Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Can't use it on gpu #64

Closed
satisl opened this issue Mar 22, 2023 · 8 comments
Closed

Can't use it on gpu #64

satisl opened this issue Mar 22, 2023 · 8 comments

Comments

@satisl
Copy link

satisl commented Mar 22, 2023

My cuda version is 11.7.1, cuDNN is 8.8.0.121, gpu is 3060laptop, but I can't run the project, which returns: -1073740791 (0xC0000409).Through debugging I found that it is

results = self.model.detect_language(input)

in line 206 of transcribe.py that is wrong, while the input is

-0.830495 -0.830495 -0.830495 ... -0.830495 -0.830495 -0.830495 ... -0.830495 -0.830495 -0.830495
[cpu:0 float32 storage viewed as 1x80x3000]

What happened to it.

@guillaumekln
Copy link
Contributor

Hi,

The same error is reported here: #42 (comment)

I suppose you are using Windows.

The most probable reason is that you did not update the PATH environment variable for cuDNN. See the step 4. in the cuDNN installation guide: https://docs.nvidia.com/deeplearning/cudnn/install-guide/index.html#installwindows

@satisl
Copy link
Author

satisl commented Mar 22, 2023

image
well,maybe the PATH is not the reason....I don't know

@guillaumekln
Copy link
Contributor

Where is your cuDNN installation?

@satisl
Copy link
Author

satisl commented Mar 22, 2023

According to the tutorial, it covers this directory.Maybe my tutorial is out of date,I'm trying reinstall it.

@satisl
Copy link
Author

satisl commented Mar 22, 2023

Okay,it works.Thanks for help

@zara0m
Copy link

zara0m commented Mar 26, 2023

Hello,
I have Cuda 11.7 and cudnn 8.6.0 on my PC. In a virtual environment, when I first install torch with Cuda and try torch.cuda.is_available, it is true. Then, when I try to install this module, it cannot install ctranslate2 unless I downgrade my numpy to 1.23.1 version; and after that the torch.cuda.is_available became false.
I don't know why I cannot use my gpu for faster whisper and it crashes.
I appreciate it if you could help me fix this.
Thanks.

@satisl
Copy link
Author

satisl commented Mar 27, 2023

Pytorch has a built-in cuda, so the result of torch.cuda.is_available cannot be used to determine whether the cuda installation is complete.
And, this package does not use pythoch. Just like this:
Package Version


av 10.0.0
certifi 2022.12.7
charset-normalizer 3.1.0
colorama 0.4.6
ctranslate2 3.10.1
faster-whisper 0.3.0
filelock 3.10.6
huggingface-hub 0.13.3
idna 3.4
numpy 1.24.2
packaging 23.0
pip 23.0.1
PyYAML 6.0
requests 2.28.2
setuptools 67.6.0
tokenizers 0.13.2
tqdm 4.65.0
typing_extensions 4.5.0
urllib3 1.26.15
wheel 0.40.0

@zara0m
Copy link

zara0m commented Mar 27, 2023

Thanks!
I tried reinstalling and it worked finally.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

3 participants