transforms (#272)

encoding/lib/cpu/operator.h

@@ -93,7 +93,7 @@ py::array_t<float> apply_transform(int H, int W, int C, py::array_t<float> img,
   auto ctm_buf = ctm.request();
 
   // printf("H: %d, W: %d, C: %d\n", H, W, C);
-  py::array_t<float> result{img_buf.size};
+  py::array_t<float> result{(unsigned long)img_buf.size};
   auto res_buf = result.request();
 
   float *img_ptr = (float *)img_buf.ptr;

encoding/transforms/get_transform.py

@@ -36,7 +36,7 @@ def get_transform(dataset, base_size=None, crop_size=224, rand_aug=False, etrans
                 CenterCrop(crop_size),
             ])
         train_transforms.extend([
-                RandomHorizontalFlip(),
+            RandomHorizontalFlip(),
             ColorJitter(0.4, 0.4, 0.4),
             ToTensor(),
             Lighting(0.1, _imagenet_pca['eigval'], _imagenet_pca['eigvec']),
@@ -65,16 +65,16 @@ def get_transform(dataset, base_size=None, crop_size=224, rand_aug=False, etrans
             normalize,
         ])
     elif dataset == 'cifar10':
-        transform_train = transforms.Compose([
-            transforms.RandomCrop(32, padding=4),
-            transforms.RandomHorizontalFlip(),
-            transforms.ToTensor(),
-            transforms.Normalize((0.4914, 0.4822, 0.4465), 
-                                 (0.2023, 0.1994, 0.2010)),
+        transform_train = Compose([
+            RandomCrop(32, padding=4),
+            RandomHorizontalFlip(),
+            ToTensor(),
+            Normalize((0.4914, 0.4822, 0.4465), 
+                      (0.2023, 0.1994, 0.2010)),
         ])
-        transform_val = transforms.Compose([
-            transforms.ToTensor(),
-            transforms.Normalize((0.4914, 0.4822, 0.4465), 
+        transform_val = Compose([
+            ToTensor(),
+            Normalize((0.4914, 0.4822, 0.4465), 
                     (0.2023, 0.1994, 0.2010)),
         ])
     return transform_train, transform_val

encoding/utils/lr_scheduler.py

@@ -29,9 +29,11 @@ class LR_Scheduler(object):
         iters_per_epoch: number of iterations per epoch
     """
     def __init__(self, mode, base_lr, num_epochs, iters_per_epoch=0,
-                 lr_step=0, warmup_epochs=0):
+                 lr_step=0, warmup_epochs=0, quiet=False):
         self.mode = mode
-        print('Using {} LR scheduler with warm-up epochs of {}!'.format(self.mode, warmup_epochs))
+        self.quiet = quiet
+        if not quiet:
+            print('Using {} LR scheduler with warm-up epochs of {}!'.format(self.mode, warmup_epochs))
         if mode == 'step':
             assert lr_step
         self.base_lr = base_lr
@@ -57,8 +59,9 @@ class LR_Scheduler(object):
         else:
             raise NotImplemented
         if epoch > self.epoch and (epoch == 0 or best_pred > 0.0):
-            print('\n=>Epoch %i, learning rate = %.4f, \
-                previous best = %.4f' % (epoch, lr, best_pred))
+            if not self.quiet:
+                print('\n=>Epoch %i, learning rate = %.4f, \
+                    previous best = %.4f' % (epoch, lr, best_pred))
             self.epoch = epoch
         assert lr >= 0
         self._adjust_learning_rate(optimizer, lr)