Update llama.py

2024-02-03 23:12:16 +11:00 · 2024-02-03 23:12:16 +11:00 · e500b785f7
commit e500b785f7
parent cf0fae9a55
1 changed files with 4 additions and 3 deletions
--- a/unsloth/models/llama.py
+++ b/unsloth/models/llama.py
@ -680,9 +680,10 @@ def LlamaModel_fast_forward_inference(
            decoder_layer.self_attn,
            hidden_states,
            past_key_values[idx],
-            temp_QA,
-            temp_KV,
-            RH_Q,
+            dp_prefill = False,
+            temp_QA = temp_QA,
+            temp_KV = temp_KV,
+            RH_Q = RH_Q,
        )
        hidden_states += residual