From eacdc75a6e9235ba4c5a8b980e5af9694bbd646d Mon Sep 17 00:00:00 2001
From: Michelia_zhx <zhanghanxiao000@gmail.com>
Date: Fri, 14 Apr 2023 21:12:04 +0800
Subject: [PATCH] fix the while loop

---
 finetune.py | 38 +++++++++++++++++++-------------------
 1 file changed, 19 insertions(+), 19 deletions(-)

diff --git a/finetune.py b/finetune.py
index 0fed260..976204a 100644
--- a/finetune.py
+++ b/finetune.py
@@ -39,11 +39,11 @@ def end_to_end_finetune(train_loader, test_loader, model, t_model, args):
     top5 = AverageMeter()
     finish = False
     while not finish:
+        iter_nums += 1
+        if iter_nums > args.epoch:
+            finish = True
+            break
         for batch_idx, (data, target) in enumerate(train_loader):
-            iter_nums += 1
-            if iter_nums > args.epoch:
-                finish = True
-                break
             # measure data loading time
             data = data.cuda()
             target = target.cuda()
@@ -66,21 +66,21 @@ def end_to_end_finetune(train_loader, test_loader, model, t_model, args):
             # measure elapsed time
             batch_time.update(time.time() - end)
             end = time.time()
-            if iter_nums % args.print_freq == 0:
-                print('Train: [{0}/{1}]\t'
-                      'Time {batch_time.val:.3f} ({batch_time.avg:.3f})\t'
-                      'Data {data_time.val:.3f} ({data_time.avg:.3f})\t'
-                      'LR {lr}\t'
-                      'Loss {loss.val:.4f} ({loss.avg:.4f})\t'
-                      'Prec@1 {top1.val:.3f} ({top1.avg:.3f})\t'
-                      'Prec@5 {top5.val:.3f} ({top5.avg:.3f})'.format(
-                    iter_nums, args.epoch, batch_time=batch_time, lr=lr,
-                    data_time=data_time, loss=losses, top1=top1, top5=top5))
-            if iter_nums % args.eval_freq == 0:
-                validate(test_loader, model)
-                model.train()
-                model.get_feat = 'pre_GAP'
-            scheduler.step()
+        if iter_nums % args.print_freq == 0:
+            print('Train: [{0}/{1}]\t'
+                    'Time {batch_time.val:.3f} ({batch_time.avg:.3f})\t'
+                    'Data {data_time.val:.3f} ({data_time.avg:.3f})\t'
+                    'LR {lr}\t'
+                    'Loss {loss.val:.4f} ({loss.avg:.4f})\t'
+                    'Prec@1 {top1.val:.3f} ({top1.avg:.3f})\t'
+                    'Prec@5 {top5.val:.3f} ({top5.avg:.3f})'.format(
+                iter_nums, args.epoch, batch_time=batch_time, lr=lr,
+                data_time=data_time, loss=losses, top1=top1, top5=top5))
+        if iter_nums % args.eval_freq == 0:
+            validate(test_loader, model)
+            model.train()
+            model.get_feat = 'pre_GAP'
+        scheduler.step()
     validate(test_loader, model)