{ "architectures": [ "KimiVLForConditionalGeneration" ], "auto_map": { "AutoConfig": "configuration_kimi_vl.KimiVLConfig", "AutoModel": "modeling_kimi_vl.KimiVLForConditionalGeneration", "AutoModelForCausalLM": "modeling_kimi_vl.KimiVLForConditionalGeneration" }, "ignore_index": -100, "media_placeholder_token_id": 163605, "model_type": "kimi_vl", "pad_token_id": 0, "quantization_config": { "config_groups": { "group_0": { "input_activations": { "actorder": null, "block_structure": null, "dynamic": true, "group_size": null, "num_bits": 8, "observer": null, "observer_kwargs": {}, "strategy": "token", "symmetric": true, "type": "float" }, "output_activations": null, "targets": [ "Linear" ], "weights": { "actorder": null, "block_structure": null, "dynamic": false, "group_size": null, "num_bits": 8, "observer": "minmax", "observer_kwargs": {}, "strategy": "channel", "symmetric": true, "type": "float" } } }, "format": "float-quantized", "global_compression_ratio": null, "ignore": [ "vision_tower.encoder.blocks.0.mlp.fc0", "vision_tower.encoder.blocks.0.mlp.fc1", "vision_tower.encoder.blocks.0.wqkv", "vision_tower.encoder.blocks.0.wo", "vision_tower.encoder.blocks.1.mlp.fc0", "vision_tower.encoder.blocks.1.mlp.fc1", "vision_tower.encoder.blocks.1.wqkv", "vision_tower.encoder.blocks.1.wo", "vision_tower.encoder.blocks.2.mlp.fc0", "vision_tower.encoder.blocks.2.mlp.fc1", "vision_tower.encoder.blocks.2.wqkv", "vision_tower.encoder.blocks.2.wo", "vision_tower.encoder.blocks.3.mlp.fc0", "vision_tower.encoder.blocks.3.mlp.fc1", "vision_tower.encoder.blocks.3.wqkv", "vision_tower.encoder.blocks.3.wo", "vision_tower.encoder.blocks.4.mlp.fc0", "vision_tower.encoder.blocks.4.mlp.fc1", "vision_tower.encoder.blocks.4.wqkv", "vision_tower.encoder.blocks.4.wo", "vision_tower.encoder.blocks.5.mlp.fc0", "vision_tower.encoder.blocks.5.mlp.fc1", "vision_tower.encoder.blocks.5.wqkv", "vision_tower.encoder.blocks.5.wo", "vision_tower.encoder.blocks.6.mlp.fc0", "vision_tower.encoder.blocks.6.mlp.fc1", "vision_tower.encoder.blocks.6.wqkv", "vision_tower.encoder.blocks.6.wo", "vision_tower.encoder.blocks.7.mlp.fc0", "vision_tower.encoder.blocks.7.mlp.fc1", "vision_tower.encoder.blocks.7.wqkv", "vision_tower.encoder.blocks.7.wo", "vision_tower.encoder.blocks.8.mlp.fc0", "vision_tower.encoder.blocks.8.mlp.fc1", "vision_tower.encoder.blocks.8.wqkv", "vision_tower.encoder.blocks.8.wo", "vision_tower.encoder.blocks.9.mlp.fc0", "vision_tower.encoder.blocks.9.mlp.fc1", "vision_tower.encoder.blocks.9.wqkv", "vision_tower.encoder.blocks.9.wo", "vision_tower.encoder.blocks.10.mlp.fc0", "vision_tower.encoder.blocks.10.mlp.fc1", "vision_tower.encoder.blocks.10.wqkv", "vision_tower.encoder.blocks.10.wo", "vision_tower.encoder.blocks.11.mlp.fc0", "vision_tower.encoder.blocks.11.mlp.fc1", "vision_tower.encoder.blocks.11.wqkv", "vision_tower.encoder.blocks.11.wo", "vision_tower.encoder.blocks.12.mlp.fc0", "vision_tower.encoder.blocks.12.mlp.fc1", "vision_tower.encoder.blocks.12.wqkv", "vision_tower.encoder.blocks.12.wo", "vision_tower.encoder.blocks.13.mlp.fc0", "vision_tower.encoder.blocks.13.mlp.fc1", "vision_tower.encoder.blocks.13.wqkv", "vision_tower.encoder.blocks.13.wo", "vision_tower.encoder.blocks.14.mlp.fc0", "vision_tower.encoder.blocks.14.mlp.fc1", "vision_tower.encoder.blocks.14.wqkv", "vision_tower.encoder.blocks.14.wo", "vision_tower.encoder.blocks.15.mlp.fc0", "vision_tower.encoder.blocks.15.mlp.fc1", "vision_tower