pytorch / torchtitan

A native PyTorch Library for large model training
BSD 3-Clause "New" or "Revised" License
1.28k stars 115 forks source link

fix periodic integration test and add helper message on torchdata import failure #353

Closed tianyu-l closed 1 month ago

tianyu-l commented 1 month ago

Stack from ghstack (oldest at bottom):

  1. use the same generic torch CI workflow for periodic integration test, as in #325 for cpu/gpu unit tests.
  2. StatefulDataloader is not in torchdata official release yet. Print helper message if user doesn't have a recent nightly installed.