use even if not using sample packing

2023-10-13 17:54:35 -04:00
parent f95858d369
commit 080612219b
1 changed files with 5 additions and 1 deletions
--- a/src/axolotl/utils/models.py
+++ b/src/axolotl/utils/models.py
@@ -136,7 +136,11 @@ def load_model(

            replace_stablelm_attn_with_flash_attn(cfg.base_model)

-    if cfg.is_llama_derived_model and cfg.flash_attention and cfg.sample_packing:
+    if (
+        cfg.is_llama_derived_model
+        and cfg.flash_attention
+        and (cfg.noisy_embeddings_alpha or cfg.sample_packing)
+    ):
        if cfg.device not in ["mps", "cpu"] and not inference:
            from axolotl.monkeypatch.llama_attn_hijack_flash import (
                replace_llama_attn_with_flash_attn,