yanziang/InternVideo3-8B-Instruct

1{2    "architectures": [3        "InternVideo3ForConditionalGeneration"4    ],5    "auto_map": {6        "AutoModel": "modeling_internvideo3.InternVideo3Model",7        "AutoModelForCausalLM": "modeling_internvideo3.InternVideo3ForConditionalGeneration",8        "AutoConfig": "configuration_internvideo3.InternVideo3Config",9        "AutoProcessor": "processing_internvideo3.InternVideo3Processor"10    },11    "dtype": "bfloat16",12    "image_token_id": 151655,13    "model_type": "internvideo3",14    "text_config": {15        "attention_bias": true,16        "attention_dropout": 0.0,17        "bos_token_id": 151643,18        "dtype": "bfloat16",19        "eos_token_id": 151645,20        "head_dim": 128,21        "hidden_act": "silu",22        "hidden_size": 4096,23        "initializer_range": 0.02,24        "intermediate_size": 12288,25        "max_position_embeddings": 262144,26        "model_type": "internvideo3_text",27        "num_attention_heads": 32,28        "num_hidden_layers": 36,29        "num_key_value_heads": 32,30        "rms_norm_eps": 1e-06,31        "rope_scaling": {32            "mrope_interleaved": true,33            "mrope_section": [34                24,35                20,36                2037            ],38            "rope_type": "default"39        },40        "rope_theta": 5000000,41        "use_cache": true,42        "vocab_size": 151936,43        "qk_rope_head_dim": 128,44        "qk_nope_head_dim": 128,45        "v_head_dim": 128,46        "q_lora_rank": null,47        "rope_interleave": true,48        "kv_lora_rank_list": [49            896,50            896,51            896,52            896,53            896,54            896,55            896,56            896,57            896,58            896,59            896,60            896,61            896,62            896,63            896,64            896,65            896,66            896,67            896,68            896,69            896,70            896,71            896,72            896,73            896,74            896,75            896,76            896,77            896,78            896,79            896,80            896,81            896,82            896,83            896,84            89685        ],86        "qk_latent_layernorm": false87    },88    "tie_word_embeddings": false,89    "transformers_version": "4.57.3",90    "video_token_id": 151656,91    "vision_config": {92        "deepstack_visual_indexes": [],93        "depth": 27,94        "dtype": "bfloat16",95        "hidden_act": "gelu_pytorch_tanh",96        "hidden_size": 1152,97        "in_channels": 3,98        "initializer_range": 0.02,99        "intermediate_size": 4304,100        "model_type": "internvideo3",101        "num_heads": 16,102        "num_position_embeddings": 2304,103        "out_hidden_size": 4096,104        "patch_size": 16,105        "spatial_merge_size": 2,106        "temporal_patch_size": 2107    },108    "vision_end_token_id": 151653,109    "vision_start_token_id": 151652110}