40 lines
2.0 KiB
Markdown
Vendored
40 lines
2.0 KiB
Markdown
Vendored
# min(DALL·E)
|
|
|
|
[](https://colab.research.google.com/github/kuprel/min-dalle/blob/main/min_dalle.ipynb)
|
|
[](https://replicate.com/kuprel/min-dalle)
|
|
|
|
This is a minimal implementation of Boris Dayma's [DALL·E Mini](https://github.com/borisdayma/dalle-mini). It has been stripped to the bare essentials necessary for doing inference, and converted to PyTorch. To run the torch model, the only third party dependencies are numpy and torch. Flax is used to convert the weights (which are saved with `torch.save` the first time the model is loaded), and wandb is only used to download the models.
|
|
|
|
It currently takes **7.4 seconds** to generate an image with DALL·E Mega with PyTorch on a standard GPU runtime in Colab
|
|
|
|
### Setup
|
|
|
|
Run `sh setup.sh` to install dependencies and download pretrained models. The models can also be downloaded manually here:
|
|
[VQGan](https://huggingface.co/dalle-mini/vqgan_imagenet_f16_16384),
|
|
[DALL·E Mini](https://wandb.ai/dalle-mini/dalle-mini/artifacts/DalleBart_model/mini-1/v0/files),
|
|
[DALL·E Mega](https://wandb.ai/dalle-mini/dalle-mini/artifacts/DalleBart_model/mega-1-fp16/v14/files)
|
|
|
|
### Usage
|
|
|
|
Use the python script `image_from_text.py` to generate images from the command line. Note: the command line script loads the models and parameters each time. To load a model once and generate multiple times, initialize either `MinDalleTorch` or `MinDalleFlax`, then call `generate_image` with some text and a seed. See the colab for an example.
|
|
|
|
### Examples
|
|
|
|
```
|
|
python image_from_text.py --text='artificial intelligence' --torch
|
|
```
|
|
|
|
|
|
|
|
```
|
|
python image_from_text.py --text='a comfy chair that looks like an avocado' --torch --mega --seed=10
|
|
```
|
|
|
|
|
|
|
|
```
|
|
python image_from_text.py --text='court sketch of godzilla on trial' --mega --seed=100
|
|
```
|
|
|
|
|