DeOldify / docs /DEPLOYMENT_GUIDE.md
thookham's picture
Initial commit for Hugging Face sync (Clean History)
e9f9fd3
|
raw
history blame
2.47 kB
# DeOldify Deployment Guide
This guide covers the various ways you can deploy and run DeOldify.
## 🏠 Local Deployment
Running DeOldify locally gives you the best performance and privacy, provided you have the necessary hardware.
### Option 1: Conda (Recommended)
We recommend using [Anaconda](https://www.anaconda.com/) or [Miniconda](https://docs.conda.io/en/latest/miniconda.html) to manage dependencies.
1. **Clone the repository**:
```bash
git clone https://github.com/jantic/DeOldify.git
cd DeOldify
```
2. **Create the environment**:
* **For NVIDIA GPU**:
```bash
conda env create -f environment.yml
conda activate deoldify
```
* **For Intel GPU**:
```bash
conda env create -f environment_intel.yml
conda activate deoldify-intel
```
3. **Run Jupyter Lab**:
```bash
jupyter lab
```
### Option 2: Pip
If you prefer standard Python venv:
1. **Install Python 3.10+**.
2. **Install dependencies**:
```bash
pip install -r requirements.txt
```
*Note: You may need to install PyTorch manually first to ensure you get the correct CUDA version for your hardware.*
---
## ☁️ Cloud Deployment
### Google Colab
The easiest way to try DeOldify without installing anything.
* **Image Colorizer**: [Open in Colab](https://colab.research.google.com/github/jantic/DeOldify/blob/master/ImageColorizerColab.ipynb)
* **Video Colorizer**: [Open in Colab](https://colab.research.google.com/github/jantic/DeOldify/blob/master/VideoColorizerColab.ipynb)
**Notes**:
* Requires a Google account.
* Free tier GPUs are sufficient for images and short videos.
* Pro tier recommended for longer videos or faster rendering.
### Google Cloud Platform (Vertex AI)
*Coming Soon* - We are working on official scripts to deploy DeOldify as a scalable API endpoint on Vertex AI.
### Docker
*Coming Soon* - Official Docker images will be available to simplify deployment on any container orchestration platform.
---
## 📦 Model Weights
DeOldify relies on pre-trained model weights. These are downloaded automatically by the notebooks/scripts when you first run them.
* **Artistic**: `ColorizeArtistic_gen.pth`
* **Stable**: `ColorizeStable_gen.pth`
* **Video**: `ColorizeVideo_gen.pth`
If you are deploying in an air-gapped environment, you will need to download these weights manually and place them in the `models/` directory.