glimpse-ui/scibert/scibert_topic/scibert_topic_train.py

import pandas as pd
from datasets import Dataset
from transformers import AutoTokenizer, AutoModelForSequenceClassification, Trainer, TrainingArguments
from sklearn.metrics import accuracy_score, precision_recall_fscore_support
import numpy as np

# Load data
dev_df = pd.read_csv("./data/DISAPERE-main/SELFExtractedData/disapere_topic_dev.csv")
train_df = pd.read_csv("./data/DISAPERE-main/SELFExtractedData/disapere_topic_train.csv")
test_df = pd.read_csv("./data/DISAPERE-main/SELFExtractedData/disapere_topic_test.csv")

# Convert to HuggingFace Datasets
train_ds = Dataset.from_pandas(train_df)
dev_ds = Dataset.from_pandas(dev_df)
test_ds = Dataset.from_pandas(test_df)

# Tokenize
model_name = "allenai/scibert_scivocab_uncased"
tokenizer = AutoTokenizer.from_pretrained(model_name)

def tokenize(batch):
    return tokenizer(batch["text"], padding="max_length", truncation=True, max_length=256)

train_ds = train_ds.map(tokenize, batched=True)
dev_ds = dev_ds.map(tokenize, batched=True)
test_ds = test_ds.map(tokenize, batched=True)

# Set format for PyTorch
train_ds.set_format(type="torch", columns=["input_ids", "attention_mask", "label"])
dev_ds.set_format(type="torch", columns=["input_ids", "attention_mask", "label"])
test_ds.set_format(type="torch", columns=["input_ids", "attention_mask", "label"])

print(train_df['label'].value_counts().sort_index())


# Load model
model = AutoModelForSequenceClassification.from_pretrained(model_name, num_labels=8)

# Metrics
def compute_metrics(eval_pred):
    logits, labels = eval_pred
    preds = np.argmax(logits, axis=1)
    precision, recall, f1, _ = precision_recall_fscore_support(labels, preds, average="macro")
    acc = accuracy_score(labels, preds)
    return {"accuracy": acc, "f1": f1, "precision": precision, "recall": recall}

# Training arguments
args = TrainingArguments(
    output_dir="./scibert/scibert_topic/checkpoints",
    eval_strategy="epoch",
    save_strategy="epoch",
    learning_rate=2e-5,
    per_device_train_batch_size=8,
    per_device_eval_batch_size=16,
    num_train_epochs=4,
    weight_decay=0.01,
    load_best_model_at_end=True,
    metric_for_best_model="f1"
)

# Trainer
trainer = Trainer(
    model=model,
    args=args,
    train_dataset=train_ds,
    eval_dataset=dev_ds,
    tokenizer=tokenizer,
    compute_metrics=compute_metrics
)

# Train
trainer.train()

# Evaluate on test
results = trainer.evaluate(test_ds)
print("Test results:", results)

# Save the model and tokenizer
model.save_pretrained("./scibert/scibert_topic/final_model")
tokenizer.save_pretrained("./scibert/scibert_topic/final_model")