Update context_flashattention_nopad.py To fix triton 2.0.0 import err…

…or. (#364)
ModelTC · Mar 19, 2024 · 56940dd · 56940dd
1 parent c96e5d5
commit 56940dd
Showing 1 changed file with 2 additions and 0 deletions.
diff --git a/lightllm/models/llama/triton_kernel/context_flashattention_nopad.py b/lightllm/models/llama/triton_kernel/context_flashattention_nopad.py
@@ -506,6 +506,8 @@ def context_attention_fwd(q, k, v, o, b_start_loc, b_seq_len, max_input_len):
             num_stages=1,
         )
         return
+
+    context_attention_fwd_no_prompt_cache = None
 
 else:
     raise Exception("error triton version!")