Long(Tony) Lian

icon indicating a website link https://tonylian.com/ icon indicating an email [email protected]

icon company UC Berkeley icon location Berkeley, California icon location PhD Student at UC Berkeley (EECS) | UC Berkeley 22' (CS) | Research interests: Improved Reasoning on LLMs/VLMs through RL

Results 82 comments of Long(Tony) Lian

Question about the refiner step

Two updates: 1. If the intermediate images are not needed (i.e., we don't want to compare before/after), no decoding and re-encoding between the base generation and refinement stage are used....

Question about the refiner step

Thanks for this example. Is using images consistently better than using latents?

Suggesting features to advanced settings in demo colab

> I noticed that there is a little bit of GPU memory to generate images with a bit more resolution than 1024px. Could you please add the option to select...

Not an issue but more of a question

Do you mind explaining how this can be useful? The current demo supports setting a seed for each generation.

Not an issue but more of a question

Seems like it was commented out. It is added.

[suggestion] add aspect ratio options

Great idea! Do you mind giving us a PR for this?

Google Colab only local URL

This might be because gradio is temporarily down. Could you try this in a while?

Google Colab only local URL

Seems like if you update gradio with: ``` !pip install gradio==3.14.0 ``` You can get a sharing link. I will update the `requirements.txt`.

Google Colab only local URL

I updated `requirements.txt`. It should work if you run the Colab (pip will download the right version).

Do you consider compatibility with stable diffusion XL?

Since this work is training-free, it should be compatible.