search

pytorch / torchtitan

A native PyTorch Library for large model training
BSD 3-Clause "New" or "Revised" License
2.68k stars 212 forks source link

Port #642's loss changes to estimation.py #656

Closed carmocca closed 1 month ago

carmocca commented 1 month ago

Fixes https://github.com/pytorch/torchtitan/pull/642#issuecomment-2442220336