search

yxlu-0102 / MP-SENet

MP-SENet: A Speech Enhancement Model with Parallel Denoising of Magnitude and Phase Spectra
MIT License
257 stars 36 forks source link

Out of Memory when inferencing with 60secs file #37

Open HimaJyothi17 opened 2 weeks ago

HimaJyothi17 commented 2 weeks ago

torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 137.36 GiB. GPU 0 has a total capacity of 47.54 GiB of which 44.17 GiB is free. Process 1932274 has 3.36 GiB memory in use. Of the allocated memory 1.89 GiB is allocated by PyTorch, and 23.82 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)

Why it is expecting 137 GB to infer just 60 sec file? Is this model only for real time purpose?

yxlu-0102 commented 2 weeks ago

Current mainstream speech enhancement models do indeed occupy a large amount of memory.

You can try enhancing the speech by dividing it into segments like this:

截屏2024-06-26 15 32 36
HimaJyothi17 commented 2 weeks ago

What are the 8 type of noises used for training in DEMAND dataset?