From 7386e3950c7e5e55260dbfa0d38198d7c51dac49 Mon Sep 17 00:00:00 2001 From: Robert Young Date: Tue, 4 Nov 2025 16:36:08 -0500 Subject: [PATCH] forgot to add attribution to new PyTorch page --- docs/hpc/08_ml_ai_hpc/02_pytorch_intro.md | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/docs/hpc/08_ml_ai_hpc/02_pytorch_intro.md b/docs/hpc/08_ml_ai_hpc/02_pytorch_intro.md index 2335384e03..7d66d6888a 100644 --- a/docs/hpc/08_ml_ai_hpc/02_pytorch_intro.md +++ b/docs/hpc/08_ml_ai_hpc/02_pytorch_intro.md @@ -1,5 +1,9 @@ # Single-GPU Training with PyTorch +:::info +This was adapted from [Princeton University Multi-GPU Training with PyTorch](https://github.com/PrincetonUniversity/multi_gpu_training) +::: + It is important to optimize your script for the single-GPU case before moving to multi-GPU training. This is because as you request more resources, your queue time increases. We also want to avoid wasting resources by running code that is not optimized. Here we train a CNN on the MNIST dataset using a single GPU as an example. We profile the code and make performance improvements.