# min(DALL·E)
[](https://colab.research.google.com/github/kuprel/min-dalle/blob/main/min_dalle.ipynb)
[](https://replicate.com/kuprel/min-dalle)
[](https://discord.com/channels/823813159592001537/912729332311556136)
This is a fast, minimal port of Boris Dayma's [DALL·E Mini](https://github.com/borisdayma/dalle-mini) (with mega weights). It has been stripped down for inference and converted to PyTorch. The only third party dependencies are numpy, requests, pillow and torch.
To generate a 4x4 grid of DALL·E Mega images it takes:
- 89 sec with a T4 in Colab
- 48 sec with a P100 in Colab
- 13 sec with an A100 on Replicate
Here's a more detailed breakdown of total inference time vs number of generated images on an A100:
Credit to @technobird22 and his [NeoGen](https://github.com/technobird22/NeoGen) discord bot for the graph
The flax model and code for converting it to torch can be found [here](https://github.com/kuprel/min-dalle-flax).
## Install
```bash
$ pip install min-dalle
```
## Usage
Load the model parameters once and reuse the model to generate multiple images.
```python
from min_dalle import MinDalle
model = MinDalle(
models_root='./pretrained',
dtype=torch.float32,
is_mega=True,
is_reusable=True
)
```
The required models will be downloaded to `models_root` if they are not already there. Set the `dtype` to `torch.float16` to save GPU memory. If you have an Ampere architecture GPU you can use `torch.bfloat16`. Once everything has finished initializing, call `generate_image` with some text as many times as you want. Use a positive `seed` for reproducible results. Higher values for `log2_supercondition_factor` result in better agreement with the text but a narrower variety of generated images. Every image token is sampled from the top-$k$ most probable tokens.
```python
image = model.generate_image(
text='Nuclear explosion broccoli',
seed=-1,
grid_size=4,
log2_k=6,
log2_supercondition_factor=5,
is_verbose=False
)
display(image)
```
credit: https://twitter.com/hardmaru/status/1544354119527596034
### Saving Individual Images
The images can also be generated as a `FloatTensor` in case you want to process them manually.
```python
images = model.generate_images(
text='Nuclear explosion broccoli',
seed=-1,
image_count=7,
log2_k=6,
log2_supercondition_factor=5,
is_verbose=False
)
```
To get an image into PIL format you will have to first move the images to the CPU and convert the tensor to a numpy array.
```python
images = images.to('cpu').numpy()
```
Then image $i$ can be coverted to a PIL.Image and saved
```python
image = Image.fromarray(images[i])
image.save('image_{}.png'.format(i))
```
### Interactive
If the model is being used interactively (e.g. in a notebook) `generate_image_stream` can be used to generate a stream of images as the model is decoding. The detokenizer adds a slight delay for each image. Setting `log2_mid_count` to 3 results in a total of `2 ** 3 = 8` generated images. The only valid values for `log2_mid_count` are 0, 1, 2, 3, and 4. This is implemented in the colab.
```python
image_stream = model.generate_image_stream(
text='Dali painting of WALL·E',
seed=-1,
grid_size=3,
log2_mid_count=3,
log2_k=6,
log2_supercondition_factor=3,
is_verbose=False
)
for image in image_stream:
display(image)
```
### Command Line
Use `image_from_text.py` to generate images from the command line.
```bash
$ python image_from_text.py --text='artificial intelligence' --no-mega
```
