Fix #566, summary.csv writing to pwd on local_rank != 0. Tweak benchmark mem handling to see if it reduces likelihood of 'bad' exceptions on OOM.

benchmark.py

@@ -374,14 +374,14 @@ def _try_run(model_name, bench_fn, initial_batch_size, bench_kwargs):
     batch_size = initial_batch_size
     results = dict()
     while batch_size >= 1:
+        torch.cuda.empty_cache()
         try:
             bench = bench_fn(model_name=model_name, batch_size=batch_size, **bench_kwargs)
             results = bench.run()
             return results
         except RuntimeError as e:
-            torch.cuda.empty_cache()
-            batch_size = decay_batch_exp(batch_size)
             print(f'Error: {str(e)} while running benchmark. Reducing batch size to {batch_size} for retry.')
+        batch_size = decay_batch_exp(batch_size)
     return results
 
 

train.py

@@ -560,7 +560,7 @@ def main():
     best_metric = None
     best_epoch = None
     saver = None
-    output_dir = ''
+    output_dir = None
     if args.local_rank == 0:
         if args.experiment:
             exp_name = args.experiment
@@ -606,6 +606,7 @@ def main():
                 # step LR for next epoch
                 lr_scheduler.step(epoch + 1, eval_metrics[eval_metric])
 
+            if output_dir is not None:
                 update_summary(
                     epoch, train_metrics, eval_metrics, os.path.join(output_dir, 'summary.csv'),
                     write_header=best_metric is None, log_wandb=args.log_wandb and has_wandb)
@@ -623,7 +624,7 @@ def main():
 
 def train_one_epoch(
         epoch, model, loader, optimizer, loss_fn, args,
-        lr_scheduler=None, saver=None, output_dir='', amp_autocast=suppress,
+        lr_scheduler=None, saver=None, output_dir=None, amp_autocast=suppress,
         loss_scaler=None, model_ema=None, mixup_fn=None):
 
     if args.mixup_off_epoch and epoch >= args.mixup_off_epoch: