camenduru

Hi @adefossez 👋 If we convert the video to a 3:1 aspect ratio, then people can [share it on Twitter](https://developer.twitter.com/en/docs/twitter-api/v1/media/upload-media/uploading-media/media-best-practices#:~:text=Aspect%20ratio%20must%20be%20between%201:3%20and%203:1) https://github.com/facebookresearch/audiocraft/compare/main...camenduru:audiocraft:dev https://github.com/camenduru/MusicGen-colab/assets/54370274/4bf56eb6-4e0e-48cc-b8b7-828d8f81d188

also when we switch the model, we should free the vram otherwise the vm will crash 

Hi @Tanmay09516 ❤ thanks for the good ide 🤩 please join discord workshop->aica channel we can work on this

Hi @Surlix thanks ❤ for the cool ideas 🤩 we are working on this with @Tanmay09516 if possible please join our discord server https://discord.gg/k5BwmmvJJU

copy paste your colab code here please

`Free T4` has only 12.7 GB of system RAM. We need torch.float8_e4m3fn for T5EncoderModel, or maybe this: https://github.com/pytorch/ao. I will try that. ```py from transformers import T5EncoderModel text_encoder = T5EncoderModel.from_pretrained("/content/CogVideoX-5b",...

But why does `torch_dtype=torch.float8_e4m3fn` work with transformer and vae but not with text_encoder on a T4? ```py import torch from diffusers import AutoencoderKLCogVideoX, CogVideoXTransformer3DModel, CogVideoXDDIMScheduler from transformers import T5EncoderModel, T5Tokenizer...

same with H100 

Maybe with this https://github.com/kijai/ComfyUI-CogVideoXWrapper and this https://huggingface.co/mcmonkey/google_t5-v1_1-xxl_encoderonly/tree/main, we can run it on a Free T4

@PyroFilmsFX here: https://github.com/camenduru/CogVideoX-5B-jupyter/blob/main/CogVideoX_5B_jupyter_free.ipynb