Convert directory fbcode/hammer to use the Ruff Formatter

generative_recommenders/data/dataset.py

@@ -192,7 +192,6 @@ def _truncate_or_pad_seq(
 
 
 class MultiFileDatasetV2(DatasetV2, torch.utils.data.Dataset):
-
     def __init__(
         self,
         file_prefix: str,

generative_recommenders/data/preprocessor.py

@@ -430,10 +430,14 @@ def preprocess_rating(self) -> int:
         return num_unique_items
 
 
-def get_common_preprocessors() -> Dict[
-    str,
-    Union[AmazonDataProcessor, MovielensDataProcessor, MovielensSyntheticDataProcessor],
-]:
+def get_common_preprocessors() -> (
+    Dict[
+        str,
+        Union[
+            AmazonDataProcessor, MovielensDataProcessor, MovielensSyntheticDataProcessor
+        ],
+    ]
+):
     ml_1m_dp = MovielensDataProcessor(  # pyre-ignore [45]
         "http://files.grouplens.org/datasets/movielens/ml-1m.zip",
         "tmp/movielens1m.zip",

generative_recommenders/indexing/candidate_index.py

@@ -23,7 +23,6 @@
 
 
 class TopKModule(torch.nn.Module):
-
     @abc.abstractmethod
     def forward(
         self,
@@ -44,7 +43,6 @@ def forward(
 
 
 class CandidateIndex(object):
-
     def __init__(
         self,
         ids: torch.Tensor,
@@ -226,7 +224,9 @@ def get_top_k_outputs(
                 else self.embeddings
             )
             top_k_embeddings = batch_gather_embeddings(
-                rowwise_indices=top_k_indices, embeddings=expanded_embeddings  # pyre-ignore[10]
+                # pyre-fixme[10]: Name `top_k_indices` is used but not defined.
+                rowwise_indices=top_k_indices,
+                embeddings=expanded_embeddings,  # pyre-ignore[10]
             )
         else:
             top_k_embeddings = None

generative_recommenders/indexing/mips_top_k.py

@@ -22,7 +22,6 @@
 
 
 class MIPSTopKModule(TopKModule):
-
     def __init__(
         self,
         item_embeddings: torch.Tensor,
@@ -40,7 +39,6 @@ def __init__(
 
 
 class MIPSBruteForceTopK(MIPSTopKModule):
-
     def __init__(
         self,
         item_embeddings: torch.Tensor,

generative_recommenders/modeling/sequential/autoregressive_losses.py

@@ -27,7 +27,6 @@
 
 
 class NegativesSampler(torch.nn.Module):
-
     def __init__(self, l2_norm: bool, l2_norm_eps: float) -> None:
         super().__init__()
 
@@ -72,7 +71,6 @@ def forward(
 
 
 class LocalNegativesSampler(NegativesSampler):
-
     def __init__(
         self,
         num_items: int,
@@ -124,7 +122,6 @@ def forward(
 
 
 class InBatchNegativesSampler(NegativesSampler):
-
     def __init__(
         self,
         l2_norm: bool,
@@ -205,7 +202,6 @@ def forward(
 
 
 class AutoregressiveLoss(torch.nn.Module):
-
     @abc.abstractmethod
     def jagged_forward(
         self,

generative_recommenders/modeling/sequential/embedding_modules.py

@@ -22,7 +22,6 @@
 
 
 class EmbeddingModule(torch.nn.Module):
-
     @abc.abstractmethod
     def debug_str(self) -> str:
         pass
@@ -38,7 +37,6 @@ def item_embedding_dim(self) -> int:
 
 
 class LocalEmbeddingModule(EmbeddingModule):
-
     def __init__(
         self,
         num_items: int,
@@ -74,7 +72,6 @@ def item_embedding_dim(self) -> int:
 
 
 class CategoricalEmbeddingModule(EmbeddingModule):
-
     def __init__(
         self,
         num_items: int,

generative_recommenders/modeling/sequential/hstu.py

@@ -47,7 +47,6 @@
 
 
 class RelativeAttentionBiasModule(torch.nn.Module):
-
     @abc.abstractmethod
     def forward(
         self,
@@ -63,7 +62,6 @@ def forward(
 
 
 class RelativePositionalBias(RelativeAttentionBiasModule):
-
     def __init__(self, max_seq_len: int) -> None:
         super().__init__()
 
