Add batch size to usage_example and move tokenize_function into main

src/genbench/tasks/latent_feature_splits/usage_example.py

@@ -8,11 +8,6 @@
 import os
 
 
-def tokenize_function(example):
-    return tokenizer(
-        example["input"])
-
-
 def compute_metrics(eval_preds):
     metric = evaluate.load("f1")
     logits, labels = eval_preds
@@ -23,16 +18,22 @@ def compute_metrics(eval_preds):
         average="macro")
 
 
-def main(split_name, num_labels, lr, epochs, checkpoint):
+def main(split_name, num_labels, bsz, lr, epochs, checkpoint):
     """
     Basic functionality to load data, train and evaluate the model.
     Args:
         - split_name: str (bert_closest_split | roberta_closest_split)
         - num_labels (int)
+        - bsz (int): batch size
         - lr (float): learning rate
         - epochs (int): number of epochs
         - checkpoint (str): should be a valid HF model name
     """
+
+    def tokenize_function(example):
+        return tokenizer(
+            example["input"])
+
     # Convert GenBench format to HF dataset format, preview dataset
     task = load_task(f"latent_feature_splits:{split_name}")
     ds = task.get_prepared_datasets(PreparationStrategy.FINETUNING)
@@ -42,7 +43,8 @@ def main(split_name, num_labels, lr, epochs, checkpoint):
 
     # Load and preprocess data
     tokenizer = AutoTokenizer.from_pretrained(checkpoint)
-    tokenized_datasets = ds.map(tokenize_function, batched=True)
+    tokenized_datasets = ds.map(
+        tokenize_function, batch_size=bsz, batched=True)
     data_collator = DataCollatorWithPadding(tokenizer=tokenizer)
 
     # Load model and HF trainer, WITH evaluation during training
@@ -52,6 +54,8 @@ def main(split_name, num_labels, lr, epochs, checkpoint):
         "test-trainer",
         learning_rate=lr,
         num_train_epochs=epochs,
+        per_device_train_batch_size=bsz,
+        per_device_eval_batch_size=bsz,
         evaluation_strategy="epoch")
     trainer = Trainer(
         model,
@@ -77,8 +81,9 @@ def main(split_name, num_labels, lr, epochs, checkpoint):
     os.environ["WANDB_DISABLED"] = "true"
     split_name = "bert_closest_split"
     num_labels = 3
+    batch_size = 16
     lr = 3e-5
     epochs = 5
     checkpoint = "prajjwal1/bert-small"
 
-    main(split_name, num_labels, lr, epochs, checkpoint)
+    main(split_name, num_labels, batch_size, lr, epochs, checkpoint)