Back out "[PyText] cache lm packed data" (facebookresearch#454)

Summary:
Pull Request resolved: facebookresearch#454

Original commit changeset: 05b0880e6d0c

Reviewed By: borguz

Differential Revision: D14807451

fbshipit-source-id: dafbc76b4660eeb296f76eff9ce14b94528d6390

pytext/main.py

@@ -36,7 +36,6 @@
     export_saved_model_to_torchscript,
     get_logits as workflow_get_logits,
     prepare_task_metadata,
-    preprocess_task,
     test_model_from_snapshot_path,
     train_model,
 )
@@ -314,7 +313,6 @@ def train(context):
     if config.use_tensorboard:
         metric_channels.append(TensorBoardChannel())
     try:
-        preprocess_task(config)
         if config.distributed_world_size == 1:
             train_model(config, metric_channels=metric_channels)
         else:

pytext/workflow.py

@@ -119,15 +119,6 @@ def prepare_task(
     return task
 
 
-def preprocess_task(config: PyTextConfig):
-    if hasattr(config.task, "data") and hasattr(config.task.data, "numberized_dir"):
-        if config.load_snapshot_path and os.path.isfile(config.load_snapshot_path):
-            task = load(config.load_snapshot_path)
-        else:
-            task = create_task(config.task)
-        task.data.initialize_numberized_data()
-
-
 def save_and_export(
     config: PyTextConfig, task: Task, metric_channels: Optional[List[Channel]] = None
 ) -> None: