| { |
| "_name_or_path": "checkpoints/VideoGPTPlus-Phi3-SAM2-8frame-tunevlproj-epoch29", |
| "architectures": [ |
| "SPARROWForCausalLM" |
| ], |
| "attention_bias": false, |
| "attention_dropout": 0.0, |
| "auto_map": { |
| "AutoConfig": "configuration_phi3.Phi3Config", |
| "AutoModelForCausalLM": "modeling_phi3.Phi3ForCausalLM" |
| }, |
| "bos_token_id": 1, |
| "box_token_idx": 32012, |
| "config": { |
| "freeze": false, |
| "mm_tune": true, |
| "vision_tune": false |
| }, |
| "embd_pdrop": 0.0, |
| "eos_token_id": 32000, |
| "freeze_mm_mlp_adapter": false, |
| "hidden_act": "silu", |
| "hidden_size": 3072, |
| "image_aspect_ratio": "square", |
| "image_grid_pinpoints": null, |
| "image_mm_hidden_size": 1024, |
| "image_mm_projector_type": "mlp2x_gelu", |
| "image_mm_vision_tower": "openai/clip-vit-large-patch14-336", |
| "image_vision_tower": "openai/clip-vit-large-patch14-336", |
| "initializer_range": 0.02, |
| "intermediate_size": 8192, |
| "mask_decoder_itm": false, |
| "max_position_embeddings": 4096, |
| "mlp_bias": false, |
| "mm_hidden_size": 1408, |
| "mm_projector_lr": 2e-05, |
| "mm_projector_type": "mlp2x_gelu", |
| "mm_use_box_start_end": false, |
| "mm_use_im_patch_token": false, |
| "mm_use_im_start_end": false, |
| "mm_vision_select_feature": "patch", |
| "mm_vision_select_layer": -2, |
| "mm_vision_tower": "OpenGVLab/InternVideo2-Stage2_1B-224p-f4/InternVideo2-stage2_1b-224p-f4.pt", |
| "model_type": "VideoGPT+", |
| "num_attention_heads": 32, |
| "num_hidden_layers": 32, |
| "num_key_value_heads": 32, |
| "original_max_position_embeddings": 4096, |
| "out_dim": 256, |
| "pad_token_id": 0, |
| "pretrain_image_mm_mlp_adapter": null, |
| "pretrain_mm_mlp_adapter": null, |
| "pretraining_tp": 1, |
| "resid_pdrop": 0.0, |
| "rms_norm_eps": 1e-05, |
| "rope_scaling": null, |
| "rope_theta": 10000.0, |
| "s2": false, |
| "s2_scales": "336,672,1008", |
| "seg_token_idx": 32011, |
| "sliding_window": 2048, |
| "tie_word_embeddings": false, |
| "torch_dtype": "float16", |
| "train_mask_decoder": false, |
| "transformers_version": "4.41.0", |
| "tune_mm_mlp_adapter": false, |
| "use_cache": false, |
| "use_mm_proj": true, |
| "use_sam2": true, |
| "vision_tower": "OpenGVLab/InternVideo2-Stage2_1B-224p-f4/InternVideo2-stage2_1b-224p-f4.pt", |
| "vocab_size": 32013 |
| } |
|
|
