Video-to-Video
Diffusers
English
Chinese
video-editing
streaming
quantization
int8
text-stream-stripping
Instructions to use kylinfires/Joy-face_change_optimized with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use kylinfires/Joy-face_change_optimized with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("kylinfires/Joy-face_change_optimized", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
File size: 789 Bytes
b9fc164 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 | # Adapted from https://github.com/hao-ai-lab/FastVideo/blob/main/fastvideo/distributed/communication_op.py
import torch
import torch.distributed as dist
def broadcast_item(item, src: int = 0):
if not dist.is_initialized():
return item
item_list = [item]
dist.broadcast_object_list(item_list, src=src)
return item_list[0]
def all_ranks_finite(value: torch.Tensor) -> bool:
"""Make the same finite-value decision on every rank before backward/step."""
from torch.distributed.tensor import DTensor
if isinstance(value, DTensor):
value = value.to_local()
finite = torch.isfinite(value.detach()).all().to(dtype=torch.int32)
if dist.is_initialized():
dist.all_reduce(finite, op=dist.ReduceOp.MIN)
return bool(finite.item())
|