-
In the first stage, the paper mentions the use of a 200GB dataset, but why does the actual code involve downloading a 2.3TB dataset?
-
According to the BLIP2 paper:
> We adopt the CapFilt method (Li et al., 2022) to create synthetic captions for the web images... We keep top-two captions per image as training data and randomly sam…
-
It was recently noticed that [laion 400m](https://laion.ai/laion-400-open-dataset/) only contains urls from 5M domains.
The same is probably true for other datasets.
Pre-resolving the domains woul…
-
[paper](https://arxiv.org/pdf/2310.03744.pdf)
see llava https://github.com/long8v/PTIR/issues/128#issue-1749571159 here
## TL;DR
- **I read this because.. :** aka LLaVA1.5 / ShareGPT4V에서 LL…
-
Has anyone trained a sample model of this? I realize full-scale training on LAION will take quite a bit of resources (SD v2 trained for 200k GPU hours), but I'm wondering (1) are there any publicly-av…
-
Dear Stable Diffusion Team,
Thanks for sharing the awesome work!
Would it be possible to provide some guidelines on training a new model on a custom dataset? E.g., how to prepare the dataset, ho…
-
Hi !
In Dall-E, we can provide a partial image in addition to the text description so that the model only completes the image. See:
![Capture](https://user-images.githubusercontent.com/35527406/…
-
Thank you for your excellent work. I'm currently training my own CLIP model and have a question. If I use LAION-2B, COYO-700M, and Datacomp datasets simultaneously for training, will it yield better r…
-
Hi there,
Thank you so much for making this library. I'm unfortunately running into the following error
```sh
./main --model '/Users/lucasigel/Downloads/laion_clip-vit-b-32-laion2b-s34b-b79k.gg…
-
Have you ever tried to train bigger models?
Like L, H or G?
If yes, what were the results