mosaicml / llm-foundry

LLM training code for Databricks foundation models
https://www.databricks.com/blog/introducing-dbrx-new-state-art-open-llm
Apache License 2.0
3.84k stars 503 forks source link

Revert to older TE version #1267

Closed mvpatel2000 closed 4 weeks ago

mvpatel2000 commented 4 weeks ago

Revert to older TE version. We're seeing some issues with flash-attn mismatch.

mvpatel2000 commented 4 weeks ago

Force merging after approval from @dakinggg