CoolFace
Modelpublic

BLIP3o/BLIP3o-Model-4B

sourceHugging Faceapache-2.0updated 1y agoView on Hugging Face
14likes609downloads
config.json85 linesDownload Raw Back to root
1{2  "architectures": [3    "blip3oQwenForCausalLM"4  ],5  "attention_dropout": 0.0,6  "bos_token_id": 151643,7  "eos_token_id": 151645,8  "freeze_mm_mlp_adapter": false,9  "gen_hidden_size": 1792,10  "gen_pooling": "early_pool2d_4",11  "gen_vision_tower": "eva-clip-E-14-plus",12  "hidden_act": "silu",13  "hidden_size": 2048,14  "image_aspect_ratio": "square",15  "image_token_id": 151655,16  "initializer_range": 0.02,17  "intermediate_size": 11008,18  "max_position_embeddings": 128000,19  "max_window_layers": 70,20  "mm_patch_merge_type": "flat",21  "mm_projector_lr": null,22  "mm_projector_type": "mlp2x_gelu",23  "mm_use_im_patch_token": false,24  "mm_use_im_start_end": false,25  "mm_vision_select_feature": "patch",26  "mm_vision_select_layer": -2,27  "model_type": "blip3o_qwen",28  "n_query": 64,29  "num_attention_heads": 16,30  "num_hidden_layers": 36,31  "num_key_value_heads": 2,32  "pad_token_id": 151643,33  "rms_norm_eps": 1e-06,34  "rope_scaling": {35    "mrope_section": [36      16,37      24,38      2439    ],40    "rope_type": "default",41    "type": "default"42  },43  "rope_theta": 1000000.0,44  "sliding_window": 32768,45  "tie_word_embeddings": true,46  "tokenizer_model_max_length": 512,47  "tokenizer_padding_side": "right",48  "torch_dtype": "bfloat16",49  "transformers_version": "4.51.0.dev0",50  "tune_mm_mlp_adapter": false,51  "use_cache": true,52  "use_mm_proj": true,53  "use_sliding_window": false,54  "video_token_id": 151656,55  "vision_config": {56    "depth": 32,57    "fullatt_block_indexes": [58      7,59      15,60      23,61      3162    ],63    "hidden_act": "silu",64    "hidden_size": 1280,65    "in_channels": 3,66    "in_chans": 3,67    "intermediate_size": 3420,68    "model_type": "qwen2_5_vl",69    "num_heads": 16,70    "out_hidden_size": 2048,71    "patch_size": 14,72    "spatial_merge_size": 2,73    "spatial_patch_size": 14,74    "temporal_patch_size": 2,75    "tokens_per_second": 2,76    "torch_dtype": "bfloat16",77    "window_size": 11278  },79  "vision_end_token_id": 151653,80  "vision_start_token_id": 151652,81  "vision_token_id": 151654,82  "vision_tower_pretrained": null,83  "vocab_size": 15166884}85