debug

mlcommons · Sep 25, 2023 · faf775e · faf775e
1 parent 67fb0e4
commit faf775e
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/submission_runner.py b/submission_runner.py
@@ -424,7 +424,7 @@ def train_once(
     train_state['last_step_end_time'] = get_time()
     if global_step == 2:
       s = torch.cuda.memory._snapshot()
-      with open(f"/experiment_runs/snapshot_{global_step}_half_bsz.pickle", "wb") as f:
+      with open(f"/experiment_runs/snapshot_{global_step}_full_bsz.pickle", "wb") as f:
         dump(s, f)
   metrics = {'eval_results': eval_results, 'global_step': global_step}