@@ -445,7 +443,6 @@ def forward(  # pyre-ignore [3]
 
 
 class HSTUJagged(torch.nn.Module):
-
     def __init__(
         self,
         modules: List[SequentialTransductionUnitJagged],

generative_recommenders/modeling/sequential/input_features_preprocessors.py

@@ -24,7 +24,6 @@
 
 
 class InputFeaturesPreprocessorModule(torch.nn.Module):
-
     @abc.abstractmethod
     def debug_str(self) -> str:
         pass
@@ -43,7 +42,6 @@ def forward(
 class LearnablePositionalEmbeddingInputFeaturesPreprocessor(
     InputFeaturesPreprocessorModule
 ):
-
     def __init__(
         self,
         max_sequence_len: int,
@@ -94,7 +92,6 @@ def forward(
 class LearnablePositionalEmbeddingRatedInputFeaturesPreprocessor(
     InputFeaturesPreprocessorModule
 ):
-
     def __init__(
         self,
         max_sequence_len: int,
@@ -156,7 +153,6 @@ def forward(
 
 
 class CombinedItemAndRatingInputFeaturesPreprocessor(InputFeaturesPreprocessorModule):
-
     def __init__(
         self,
         max_sequence_len: int,

generative_recommenders/modeling/sequential/losses/sampled_softmax.py

@@ -27,7 +27,6 @@
 
 
 class SampledSoftmaxLoss(AutoregressiveLoss):
-
     def __init__(
         self,
         num_to_sample: int,

generative_recommenders/modeling/sequential/output_postprocessors.py

@@ -21,7 +21,6 @@
 
 
 class OutputPostprocessorModule(torch.nn.Module):
-
     @abc.abstractmethod
     def debug_str(self) -> str:
         pass
@@ -35,7 +34,6 @@ def forward(
 
 
 class L2NormEmbeddingPostprocessor(OutputPostprocessorModule):
-
     def __init__(
         self,
         embedding_dim: int,
@@ -60,7 +58,6 @@ def forward(
 
 
 class LayerNormEmbeddingPostprocessor(OutputPostprocessorModule):
-
     def __init__(
         self,
         embedding_dim: int,

generative_recommenders/modeling/sequential/sasrec.py

@@ -305,6 +305,13 @@ def predict(
         batch_id: Optional[int] = None,
     ) -> torch.Tensor:
         return self.interaction(
-            self.encode(past_ids, past_ratings, past_timestamps, next_timestamps),  # pyre-ignore [6]
+            self.encode(
+                # pyre-fixme[6]: For 4th argument expected `Dict[str, Tensor]` but
+                #  got `Tensor`.
+                past_ids,
+                past_ratings,
+                past_timestamps,
+                next_timestamps,
+            ),  # pyre-ignore [6]
             target_ids,
         )  # [B, X]

generative_recommenders/modeling/similarity/mol.py

@@ -15,9 +15,10 @@
 # pyre-unsafe
 
 """
-Implements MoL (Mixture-of-Logits) in 
+Implements MoL (Mixture-of-Logits) in
 Revisiting Neural Retrieval on Accelerators (https://arxiv.org/abs/2306.04039, KDD'23).
 """
+
 from typing import Callable, Dict, Optional, Tuple
 
 import torch
@@ -29,7 +30,6 @@
 
 
 class SoftmaxDropout(torch.nn.Module):
-
     def __init__(
         self,
         dropout_rate: float,
@@ -51,7 +51,6 @@ def forward(
 
 
 class SoftmaxDropoutCombiner(torch.nn.Module):
-
     def __init__(
         self,
         dropout_rate: float,
@@ -110,7 +109,6 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:
 
 
 class TauFn(torch.nn.Module):
-
     def __init__(
         self,
         alpha: float,
@@ -132,7 +130,6 @@ def forward(
 
 
 class GeGLU(torch.nn.Module):
-
     def __init__(
         self,
         in_features: int,
@@ -201,7 +198,6 @@ def forward(self, x: torch.Tensor) -> torch.Tensor:
 
 
 class MoLGatingFn(torch.nn.Module):
-
     def __init__(
         self,
         num_logits: int,
@@ -291,7 +287,10 @@ def forward(
         if self._item_only_partial_module is not None:
             if item_sideinfo is not None:
                 item_embeddings = torch.cat([item_embeddings, item_sideinfo], dim=-1)
-            item_partial_inputs = self._item_only_partial_module(item_embeddings)  # pyre-ignore [29]
+            # pyre-fixme[29]: `Optional[Module]` is not a function.
+            item_partial_inputs = self._item_only_partial_module(
+                item_embeddings
+            )  # pyre-ignore [29]
         if self._ci_partial_module is not None:
             if self._combine_item_sideinfo_into_ci:
                 assert item_sideinfo is not None

generative_recommenders/modeling/similarity_utils.py

@@ -133,7 +133,10 @@ def create_mol_interaction_module(
                 ).apply(init_mlp_xavier_weights_zero_bias)
             )
         ),
-        gating_context_only_partial_fn=lambda input_dim, output_dim: (  # pyre-ignore [6]
+        # pyre-fixme[6]: For 12th argument expected `Optional[(int, int) -> Module]`
+        #  but got `(input_dim: Any, output_dim: Any) -> Optional[Sequential]`.
+        gating_context_only_partial_fn=lambda input_dim,
+        output_dim: (  # pyre-ignore [6]
             torch.nn.Sequential(
                 torch.nn.Linear(
                     in_features=input_dim,