# Quick Reference - Hugging Face Integration

## Upload Model to HF Hub

### Option 1: Interactive (Easiest)
```bash
python publish_to_hf.py
# Enter token when prompted
# Enter repo_id when prompted (e.g., Enzo8930302/ByteDream)
```

### Option 2: Command Line
```bash
python publish_to_hf.py hf_xxxxxxxxxxxxx Enzo8930302/ByteDream
```

### Option 3: Python Code
```python
from bytedream import ByteDreamGenerator

generator = ByteDreamGenerator(model_path="./models/bytedream")
generator.push_to_hub(
    repo_id="Enzo8930302/ByteDream",
    token="hf_xxxxx"
)
```

---

## Load Model from HF Hub

### Python - Generator
```python
from bytedream import ByteDreamGenerator

generator = ByteDreamGenerator(hf_repo_id="Enzo8930302/ByteDream")
image = generator.generate("Your prompt")
image.save("output.png")
```

### Python - Pipeline
```python
from bytedream.pipeline import ByteDreamPipeline

pipeline = ByteDreamPipeline.from_pretrained("Enzo8930302/ByteDream")
result = pipeline("Your prompt")
result[0].save("output.png")
```

### Command Line
```bash
python infer.py --prompt "Dragon flying" --hf_repo "Enzo8930302/ByteDream" --output dragon.png
```

### Web Interface
```bash
# Set environment variable
export HF_REPO_ID=Enzo8930302/ByteDream

# Run app
python app.py
```

---

## Deploy to Hugging Face Spaces

1. **Create Space**: https://huggingface.co/spaces → Create new Space
2. **Choose**: Gradio SDK + CPU Basic
3. **Push files**:
   ```bash
   git clone https://huggingface.co/spaces/YOUR_USERNAME/SPACE_NAME
   cp -r ../Byte\ Dream/* SPACE_NAME/
   cd SPACE_NAME
   git add .
   git commit -m "Deploy Byte Dream"
   git push
   ```
4. **Set env var**: In Space settings → Add `HF_REPO_ID=Enzo8930302/ByteDream`
5. **Done!** Available at: `https://huggingface.co/spaces/YOUR_USERNAME/SPACE_NAME`

---

## Common Commands

### Get HF Token
- Go to: https://huggingface.co/settings/tokens
- Click "New token"
- Copy token (starts with `hf_`)

### Check Model Exists
```bash
# List files in your HF repo
huggingface-cli ls Enzo8930302/ByteDream
```

### Download Model Manually
```python
from huggingface_hub import snapshot_download

snapshot_download(repo_id="Enzo8930302/ByteDream")
```

### Test Installation
```bash
python -c "from huggingface_hub import login; print('OK')"
```

---

## Environment Variables

```bash
# Load model from HF in app.py
export HF_REPO_ID=Enzo8930302/ByteDream

# Custom model path
export MODEL_PATH=./models/bytedream

# Set HF cache directory
export HF_HOME=~/.cache/huggingface
```

---

## Example Prompts

```python
prompts = [
    "A beautiful sunset over mountains, digital art",
    "Cyberpunk city at night, neon lights, futuristic",
    "Fantasy dragon breathing fire, dramatic lighting",
    "Peaceful cottage in meadow, flowers, sunny day",
    "Underwater coral reef, tropical fish, sunlight",
    "Abstract geometric art, colorful shapes, modern",
]

from bytedream import ByteDreamGenerator
generator = ByteDreamGenerator(hf_repo_id="Enzo8930302/ByteDream")

for i, prompt in enumerate(prompts):
    image = generator.generate(prompt, num_inference_steps=50)
    image.save(f"image_{i}.png")
```

---

## Troubleshooting Quick Fixes

**Error: Repository not found**
```python
# Make repo public or provide token
from huggingface_hub import login
login(token="hf_xxxxx")
```

**Error: Out of memory**
```python
# Reduce size
generator.generate(width=256, height=256)
# Fewer steps
generator.generate(num_inference_steps=20)
```

**Error: Model not trained**
```bash
# Train first
python train.py
# Or download from HF
python infer.py --hf_repo username/model --prompt "test"
```

---

## File Structure on HF

```
username/ByteDream/
├── unet/
│   └── pytorch_model.bin
├── vae/
│   └── pytorch_model.bin
├── scheduler/
│   └── scheduler_config.json
├── model_index.json
├── config.yaml
└── README.md
```

---

## API Methods Summary

### ByteDreamGenerator
```python
# Initialize with HF model
ByteDreamGenerator(hf_repo_id="username/repo")

# Upload to HF
generator.push_to_hub(repo_id="username/repo", token="hf_xxx")

# Save locally
generator.save_pretrained("./models/bytedream")
```

### ByteDreamPipeline
```python
# Load from HF
pipeline = ByteDreamPipeline.from_pretrained("username/repo")

# Load from local
pipeline = ByteDreamPipeline.from_pretrained("./models/bytedream")

# Save to HF
pipeline.save_pretrained("./models/bytedream")
```

---

## Complete Workflow Example

```python
# 1. Train (if needed)
# python train.py

# 2. Load model
from bytedream import ByteDreamGenerator
generator = ByteDreamGenerator(model_path="./models/bytedream")

# 3. Test generation
image = generator.generate("Test prompt")
image.save("test.png")

# 4. Upload to HF
generator.push_to_hub(
    repo_id="Enzo8930302/ByteDream",
    token="hf_xxxxxxxx"
)

print("✓ Uploaded to HF!")

# 5. Use from HF (new session)
generator2 = ByteDreamGenerator(hf_repo_id="Enzo8930302/ByteDream")
image2 = generator2.generate("Amazing prompt")
image2.save("amazing.png")
```

---

**More Info**: See `HF_INTEGRATION_GUIDE.md` for detailed documentation