Use OpenAIDiscreteVAE as pretrained model
See original GitHub issueI notice the forward function in the OpenAIDiscreteVAE
class is not implemented. I am interested in using this model as a pretrained model and train it on a medical dataset. What is the best approach to achieve this?
- implement the forward function
- Is it possible to initialize the parameters for
DiscreteVAE
class such that we can load the Openai discrete vae model in withload_state_dict( )
?
Any suggestion is welcome!
Issue Analytics
- State:
- Created 2 years ago
- Comments:6 (3 by maintainers)
Top Results From Across the Web
Loading Google AI or OpenAI pre-trained weights ...
To load one of Google AI's, OpenAI's pre-trained models or a PyTorch saved model (an instance of BertForPreTraining saved with torch.save() ), the...
Read more >State-Of-The-Art Text to Image Generation using DALL-E
DALL-E is a novel text-to-image generation model developed by OpenAI. ... Using each pixel of a 256*256 image is impractical because it increases...
Read more >github.com-lucidrains-DALLE-pytorch_-_2021-04-19_12- ...
vae = OpenAIDiscreteVAE() # loads pretrained OpenAI VAE ... Or you can just use the official CLIP model to rank the images from...
Read more >How to Use Pretrained Models from Hugging Face ... - YouTube
In this Deep Learning Tutorial, We'll take a look at the pipeline from Hugging Face for Pretrained Neural Networks. We will go over...
Read more >Chasing fireflies transformer
The wrapper class should take care of downloading and caching the model for you ... OpenAIDiscreteVAE, DALLE vae = OpenAIDiscreteVAE() # loads pretrained...
Read more >Top Related Medium Post
No results found
Top Related StackOverflow Question
No results found
Troubleshoot Live Code
Lightrun enables developers to add logs, metrics and snapshots to live code - no restarts or redeploys required.
Start FreeTop Related Reddit Thread
No results found
Top Related Hackernoon Post
No results found
Top Related Tweet
No results found
Top Related Dev.to Post
No results found
Top Related Hashnode Post
No results found
Top GitHub Comments
@richcmwang there are some slight differences, so it won’t be loadable, unfortunately. one thing to note is that it is coming to light that the VQVAE from Taming Transformers is actually much better than the OpenAI one (despite the latter’s scale)
https://github.com/CompVis/taming-transformers Their repository should be more approachable for training your own VAE (or fine tuning your data from their checkpoints)
Thanks! will give it a try!