Instructions to use meituan-longcat/LongCat-Video with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use meituan-longcat/LongCat-Video with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("meituan-longcat/LongCat-Video", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Transformers
How to use meituan-longcat/LongCat-Video with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("meituan-longcat/LongCat-Video", dtype="auto") - Inference
- Notebooks
- Google Colab
- Kaggle
mps backend support
#5
by pypry - opened
I'm trying to make the code support the MPS backend, but I found that some parts use amp.autocast to cast computation to float32. However, MPS's autocast only supports float16 and bfloat16. Is there any way to fix this issue?
As a quick try, you can omit the fp32 casting. This might slightly reduce performance, but the impact should be minor.