| { |
| "NUM_FUTURE_MAP_MASK_TOKEN": 9, |
| "NUM_SEMANTIC_MAP_MASK_TOKEN": 9, |
| "PATCH_SIZE": 16, |
| "SEMANTIC_MAP_MASK_DECODER_hidden_dim": 768, |
| "_name_or_path": "./Model/LLaMA-UAV/../../model_zoo/LLM/vicuna-7b-v1.5", |
| "_temp_global_step": 3333, |
| "_temp_labels": null, |
| "architectures": [ |
| "LlamaForCausalLM" |
| ], |
| "bert_type": "qformer_pretrain_freeze", |
| "bos_token_id": 1, |
| "compress_type": "mean", |
| "eos_token_id": 2, |
| "freeze_mm_mlp_adapter": false, |
| "hidden_act": "silu", |
| "hidden_size": 4096, |
| "history_token_num": 17, |
| "image_aspect_ratio": "square", |
| "image_grid_pinpoints": null, |
| "image_processor": "./Model/LLaMA-UAV/llamavid/processor/clip-patch14-224", |
| "initializer_range": 0.02, |
| "intermediate_size": 11008, |
| "max_position_embeddings": 4096, |
| "max_token": 2048, |
| "mm_hidden_size": 1408, |
| "mm_projector_type": "mlp2x_gelu", |
| "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": "./Model/LLaMA-UAV/../../model_zoo/LAVIS/eva_vit_g.pth", |
| "model_type": "llama", |
| "next_waypoint_mask_class_weights": [ |
| 0.1, |
| 1.0 |
| ], |
| "next_waypoint_mask_loss_scale": 0.5, |
| "num_attention_heads": 32, |
| "num_hidden_layers": 32, |
| "num_key_value_heads": 32, |
| "num_query": 32, |
| "num_semantic_map_mask_classes": 20, |
| "pad_token_id": 0, |
| "pred_future_map_mask": true, |
| "pred_map_mask": true, |
| "pred_next_waypoint_mask": true, |
| "pred_waypoint": true, |
| "pretraining_tp": 1, |
| "random_zero_map_mask_token_embed": false, |
| "random_zero_map_mask_token_embed_prob": 0.3, |
| "rms_norm_eps": 1e-05, |
| "rope_scaling": null, |
| "semantic_map_h": 224, |
| "semantic_map_mask_class_weights": [ |
| 0.5, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0, |
| 1.0 |
| ], |
| "semantic_map_mask_decoder_depth": 2, |
| "semantic_map_mask_loss_scale": 0.5, |
| "semantic_map_w": 224, |
| "tie_word_embeddings": false, |
| "torch_dtype": "float16", |
| "trajectory_direction_constraint_loss_scale": 0.001, |
| "transformers_version": "4.31.0", |
| "tune_mm_mlp_adapter": true, |
| "use_cache": true, |
| "use_custom_causal_mask": true, |
| "use_history_images": false, |
| "use_mm_proj": true, |
| "use_trajectory_direction_constraint": true, |
| "vocab_size": 32002 |
| } |
|
|
