config_used.json

{
  "_comment": "Projection-only alignment on CC3M spatial-only subset",
  "model_name": "google/gemma-3-4b-it",
  "cosmos_model_name": "nvidia/Cosmos-Predict2-2B-Video2World",
  "enable_world": true,
  "num_world_steps": 0,
  "world_projection_mode": "channel",
  "projection_architecture": "mlp",
  "freeze_gemma_vision": true,
  "freeze_gemma_language": true,
  "freeze_cosmos_vae": true,
  "learning_rate": 0.0001,
  "batch_size": 2,
  "gradient_accumulation_steps": 1,
  "num_epochs": 1,
  "warmup_steps": 100,
  "weight_decay": 0.01,
  "max_grad_norm": 1.0,
  "use_gradient_checkpointing": false,
  "mixed_precision": "bf16",
  "output_dir": "/storage/ice1/8/2/ajin37/checkpoints/theworld-cc3m-align-only-spatial-it",
  "save_steps": 1000,
  "save_total_limit": 3,
  "resume_from_checkpoint": null,
  "eval_steps": 2000,
  "eval_batch_size": 8,
  "do_eval": false,
  "logging_steps": 100,
  "log_to_wandb": false,
  "wandb_project": "theworld-cc3m",
  "wandb_run_name": "cc3m-align-only-spatial",
  "log_to_tensorboard": true,
  "max_seq_length": 2048,
  "num_workers": 4,
  "dataset_name": "cc3m_align",
  "train_dataset_path": "/storage/ice1/8/2/ajin37/data/cc3m-align/cc3m_spatial_only.jsonl",
  "eval_dataset_path": null,
  "num_samples": null,
  "streaming": false,
  "image_folder": "/storage/ice1/8/2/ajin37/data/cc3m-align/images",
  "draw_bboxes": false,
  "hf_token": null,
  "push_to_hub": true,
  "hub_model_id": "alwinjin8/theworld-cc3m-align-only-spatial",
  "hub_strategy": "every_save",
  "hub_private_repo": false
}