "group_size": 64, "bits": 8 }, "visual.merger.linear_fc2": { "group_size": 64, "bits": 8 } } }