remove empty cache call

Signed-off-by: Kyle Sayers <kylesayrs@gmail.com>
vllm-project · Jan 29, 2025 · b9bd970 · b9bd970
1 parent b61092b
commit b9bd970
Showing 1 changed file with 0 additions and 4 deletions.
diff --git a/src/llmcompressor/modifiers/utils/pytorch_helpers.py b/src/llmcompressor/modifiers/utils/pytorch_helpers.py
@@ -106,10 +106,6 @@ def run_calibration_forward(
                 # move on to next calibration sample
                 intermediates.append((e.args, e.kwargs))
 
-        # TODO: not ideal, figure out where we aren't freeing memory instead
-        # currently without this we run OOM on the 2nd forward pass
-        torch.cuda.empty_cache()
-
     return intermediates