ByteDance/Sa2VA-InternVL3-2B

1{2  "architectures": [3    "Sa2VAChatModel"4  ],5  "auto_map": {6    "AutoConfig": "configuration_sa2va_chat.Sa2VAChatConfig",7    "AutoModel": "modeling_sa2va_chat.Sa2VAChatModel",8    "AutoModelForCausalLM": "modeling_sa2va_chat.Sa2VAChatModel"9  },10  "downsample_ratio": 0.5,11  "dtype": "bfloat16",12  "dynamic_image_size": true,13  "force_image_size": 448,14  "hidden_size": 1536,15  "image_fold": null,16  "llm_config": {17    "_attn_implementation_autoset": true,18    "_name_or_path": "./pretrained/Qwen2.5-32B-Instruct",19    "architectures": [20      "Qwen2ForCausalLM"21    ],22    "attention_dropout": 0.0,23    "bos_token_id": 151643,24    "dtype": "bfloat16",25    "eos_token_id": 151643,26    "hidden_act": "silu",27    "hidden_size": 1536,28    "initializer_range": 0.02,29    "intermediate_size": 8960,30    "layer_types": [31      "full_attention",32      "full_attention",33      "full_attention",34      "full_attention",35      "full_attention",36      "full_attention",37      "full_attention",38      "full_attention",39      "full_attention",40      "full_attention",41      "full_attention",42      "full_attention",43      "full_attention",44      "full_attention",45      "full_attention",46      "full_attention",47      "full_attention",48      "full_attention",49      "full_attention",50      "full_attention",51      "full_attention",52      "full_attention",53      "full_attention",54      "full_attention",55      "full_attention",56      "full_attention",57      "full_attention",58      "full_attention"59    ],60    "max_position_embeddings": 32768,61    "max_window_layers": 70,62    "model_type": "qwen2",63    "moe_config": null,64    "num_attention_heads": 12,65    "num_hidden_layers": 28,66    "num_key_value_heads": 2,67    "rms_norm_eps": 1e-06,68    "rope_scaling": {69      "factor": 2.0,70      "rope_type": "dynamic",71      "type": "dynamic"72    },73    "rope_theta": 1000000.0,74    "sliding_window": null,75    "use_bfloat16": true,76    "use_cache": false,77    "use_sliding_window": false,78    "vocab_size": 15167979  },80  "max_dynamic_patch": 12,81  "min_dynamic_patch": 1,82  "model_type": "sa2va_chat",83  "output_attentions": false,84  "pad2square": false,85  "ps_version": "v2",86  "select_layer": -1,87  "system_message": null,88  "template": "qwen_chat",89  "tie_word_embeddings": false,90  "transformers_version": null,91  "use_backbone_lora": 0,92  "use_llm_lora": 0,93  "use_thumbnail": true,94  "vision_config": {95    "_attn_implementation_autoset": true,96    "_name_or_path": "OpenGVLab/InternViT-6B-448px-V1-5",97    "architectures": [98      "InternVisionModel"99    ],100    "attention_dropout": 0.0,101    "auto_map": {102      "AutoConfig": "configuration_intern_vit.InternVisionConfig",103      "AutoModel": "modeling_intern_vit.InternVisionModel"104    },105    "capacity_factor": 1.2,106    "drop_path_rate": 0.1,107    "dropout": 0.0,108    "dtype": "bfloat16",109    "eval_capacity_factor": 1.4,110    "hidden_act": "gelu",111    "hidden_size": 1024,112    "image_size": 448,113    "initializer_factor": 0.1,114    "initializer_range": 1e-10,115    "intermediate_size": 4096,116    "laux_allreduce": "all_nodes",117    "layer_norm_eps": 1e-06,118    "model_type": "intern_vit_6b",119    "moe_coeff_ratio": 0.5,120    "moe_intermediate_size": 768,121    "moe_output_scale": 4.0,122    "noisy_gate_policy": "RSample_before",123    "norm_type": "layer_norm",124    "num_attention_heads": 16,125    "num_channels": 3,126    "num_experts": 8,127    "num_hidden_layers": 24,128    "num_routed_experts": 4,129    "num_shared_experts": 4,130    "patch_size": 14,131    "qk_normalization": false,132    "qkv_bias": true,133    "shared_expert_intermediate_size": 3072,134    "use_bfloat16": true,135    "use_flash_attn": true,136    "use_moe": false,137    "use_residual": true,138    "use_rts": false,139    "use_weighted_residual": false140  }141}142