README.md

---
license: mit
tags:
- lora
- training
- runpod
- ai-toolkit
---

# AI Trainer - RunPod Serverless

Single-endpoint multi-model LoRA training with all models cached in this repo.

## RunPod Deployment

**Set Model field to:** `Aloukik21/trainer`

This will cache all models (~240GB) for fast cold starts.

## Cached Models

| Model Key | Subfolder | Size |
|-----------|-----------|------|
| flux_dev | flux-dev/ | ~54GB |
| flux_schnell | flux-schnell/ | ~54GB |
| wan21_14b | wan21-14b/ | ~75GB |
| wan22_14b | wan22-14b/ | ~53GB |
| qwen_image | qwen-image/ | ~54GB |
| accuracy_recovery_adapters | accuracy_recovery_adapters/ | ~3GB |

## API Usage

### List Models
```json
{"input": {"action": "list_models"}}
```

### Train LoRA
```json
{
  "input": {
    "action": "train",
    "model": "flux_dev",
    "params": {
      "dataset_path": "/workspace/dataset",
      "output_path": "/workspace/output",
      "steps": 1000
    }
  }
}
```

### Cleanup (between different models)
```json
{"input": {"action": "cleanup"}}
```

## Environment Variables

- `HF_TOKEN`: HuggingFace token (required for some gated models)

## Auto-Cleanup

Handler automatically cleans up GPU memory when switching between different model types.