tanner/min-dalle-test
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

update readme, random seed if none specified

Browse Source
This commit is contained in:
Brett Kuprel 2022-07-01 18:50:11 -04:00
parent ec9e22e4d4
commit fff44d683e
2 changed files with 22 additions and 26 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

44
README.md vendored
View File

@ -11,7 +11,7 @@ The flax model, and the code for coverting it to torch, have been moved [here](h
### Install ### Install
```bash ```zsh
$ pip install min-dalle $ pip install min-dalle
``` ```
@ -19,6 +19,16 @@ $ pip install min-dalle
Use the python script `image_from_text.py` to generate images from the command line. Use the python script `image_from_text.py` to generate images from the command line.
```zsh
$ python image_from_text.py --text='artificial intelligence' --seed=7
```
![Artificial Intelligence](examples/artificial_intelligence.png)
```zsh
$ python image_from_text.py --text='court sketch of godzilla on trial' --mega
```
![Godzilla Trial](examples/godzilla_on_trial.png)
To load a model once and generate multiple times, initialize `MinDalleTorch`, then call `generate_image` with some text and a seed. To load a model once and generate multiple times, initialize `MinDalleTorch`, then call `generate_image` with some text and a seed.
```python ```python
@ -29,33 +39,17 @@ model = MinDalleTorch(
is_reusable=True, is_reusable=True,
models_root='./pretrained' models_root='./pretrained'
) )
image = model.generate_image("court sketch of godzilla on trial", seed=40)
``` ```
Model parameters will be downloaded as needed to the directory specified. The models can also be manually downloaded [here](https://huggingface.co/kuprel/min-dalle/tree/main). ```python
image = model.generate_image("a comfy chair that looks like an avocado")
### Examples display(image)
``` ```
python image_from_text.py --text='artificial intelligence' --seed=7
```
![Alien](examples/artificial_intelligence.png)
```python
``` image = model.generate_image("trail cam footage of gollum eating watermelon", seed=1)
python image_from_text.py --text='a comfy chair that looks like an avocado' --mega --seed=10 display(image)
```
![Avocado Armchair](examples/avocado_armchair.png)
```
python image_from_text.py --text='court sketch of godzilla on trial' --mega --seed=40
```
![Godzilla Trial](examples/godzilla_on_trial.png)
```
python image_from_text.py --text='trail cam footage of gollum eating watermelon' --mega --seed=1
``` ```
![Gollum Trailcam](examples/gollum_trailcam.png) ![Gollum Trailcam](examples/gollum_trailcam.png)
Model parameters will be downloaded as needed to the directory specified. The models can also be manually downloaded [here](https://huggingface.co/kuprel/min-dalle/tree/main).

2
min_dalle/min_dalle_torch.py
View File

@ -6,6 +6,7 @@ from torch import LongTensor
import torch import torch
import json import json
import requests import requests
import random
torch.set_grad_enabled(False) torch.set_grad_enabled(False)
torch.set_num_threads(os.cpu_count()) torch.set_num_threads(os.cpu_count())
@ -167,6 +168,7 @@ class MinDalleTorch:
if not self.is_reusable: self.init_decoder() if not self.is_reusable: self.init_decoder()
print("sampling image tokens") print("sampling image tokens")
if seed < 0: seed = random.randint(0, 2 ** 31)
torch.manual_seed(seed) torch.manual_seed(seed)
image_tokens = self.decoder.forward(text_tokens, encoder_state) image_tokens = self.decoder.forward(text_tokens, encoder_state)
if not self.is_reusable: del self.decoder if not self.is_reusable: del self.decoder