[Model Runner V2] Fix pooling (#36019)
Signed-off-by: Nick Hill <nickhill123@gmail.com>
This commit is contained in:
@@ -1117,7 +1117,7 @@ class GPUModelRunner(LoRAModelRunnerMixin):
|
||||
# The prior execute_model call must have failed.
|
||||
return None
|
||||
|
||||
input_batch, _, _, _, hidden_states, _, kv_connector_output = (
|
||||
input_batch, _, _, _, hidden_states, _, kv_connector_output, _ = (
|
||||
self.execute_model_state
|
||||
)
|
||||
self.execute_model_state = None
|
||||
|
||||
Reference in New Issue
Block a user