Upload train.py with huggingface_hub

train.py

@@ -0,0 +1,87 @@
+from pathlib import Path
+import shutil
+from datasets import load_dataset, concatenate_datasets
+from transformers import TrainingArguments
+from span_marker import SpanMarkerModel, Trainer
+from span_marker.model_card import SpanMarkerModelCardData
+
+import os
+
+os.environ["CODECARBON_LOG_LEVEL"] = "error"
+
+
+def main() -> None:
+    # Load the dataset, ensure "tokens" and "ner_tags" columns, and get a list of labels
+    dataset_id = "midas/inspec"
+    dataset_name = "Inspec"
+    dataset = load_dataset(dataset_id, "extraction")
+    dataset = dataset.rename_columns({"document": "tokens", "doc_bio_tags": "ner_tags"})
+    # Map string labels to integer labels instead
+    real_labels = ["O", "B", "I"]
+    dataset = dataset.map(lambda sample: {"ner_tags": [real_labels.index(tag) for tag in sample]}, input_columns="ner_tags")
+    # Use more readable labels
+    labels = ["O", "B-KEY", "I-KEY"]
+    # Train using train + validation set.
+    train_dataset = concatenate_datasets((dataset["train"], dataset["validation"]))
+
+    # Initialize a SpanMarker model using a pretrained BERT-style encoder
+    encoder_id = "bert-base-uncased"
+    model_id = "tomaarsen/span-marker_bert-base-uncased-keyphrase-inspec"
+    model = SpanMarkerModel.from_pretrained(
+        encoder_id,
+        labels=labels,
+        # SpanMarker hyperparameters:
+        model_max_length=256,
+        marker_max_length=128,
+        entity_max_length=8,
+        # Model card variables
+        model_card_data=SpanMarkerModelCardData(
+            model_id=model_id,
+            encoder_id=encoder_id,
+            dataset_name=dataset_name,
+            dataset_id=dataset_id,
+            license="apache-2.0",
+            language="en",
+        ),
+    )
+
+    # Prepare the 🤗 transformers training arguments
+    output_dir = Path("models") / model_id
+    args = TrainingArguments(
+        output_dir=output_dir,
+        hub_model_id=model_id,
+        run_name=f"bbu_keyphrase",
+        # Training Hyperparameters:
+        learning_rate=5e-5,
+        per_device_train_batch_size=32,
+        per_device_eval_batch_size=32,
+        num_train_epochs=3,
+        weight_decay=0.01,
+        warmup_ratio=0.1,
+        bf16=True,  # Replace `bf16` with `fp16` if your hardware can't use bf16.
+        # Other Training parameters
+        logging_first_step=True,
+        logging_steps=50,
+        evaluation_strategy="no",
+        save_total_limit=2,
+        dataloader_num_workers=2,
+    )
+
+    # Initialize the trainer using our model, training args & dataset, and train
+    trainer = Trainer(
+        model=model,
+        args=args,
+        train_dataset=train_dataset
+    )
+    trainer.train()
+
+    # Compute & save the metrics on the test set
+    metrics = trainer.evaluate(dataset["test"], metric_key_prefix="test")
+    trainer.save_metrics("test", metrics)
+
+    trainer.save_model(output_dir / "checkpoint-final")
+    shutil.copy2(__file__, output_dir / "checkpoint-final" / "train.py")
+
+
+if __name__ == "__main__":
+    main()