search

ControlNet / LAV-DF

[CVIU] Glitch in the Matrix: A Large Scale Benchmark for Content Driven Audio-Visual Forgery Detection and Localization
https://www.sciencedirect.com/science/article/pii/S1077314223001984
Other
73 stars 10 forks source link

GPU Requirements for Training #7

Closed jeremy-swack closed 1 year ago

jeremy-swack commented 1 year ago

I am attempting to train BA-TFD+ on a custom data set, but I am running into memory issues. I am using a VM with 2 M60 GPUs with 8 GB of memory each, but I am not able to run the training module without crashing, even when setting the batch size to 1. Is there a recommended amount of GPU memory to be able to train the model?

ControlNet commented 1 year ago

Hi, as mentioned in the paper, we used A100 80GB for training BA-TFD+, which require much more memory than BA-TFD which I used RTX3090 24GB to train.

I think for the minimum requirement, 48GB vRAM is required.