Clay Mullis
Clay Mullis
Hey! Great work on the original notebook. @mehdidc is working on a similar project now as well. I recommended your original CLIP Decision Transformer notebook; and came here to find...
Having trouble debugging this; but I think after looking at the code briefly - there's not any webdataset retrieval I dont think? The root problem is that I am not...
Laion.ai recently has been donated access to a few 8xA100 pods, of which I'm able to use. Tinkering with using `cog` for inference on these is tricky because it is...
(disclaimer): this is code for training a _custom_ CLIP from the repository here, not the one in the OpenAI repo. For something like that I recommend open_clip. There are valid...
https://github.com/lucidrains/DALLE-pytorch/discussions/139#discussioncomment-560790 It appears as though adamw does work better but the weight decay is creating strange generations. Getting the same strange "brown" generations even though the loss continues to go...
https://gist.github.com/afiaka87/b29213684a1dd633df20cab49d05209d If there are any bugs - please make a comment below. When in doubt; restart your kernel. Tends to fix things a lot.
### Discussed in https://github.com/lucidrains/DALLE-pytorch/discussions/339   Originally posted by **afiaka87** July 17, 2021 I've been training a DALL-e with the goal of seeing whether or not a caption could be...
### Discussed in https://github.com/lucidrains/DALLE-pytorch/discussions/335 Originally posted by **afiaka87** July 11, 2021 [Full W&B training session](https://wandb.ai/dalle-pytorch-replicate/COCO512_16_16D_16H_80TSL)    Details Transformer: - Visual Dim - 512 - Max Text Length/Language Dim...
In the cogview paper they claim that by giving the text as much importance they achieve a better result. They "hypothesize" that this is because the transformer is learning both...