don't use mask expansion for inference (#392)

src/axolotl/utils/models.py

@@ -138,8 +138,10 @@ def load_model(
         LOG.info("patching with xpos rope")
         replace_llama_rope_with_xpos_rope()
 
-    if cfg.is_llama_derived_model and (
-        cfg.max_packed_sequence_len or cfg.sample_packing
+    if (
+        cfg.is_llama_derived_model
+        and (cfg.max_packed_sequence_len or cfg.sample_packing)
+        and not cfg.inference
     ):
         from axolotl.monkeypatch.llama_expand_mask import hijack_expand_mask