feat(doc): note lora kernel incompat with RLHF (#2706) [skip ci]

* feat(doc): note lora kernel incompat with RLHF * fix: add validation following comments * chore: fix typo following suggestion
2025-05-28 15:48:40 +07:00
parent 3e6948be97
commit e33f225434
2 changed files with 19 additions and 2 deletions
--- a/docs/lora_optims.qmd
+++ b/docs/lora_optims.qmd
@@ -84,6 +84,10 @@ lora_qkv_kernel: true
 lora_o_kernel: true
 ```

+::: {.callout-note}
+Currently, LoRA kernels are not supported for RLHF training, only SFT.
+:::
+
 ## Requirements

 - One or more NVIDIA or AMD GPUs (in order to use the Triton kernels)