Add missing `rng` for ImageNet-ViT and WMT

mlcommons / algorithmic-efficiency

MLCommons Algorithmic Efficiency is a benchmark and competition measuring neural network training speedups due to algorithmic improvements in both training algorithms and models.

https://mlcommons.org/en/groups/research-algorithms/

Apache License 2.0

321 stars 62 forks source link

Add missing `rng` for ImageNet-ViT and WMT #677

Closed runame closed 6 months ago

runame commented 6 months ago

I went through all workloads and found two where we also don't initialize with a separate rng for dropout -- as in #672.

github-actions[bot] commented 6 months ago

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