Bug- pdf2muse gives an error about CUDA while --use-tf is not specified

[MatNieuw]

A simple convert command of a 3 page PDF took very long. Eventually for page 001 an CUDA error popped up, page 002 finished, page 003 was never processed. The resulting page was not anywhere near the original, and MuseScore 4.6.5 complained about missing things .

But the problem is that it tried to use CUDA, which is not installed on my machine. See the screenshot.

[thedivergentai]

It should fall back to CPU if you don't have CUDA installed but running on CPU would be slower.

As for the quality I was attempting to use an open source model to try and get the workflow to work. It could be useful. But it'll need some work before it can work somewhat decently, if at all.

I'll see what more I can do!

Thanks for your feedback.

[MatNieuw]

No hurry! I know that a CPU is slower, but that is fully acceptable. I don't know how much of your code controls CUDA or not, but it should be easy to determine if CUDA (or other accelerator) is available, and then just run the CPU. Then at least it would produce 3 pages (of not-well converted music, but that's another issue).

I'm also a little confused about the cmd line option --use-tf . Tensorflow is not tied to CUDA, and should run on CPUs and non-CUDA interfaced GPUs also. Is there maybe a tf config in this that says it must use CUDA?

Thanks for any work you can do on this.