Grammar changes and spacings

README.md

@@ -6,7 +6,6 @@ Currently these are:
 
 2. TwHIN embeddings (projects/twhin) https://arxiv.org/abs/2202.05387
 
-
 This project can be run inside a python virtualenv. We have only tried this on Linux machines and because we use torchrec it works best with an Nvidia GPU. To setup run
 
 `./images/init_venv.sh` (Linux only).

checks.json

@@ -0,0 +1 @@
+{"enabled":true,"categories":{}}

common/run_training.py

@@ -32,7 +32,7 @@ def maybe_run_training(
   `train_fn(**training_kwargs)`.
 
   Otherwise, this function calls torchrun and points at the calling module
-  `module_name`.  After this call, the necessary environment variables are set
+  `module_name`. After this call, the necessary environment variables are set
   and training will commence.
 
   Args:

machines/list_ops.py

@@ -1,7 +1,7 @@
 """
 Simple str.split() parsing of input string
 
-usage example:
+Usage example:
   python list_ops.py --input_list=$INPUT [--sep=","] [--op=<len|select>] [--elem=$INDEX]
 
 Args:

projects/home/recap/data/preprocessors.py

@@ -51,7 +51,7 @@ class TruncateAndSlice(tf.keras.Model):
 
 
 class DownCast(tf.keras.Model):
-  """Class for Down casting dataset before serialization and transferring to training host.
+  """Class for Down-casting dataset before serialization and transferring to training host.
   Depends on the data type and the actual data range, the down casting can be lossless or not.
   It is strongly recommended to compare the metrics before and after down casting.
   """

projects/home/recap/data/util.py

@@ -9,7 +9,7 @@ def keyed_tensor_from_tensors_dict(
   tensor_map: Mapping[str, torch.Tensor]
 ) -> "torchrec.KeyedTensor":
   """
-  Convert a dictionary of torch tensor to torchrec keyed tensor
+  Converts a dictionary of torch tensor to torchrec keyed tensor
   Args:
     tensor_map:
 
@@ -40,7 +40,7 @@ def _compute_jagged_tensor_from_tensor(tensor: torch.Tensor) -> Tuple[torch.Tens
 
 def jagged_tensor_from_tensor(tensor: torch.Tensor) -> "torchrec.JaggedTensor":
   """
-  Convert a torch tensor to torchrec jagged tensor.
+  Converts a torch tensor to torchrec jagged tensor.
   Note: Currently only support shape of [Batch_size] or [Batch_size x N] for dense tensors.
         For sparse tensor the shape of .values() should be [Batch_size] or [Batch_size x N]; the
           dense_shape of the sparse tensor can be arbitrary.

projects/home/recap/model/config.py

@@ -82,7 +82,7 @@ class ZScoreLogConfig(base_config.BaseConfig):
   analysis_path: str
   schema_path: str = pydantic.Field(
     None,
-    description="Schema path which feaure statistics are generated with. Can be different from scehma in data config.",
+    description="Schema path which feature statistics are generated with. Can be different from scehma in data config.",
   )
   clip_magnitude: float = pydantic.Field(
     5.0, description="Threshold to clip the normalized input values."

projects/home/recap/model/entrypoint.py

@@ -26,7 +26,7 @@ def unsanitize(sanitized_task_name):
 
 
 def _build_single_task_model(task: model_config_mod.TaskModel, input_shape: int):
-  """ "Builds a model for a single task"""
+  """Builds a model for a single task"""
   if task.mlp_config:
     return mlp.Mlp(in_features=input_shape, mlp_config=task.mlp_config)
   elif task.dcn_config:

projects/home/recap/model/model_and_loss.py

@@ -15,7 +15,7 @@ class ModelAndLoss(torch.nn.Module):
     Args:
       model: torch module to wrap.
       loss_fn: Function for calculating loss, should accept logits and labels.
-      straitifiers: mapping of stratifier name and index of discrete features to emit for metrics stratification.
+      straitifiers: mapping of a stratifier name and index of discrete features to emit for metrics stratification.
     """
     super().__init__()
     self.model = model