[codemod][pyfmt] apply Black 2024 style in fbcode (9/16)

Summary:
Formats the covered files with pyfmt.

paintitblack

Reviewed By: aleivag

Differential Revision: D54447729

fbshipit-source-id: fc781322b254f7027c24888cdadd5f1e90325ba8

mmf/common/registry.py

@@ -35,6 +35,7 @@ class Registry:
     r"""Class for registry object which acts as central source of truth
     for MMF
     """
+
     mapping = {
         # Mappings of builder name to their respective classes
         # Use `registry.register_builder` to register a builder class

mmf/datasets/builders/charades/_utils.py

@@ -21,9 +21,9 @@ def make_charades_df(csv_path, video_dir, classes_file):
 
     # parse action labels
     df["action_labels"] = df["actions"].map(
-        lambda x: [label.split(" ")[0] for label in x.split(";")]
-        if pd.notnull(x)
-        else []
+        lambda x: (
+            [label.split(" ")[0] for label in x.split(";")] if pd.notnull(x) else []
+        )
     )
 
     # load id to class map

mmf/datasets/databases/readers/feature_readers.py

@@ -275,9 +275,7 @@ def read(self, image_feat_path):
         tmp_image_feat = image_feat_bbox.item().get("image_feature")
         image_loc, image_dim = tmp_image_feat.shape
         tmp_image_feat_2 = np.zeros((self.max_loc, image_dim), dtype=np.float32)
-        tmp_image_feat_2[
-            0:image_loc,
-        ] = tmp_image_feat  # noqa
+        tmp_image_feat_2[0:image_loc,] = tmp_image_feat  # noqa
         tmp_image_feat_2 = torch.from_numpy(tmp_image_feat_2)
         tmp_image_box = np.zeros((self.max_loc, 4), dtype=np.int32)
         tmp_image_box[0:image_loc] = image_boxes

mmf/datasets/multi_datamodule.py

@@ -29,9 +29,9 @@ def __init__(self, config: DictConfig):
         self.batch_size = get_batch_size()
 
         self.dataset_list: List[str] = dataset_list_from_config(self.config)
-        self.datamodules: Dict[
-            str, pl.LightningDataModule
-        ] = build_multiple_datamodules(self.dataset_list, self.config.dataset_config)
+        self.datamodules: Dict[str, pl.LightningDataModule] = (
+            build_multiple_datamodules(self.dataset_list, self.config.dataset_config)
+        )
         self.train_loader: Optional[MultiDataLoader] = None
         self.val_loader: Optional[MultiDataLoader] = None
         self.test_loader: Optional[MultiDataLoader] = None

mmf/models/albef/vit.py

@@ -660,9 +660,7 @@ def interpolate_pos_embed(pos_embed_checkpoint, visual_encoder):
         )
         pos_tokens = pos_tokens.permute(0, 2, 3, 1).flatten(1, 2)
         new_pos_embed = torch.cat((extra_tokens, pos_tokens), dim=1)
-        print(
-            "reshape position embedding from %d to %d" % (orig_size**2, new_size**2)
-        )
+        print("reshape position embedding from %d to %d" % (orig_size**2, new_size**2))
 
         return new_pos_embed
     else:

mmf/models/krisp.py

@@ -123,9 +123,9 @@ def build(self):
         # Answer indices not in graph
         if self.config.output_combine == "add":
             self.missing_ans_inds = torch.LongTensor(self.config.num_labels).fill_(1)
-            self.missing_ans_inds[
-                self.graph_module.index_in_ans
-            ] = 0  # Now any index stil set to 1 is missing from graph
+            self.missing_ans_inds[self.graph_module.index_in_ans] = (
+                0  # Now any index stil set to 1 is missing from graph
+            )
 
     # Each model in MMF gets a dict called sample_list which contains
     # all of the necessary information returned from the image

mmf/models/visual_bert.py

@@ -459,9 +459,9 @@ def add_post_flatten_params(
             new_lm_labels = torch.ones_like(attention_mask) * -1
             size_masked_lm_labels = sample_list["masked_lm_labels"].size()
             assert len(size_masked_lm_labels) == 2
-            new_lm_labels[
-                : size_masked_lm_labels[0], : size_masked_lm_labels[1]
-            ] = sample_list["masked_lm_labels"]
+            new_lm_labels[: size_masked_lm_labels[0], : size_masked_lm_labels[1]] = (
+                sample_list["masked_lm_labels"]
+            )
             sample_list["masked_lm_labels"] = new_lm_labels
 
         return sample_list

mmf/modules/losses.py

@@ -880,7 +880,6 @@ def calc_ms_loss(pair, base, param, multiplier):
 
 @registry.register_loss("refiner_ms")
 class RefinerMSLoss(nn.Module):
-
     """
     A Multi-Similarity loss between the decoder outputs of a given embedding size
     and its targets
@@ -949,7 +948,6 @@ def forward(self, sample_list, model_output):
 
 @registry.register_loss("ms_loss")
 class MSLoss(nn.Module):
-
     """
     A Multi-Similarity loss between embeddings of similar and dissimilar
     labels is implemented here.
@@ -1049,7 +1047,6 @@ def forward(self, sample_list, model_output):
 
 @registry.register_loss("refiner_contrastive_loss")
 class RefinerContrastiveLoss(nn.Module):
-
     """
     A contrastive loss between the decoder outputs of a given embedding size
     and its targets

tests/models/transformers/test_heads.py

@@ -247,9 +247,9 @@ def setUp(self):
         )
         self.processed_sample_list = Sample()
         feat_targets = torch.zeros((bs, num_feat, img_dim))
-        self.processed_sample_list[
-            "mrfr_region_target"
-        ] = feat_targets.contiguous().view(-1, img_dim)
+        self.processed_sample_list["mrfr_region_target"] = (
+            feat_targets.contiguous().view(-1, img_dim)
+        )
         self.processed_sample_list["mrfr_region_mask"] = torch.ones(
             (bs, num_feat)
         ).bool()