| { |
| "lr_mp": 5e-05, |
| "lr_vision_backbone": 0.0, |
| "lr_language_backbone": 5e-05, |
| "lr_full_decoder": 0.0, |
| "batch_size": 16, |
| "gradient_accumulation_steps": 8, |
| "max_grad_norm": 1.0, |
| "max_training_steps": 5000, |
| "stop_after_step": 5000, |
| "warmup_ratio": 0.03, |
| "stats_log_interval": 100, |
| "precision": "bf16", |
| "compile": false, |
| "log_train_samples": false, |
| "do_eval": true, |
| "eval_interval": 500, |
| "max_val_batches": 32, |
| "max_images_per_example": 1, |
| "max_sample_length": 6144, |
| "train_dataset_path": "patrickamadeus/the_cauldron", |
| "train_dataset_name": [ |
| "all" |
| ], |
| "train_split": "train", |
| "val_split": "validation", |
| "stream_dataset": true, |
| "dataloader_num_workers": 2, |
| "dataloader_pin_memory": true, |
| "enable_source_filter": true, |
| "allowed_dataset_sources": [ |
| "dvqa", |
| "tallyqa", |
| "figureqa", |
| "ocrvqa", |
| "vqav2", |
| "clevr", |
| "cocoqa", |
| "textvqa", |
| "visual7w", |
| "st_vqa", |
| "docvqa", |
| "aokvqa", |
| "chartqa" |
| ], |
| "relevance_min_rating": 1, |
| "image_correspondence_min_rating": 1, |
| "visual_dependency_min_rating": 1, |
| "formatting_min_rating": 1, |
| "wandb_entity": "HuggingFace", |
| "log_wandb": false, |
| "push_checkpoints_to_hub": true, |
| "save_training_state_to_hub": true, |
| "checkpoint_repo_pattern": "patrickamadeus/stackvlm-{i}", |
| "hf_private": false, |
| "push_final_model_to_hub": false, |
| "resume_from_vlm_checkpoint": false, |
| "resume_checkpoint_path": "patrickamadeus/stackvlm-1500", |
| "resume_start_step": 1500 |
| } |
