# Training Command Documentation

## Training Command
```bash
python scripts/gr00t_finetune.py \
    --dataset-path ./dataset/pickup-boxs26 \
    --num-gpus 1 \
    --output-dir ./checkpoints/pickup-boxs26 \
    --max-steps 10000 \
    --data-config so100_dualcam \
    --video-backend torchvision_av \
    --batch-size 8 \
    --dataloader-num-workers 1 \
    --dataloader-prefetch-factor 1
```

## Parameters Description

| Parameter | Value | Description |
|-----------|-------|-------------|
| `--dataset-path` | `./dataset/pickup-boxs26` | Path to the training dataset |
| `--num-gpus` | `1` | Number of GPUs to use for training |
| `--output-dir` | `./checkpoints/pickup-boxs26` | Directory to save model checkpoints |
| `--max-steps` | `10000` | Maximum number of training steps |
| `--data-config` | `so100_dualcam` | Data configuration for dual camera setup |
| `--video-backend` | `torchvision_av` | Video backend for data loading |
| `--batch-size` | `8` | Training batch size |
| `--dataloader-num-workers` | `1` | Number of data loader workers |
| `--dataloader-prefetch-factor` | `1` | Data loader prefetch factor |

## Training Notes
- This training uses the GR00T fine-tuning script
- Dataset: pickup-boxs26
- Single GPU training setup
- Dual camera configuration (so100_dualcam)
- Total training steps: 10,000
- Batch size: 8

## Date Created
Created on: $(date)