search

fschmid56 / EfficientAT

This repository aims at providing efficient CNNs for Audio Tagging. We provide AudioSet pre-trained models ready for downstream training and extraction of audio embeddings.
MIT License
218 stars 41 forks source link

multiple gpus support #4

Closed didadida-r closed 7 months ago

didadida-r commented 1 year ago

Hi,

how to train the script using multiple gpus? thanks

fschmid56 commented 1 year ago

Hi, this script is kept as simple as possible, also because most of the models can be trained in a single day on a consumer-grade GPU on AudioSet. For what purpose would you need multiple GPU support?

didadida-r commented 1 year ago

Well, I just think it would be more convenient to do some experiments if we support multi-gpu training

fschmid56 commented 1 year ago

Okay, I'm currently busy with a deadline, so if you implement it and create a pull request that would be nice. Otherwise, I'll do it but it will take a bit.

turian commented 8 months ago

@didadida-r If you just use pytorch-lightning, you get multi GPU training on any torch.nn.Module. Seems like that would be a better and more generic solution for you

fschmid56 commented 7 months ago

I added the ex_pl_audioset.py file for multi GPU training on AudioSet.