min-dalle-test/README.md

# min(DALL·E)

[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/kuprel/min-dalle/blob/main/min_dalle.ipynb) &nbsp;
[![Replicate](https://replicate.com/kuprel/min-dalle/badge)](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 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
```
![Alien](examples/artificial_intelligence.png)


```
python image_from_text.py --text='a comfy chair that looks like an avocado' --torch --mega --seed=10
```
![Avocado Armchair](examples/avocado_armchair.png)


```
python image_from_text.py --text='court sketch of godzilla on trial' --torch --mega --seed=40
```

![Godzilla Trial](examples/godzilla_trial.png)
simplified 2022-06-27 19:46:04 +00:00			`# min(DALL·E)`
examples 2022-06-27 17:38:35 +00:00
Update README.md 2022-06-29 19:24:09 +00:00			`[![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/kuprel/min-dalle/blob/main/min_dalle.ipynb)  `
			`[![Replicate](https://replicate.com/kuprel/min-dalle/badge)](https://replicate.com/kuprel/min-dalle)`
updated readme 2022-06-27 21:26:05 +00:00
mega works with latest flax version 0.5.2 now, removing 0.4.2 pin 2022-07-01 06:58:43 +00:00			`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 the first time the model is loaded), and wandb is only used to download the models.`
simplified attention and keys_values state resulted in decrease in inference time to 7.3 seconds (from ~10 seconds) 2022-06-29 17:56:29 +00:00
update readme 2022-07-01 01:05:02 +00:00			`It currently takes 7.4 seconds to generate an image with DALL·E Mega with PyTorch on a standard GPU runtime in Colab`
readme 2022-06-27 17:51:48 +00:00
license and cleanup 2022-06-27 18:34:10 +00:00			`### Setup`
readme 2022-06-27 17:51:48 +00:00
readme wording 2022-06-29 15:01:46 +00:00			Run `sh setup.sh` to install dependencies and download pretrained models. The models can also be downloaded manually here:
updated colab 2022-06-28 00:59:39 +00:00			`[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)`
updated readme 2022-06-27 21:26:05 +00:00
updated colab 2022-06-28 00:59:39 +00:00			`### Usage`
readme formatting 2022-06-27 21:28:40 +00:00
readme wording 2022-06-29 15:00:38 +00:00			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.
readme wording 2022-06-29 14:47:08 +00:00
readme wording 2022-06-29 14:54:01 +00:00			`### Examples`
examples 2022-06-27 17:38:35 +00:00
			```
works with latest flax version 0.5.2, updated requirements.txt 2022-06-29 15:46:19 +00:00			`python image_from_text.py --text='artificial intelligence' --torch`
simplified 2022-06-27 19:46:04 +00:00			```
updated readme with torch examples 2022-06-29 14:43:46 +00:00			`![Alien](examples/artificial_intelligence.png)`
simplified 2022-06-27 19:46:04 +00:00
fixed an issue with argument parser 2022-06-27 20:49:42 +00:00
simplified 2022-06-27 19:46:04 +00:00			```
updated readme with torch examples 2022-06-29 14:43:46 +00:00			`python image_from_text.py --text='a comfy chair that looks like an avocado' --torch --mega --seed=10`
examples 2022-06-27 17:38:35 +00:00			```
			`![Avocado Armchair](examples/avocado_armchair.png)`


			```
mega works with latest flax version 0.5.2 now, removing 0.4.2 pin 2022-07-01 06:58:43 +00:00			`python image_from_text.py --text='court sketch of godzilla on trial' --torch --mega --seed=40`
examples 2022-06-27 17:38:35 +00:00			```

Update README.md 2022-06-29 19:24:09 +00:00			`![Godzilla Trial](examples/godzilla_trial.png)`