.encoder.blocks.15.wqkv", "vision_tower.encoder.blocks.15.wo", "vision_tower.encoder.blocks.16.mlp.fc0", "vision_tower.encoder.blocks.16.mlp.fc1", "vision_tower.encoder.blocks.16.wqkv", "vision_tower.encoder.blocks.16.wo", "vision_tower.encoder.blocks.17.mlp.fc0", "vision_tower.encoder.blocks.17.mlp.fc1", "vision_tower.encoder.blocks.17.wqkv", "vision_tower.encoder.blocks.17.wo", "vision_tower.encoder.blocks.18.mlp.fc0", "vision_tower.encoder.blocks.18.mlp.fc1", "vision_tower.encoder.blocks.18.wqkv", "vision_tower.encoder.blocks.18.wo", "vision_tower.encoder.blocks.19.mlp.fc0", "vision_tower.encoder.blocks.19.mlp.fc1", "vision_tower.encoder.blocks.19.wqkv", "vision_tower.encoder.blocks.19.wo", "vision_tower.encoder.blocks.20.mlp.fc0", "vision_tower.encoder.blocks.20.mlp.fc1", "vision_tower.encoder.blocks.20.wqkv", "vision_tower.encoder.blocks.20.wo", "vision_tower.encoder.blocks.21.mlp.fc0", "vision_tower.encoder.blocks.21.mlp.fc1", "vision_tower.encoder.blocks.21.wqkv", "vision_tower.encoder.blocks.21.wo", "vision_tower.encoder.blocks.22.mlp.fc0", "vision_tower.encoder.blocks.22.mlp.fc1", "vision_tower.encoder.blocks.22.wqkv", "vision_tower.encoder.blocks.22.wo", "vision_tower.encoder.blocks.23.mlp.fc0", "vision_tower.encoder.blocks.23.mlp.fc1", "vision_tower.encoder.blocks.23.wqkv", "vision_tower.encoder.blocks.23.wo", "vision_tower.encoder.blocks.24.mlp.fc0", "vision_tower.encoder.blocks.24.mlp.fc1", "vision_tower.encoder.blocks.24.wqkv", "vision_tower.encoder.blocks.24.wo", "vision_tower.encoder.blocks.25.mlp.fc0", "vision_tower.encoder.blocks.25.mlp.fc1", "vision_tower.encoder.blocks.25.wqkv", "vision_tower.encoder.blocks.25.wo", "vision_tower.encoder.blocks.26.mlp.fc0", "vision_tower.encoder.blocks.26.mlp.fc1", "vision_tower.encoder.blocks.26.wqkv", "vision_tower.encoder.blocks.26.wo", "multi_modal_projector.linear_1", "multi_modal_projector.linear_2", "language_model.lm_head" ], "kv_cache_scheme": null, "quant_method": "compressed-tensors", "quantization_status": "compressed" }, "text_config": { "attention_bias": false, "attention_dropout": 0.0, "aux_loss_alpha": 0.001, "bos_token_id": 163584, "eos_token_id": 163585, "ep_size": 1, "first_k_dense_replace": 1, "hidden_act": "silu", "hidden_size": 2048, "initializer_range": 0.02, "intermediate_size": 11264, "kv_lora_rank": 512, "max_position_embeddings": 131072, "model_type": "deepseek_v3", "moe_intermediate_size": 1408, "moe_layer_freq": 1, "n_group": 1, "n_routed_experts": 64, "n_shared_experts": 2, "norm_topk_prob": true, "num_attention_heads": 16, "num_experts_per_tok": 6, "num_hidden_layers": 27, "num_key_value_heads": 16, "num_nextn_predict_layers": 1, "pad_token_id": 163839, "pretraining_tp": 1, "q_lora_rank": null, "qk_nope_head_dim": 128, "qk_rope_head_dim": 64, "rms_norm_eps": 1e-05, "rope_scaling": null, "rope_theta": 800000.0, "routed_scaling_factor": 2.446, "scoring_func": "sigmoid", "seq_aux": true, "topk_group": 1, "topk_method": "noaux_tc", "torch_dtype": "bfloat16", "use_cache": true, "v_head_dim": 128, "vocab_size": 163840 }, "tie_word_embeddings": false, "torch_dtype": "bfloat16", "transformers_version": "4.53.2", "vision_config": { "hidden_size": 1152, "init_pos_emb_height": 64, "init_pos_emb_width": 64, "intermediate_size": 4304, "merge_kernel_size": [ 2, 2 ], "model_type": "moonvit", "num_attention_heads": 16, "num_hidden_layers": 27, "patch_size": 14, "torch_dtype": "bfloat16" }, "vocab_size": 163840 }