search

Vchitect / Latte

Latte: Latent Diffusion Transformer for Video Generation.
Apache License 2.0
1.45k stars 147 forks source link

Issue about "LayerNormKernelImpl" not implemented for 'Half' #38

Closed kaiw7 closed 4 months ago

kaiw7 commented 4 months ago

Hi, I would like to run bash sample/ucf101.sh for sampling but I met this "LayerNormKernelImpl" not implemented for 'Half'. Could you give some insights about how to resolve it? Many thanks.

maxin-cn commented 4 months ago

Hi, I would like to run bash sample/ucf101.sh for sampling but I met this "LayerNormKernelImpl" not implemented for 'Half'. Could you give some insights about how to resolve it? Many thanks.

Hi, thanks for your interest. I did not find the function LayerNormKernelImpl in Latte. Could you provide more information?

kaiw7 commented 4 months ago

Hi, I would like to run bash sample/ucf101.sh for sampling but I met this "LayerNormKernelImpl" not implemented for 'Half'. Could you give some insights about how to resolve it? Many thanks.

Hi, thanks for your interest. I did not find the function LayerNormKernelImpl in Latte. Could you provide more information?

Hi, it seems that this issue is caused by the environment. After I install the environment based on your environment.yaml, it seems that the pytorch version is not GPU version. Could you check the environment file again? Many thanks

image
maxin-cn commented 4 months ago

Hi, I would like to run bash sample/ucf101.sh for sampling but I met this "LayerNormKernelImpl" not implemented for 'Half'. Could you give some insights about how to resolve it? Many thanks.

Hi, thanks for your interest. I did not find the function LayerNormKernelImpl in Latte. Could you provide more information?

Hi, it seems that this issue is caused by the environment. After I install the environment based on your environment.yaml, it seems that the pytorch version is not GPU version. Could you check the environment file again? Many thanks image

I used the installation method recommended by pytorch. image

If you still have environmental problems, please install as recommended by Pytorch.

RuslanPeresy commented 4 months ago

First, you need to install pytorch for your CUDA version - in the environment.yml CUDA version 11.7 is hardcoded. For example, if you have CUDA 12.1: conda install pytorch torchvision torchaudio pytorch-cuda=12.1 -c pytorch -c nvidia

Then, the bash script sample/ucf101.sh exports hardcoded CUDA_VISIBLE_DEVICES=7. Check your GPU index(es) using nvidia-smi command, and change the number if needed.

kaiw7 commented 4 months ago

Hi, thanks, I solved it.