Update cutmixup_collator.py (facebookresearch#196)

Summary: Fixed collator compatibility issues w/ MoCo and SimCLR Pull Request resolved: facebookresearch#196 Reviewed By: blefaudeux Differential Revision: D26651453 Pulled By: growlix fbshipit-source-id: 120d14005236ca9e7ced948fdfd2a3a0ca5770bb
yangsenwxy · Feb 25, 2021 · a7a2828 · a7a2828
1 parent 4cffc1c
commit a7a2828
Show file tree

Hide file tree

Showing 7 changed files with 15 additions and 16 deletions.
diff --git a/...ging/pretrain/supervised/supervised_1gpu_vision_transformer_debug_integration_cutmix.yaml b/...ging/pretrain/supervised/supervised_1gpu_vision_transformer_debug_integration_cutmix.yaml
@@ -19,7 +19,7 @@ config:
       TRANSFORMS:
         - name: RandomResizedCrop
           size: 224
-        - name: VisslRandAugment
+        - name: RandAugment
           magnitude: 5
           weight_choice: 0
         - name: ToTensor
@@ -35,7 +35,6 @@ config:
       }
     TEST:
       DATA_SOURCES: [disk_folder]
-      #      DATA_PATHS: ["<path to test folder>"]
       LABEL_SOURCES: [disk_folder]
       DATASET_NAMES: [imagenet1k_debug_folder]
       BATCHSIZE_PER_REPLICA: 32

diff --git a/configs/config/debugging/pretrain/supervised/supervised_deit_b_integration_debug.yaml b/configs/config/debugging/pretrain/supervised/supervised_deit_b_integration_debug.yaml
@@ -20,7 +20,7 @@ config:
         - name: RandomResizedCrop
           size: 224
         - name: RandomHorizontalFlip
-        - name: VisslRandAugment
+        - name: RandAugment
           magnitude: 9
           magnitude_std: 0.5
           increasing_severity: True

diff --git a/configs/config/pretrain/vision_transformer/supervised/supervised_16gpu_deit_b_example.yaml b/configs/config/pretrain/vision_transformer/supervised/supervised_16gpu_deit_b_example.yaml
@@ -16,11 +16,12 @@ config:
       LABEL_SOURCES: [disk_folder]
       DATASET_NAMES: [imagenet1k_folder]
       BATCHSIZE_PER_REPLICA: 32
+      DROP_LAST: True
       TRANSFORMS:
         - name: RandomResizedCrop
           size: 224
         - name: RandomHorizontalFlip
-        - name: VisslRandAugment
+        - name: RandAugment
           magnitude: 9
           magnitude_std: 0.5
           increasing_severity: True

diff --git a/configs/config/pretrain/vision_transformer/swav/swav_deit_b_2nodes.yaml b/configs/config/pretrain/vision_transformer/swav/swav_deit_b_2nodes.yaml
@@ -3,7 +3,6 @@ config:
   HOOKS:
     TENSORBOARD_SETUP:
       USE_TENSORBOARD: True
-      EXPERIMENT_LOG_DIR: "/checkpoint/ito/vision_transformer/1gpu_test"
       FLUSH_EVERY_N_MIN: 20
   CHECKPOINT:
     DIR: "."
@@ -13,19 +12,18 @@ config:
     NUM_DATALOADER_WORKERS: 8
     TRAIN:
       DATA_SOURCES: [disk_folder]
-      #      DATA_PATHS: ["<path to train folder>"]
-      LABEL_SOURCES: [disk_folder]
-      DATASET_NAMES: [imagenet1k_debug_folder]
+      DATASET_NAMES: [imagenet1k_folder]
       LABEL_TYPE: "zero"
       BATCHSIZE_PER_REPLICA: 16
+      DROP_LAST: True
       TRANSFORMS:
         - name: ImgPilToMultiCrop
           total_num_crops: 2
           size_crops: [224]
           num_crops: [2]
           crop_scales: [[0.14, 1]]
         - name: RandomHorizontalFlip
-        - name: VisslRandAugment
+        - name: RandAugment
           magnitude: 9
           magnitude_std: 0.5
           increasing_severity: True
@@ -42,14 +40,15 @@ config:
           std: [0.229, 0.224, 0.225]
       COLLATE_FUNCTION: cutmixup_collator
       COLLATE_FUNCTION_PARAMS: {
+        "ssl_method": "swav",
         "mixup_alpha": 1.0, # mixup alpha value, mixup is active if > 0.
         "cutmix_alpha": 1.0, # cutmix alpha value, cutmix is active if > 0.
         "prob": 1.0, # probability of applying mixup or cutmix per batch or element
         "switch_prob": 0.5, # probability of switching to cutmix instead of mixup when both are active
         "mode": "batch", # how to apply mixup/cutmix params (per 'batch', 'pair' (pair of elements), 'elem' (element)
         "correct_lam": True, # apply lambda correction when cutmix bbox clipped by image borders
         "label_smoothing": 0.1, # apply label smoothing to the mixed target tensor
-        "num_classes": 1000 # number of classes for target
+        "num_classes": 1 # number of classes for target
       }
   MODEL:
     TRUNK:

diff --git a/vissl/config/defaults.yaml b/vissl/config/defaults.yaml
@@ -195,7 +195,7 @@ config:
       # no valid values other than 0 or None), unclear if beneficial. Default =
       # None.
       # TRANSFORMS:
-      #   - name: VisslRandAugment
+      #   - name: RandAugment
       #     magnitude: 9
       #     magnitude_std: 0.5
       #     num_layers: 2

diff --git a/vissl/data/collators/cutmixup_collator.py b/vissl/data/collators/cutmixup_collator.py
@@ -147,11 +147,11 @@ def data_back_to_input_form(data, labels, data_valid, data_idx):
     for sample_i in range(num_images):
         sample_input_form = {"data": [], "data_valid": [], "data_idx": [], "label": []}
         for duplicate_i in range(num_duplicates):
-            sample_input_form["data"].append(data[duplicate_i][sample_i])
-            sample_input_form["label"].append(labels[duplicate_i][sample_i])
             valid_and_idx_i = sample_i + (num_duplicates * duplicate_i)
-            sample_input_form["data_idx"].append(data_idx[valid_and_idx_i])
-            sample_input_form["data_valid"].append(data_valid[valid_and_idx_i])
+            sample_input_form["data"].append(data[duplicate_i][sample_i])
+            sample_input_form["label"].append(labels[duplicate_i][sample_i].tolist())
+            sample_input_form["data_idx"].append(data_idx[valid_and_idx_i].item())
+            sample_input_form["data_valid"].append(data_valid[valid_and_idx_i].item())
         data_input_form.append(sample_input_form)
     return data_input_form
 

diff --git a/vissl/data/ssl_transforms/rand_auto_aug.py b/vissl/data/ssl_transforms/rand_auto_aug.py
@@ -69,7 +69,7 @@
 
 
 # Modification/Addition
-@register_transform("VisslRandAugment")
+@register_transform("RandAugment")
 class RandAugment(ClassyTransform):
     """
     Create a RandAugment transform.