OSError: tencent/HunyuanVideo does not appear to have a file named config.json.
#27
by
reppane
- opened
Hello,
Running the example code in here fails due to a missing config file. Seems to be the one for the transformer. Can someone verify this? In the example, there seems to be a typo also in the subfolder name (transformer -> tansformers), but changing it does not make a difference either.
requests.exceptions.HTTPError: 404 Client Error: Not Found for url: https://huggingface.co/tencent/HunyuanVideo/resolve/main/transformer/config.json
Thank you!
+1 Same issue
Hi, the Hunyuan team is working on adding the weights in Diffusers format soon (see comment here). In the meantime, you can still load it now if you specify revision="refs/pr/18" as shown below.
import torch
from diffusers import HunyuanVideoPipeline, HunyuanVideoTransformer3DModel
from diffusers.utils import export_to_video
model_id = "tencent/HunyuanVideo"
transformer = HunyuanVideoTransformer3DModel.from_pretrained(
model_id, subfolder="transformer", torch_dtype=torch.bfloat16, revision='refs/pr/18'
)
pipe = HunyuanVideoPipeline.from_pretrained(model_id, transformer=transformer, revision='refs/pr/18', torch_dtype=torch.float16)
Thanks
@stevhliu
I had tried that branch and ran into a CUDA out of memory error with the code on this pull request on a 48GB Ada gpu.
https://huggingface.co/tencent/HunyuanVideo/discussions/18