HuggingFaceTB/python-edu-scorer

Python-Edu Scorer

This model is a fine-tuned version of Snowflake/snowflake-arctic-embed-m on a dataset of Python files labeled by Llama3 for educational value. We used this classifier to build the Python-Edu dataset.

How to use in transformers

To load the Python-Edu classifier, use the following code:

from transformers import AutoTokenizer, AutoModelForSequenceClassification

tokenizer = AutoTokenizer.from_pretrained("HuggingFaceTB/python-edu-scorer")
model = AutoModelForSequenceClassification.from_pretrained("HuggingFaceTB/python-edu-scorer")

text = "This is a test sentence."
inputs = tokenizer(text, return_tensors="pt", padding="longest", truncation=True)
outputs = model(**inputs)
logits = outputs.logits.squeeze(-1).float().detach().numpy()
score = logits.item()
result = {
    "text": text,
    "score": score,
    "int_score": int(round(max(0, min(score, 5)))),
}

print(result)
# {'text': 'This is a test sentence.', 'score': 0.07964489609003067, 'int_score': 0}

Intended uses & limitations

While the Python-Edu classifier performs well in distinguishing high-quality python code, there are some limitations:

• Scope: The model's performance might change for other datasets, in particular for out of distribution samples. It is also focused on educational content relevant to beginners and may not perform as well on content intended for higher education or specialized domains.
• Bias: The model's performance is dependent on the quality and representativeness of the training data and the LLM used for the annotation. Biases in both can affect the classifier's judgments. It might overfit to thoroughly commented code.
• Context: The classifier evaluates individual code files without considering broader context, which might impact its effectiveness in certain scenarios.

The training and inference code is available on GitHub https://github.com/huggingface/cosmopedia/tree/main/classification

Training procedure

The classifier was trained on 450,000 pairs of python code files and their scores from 1 to 5, generated by Llama3. The samples were annotated based on their educational quality with 1 being not educational and 5 being highly educational.

We added a classification head with a single regression output to Snowflake-arctic-embed and trained the model for 20 epochs with a learning rate of 3e-4. During training, the embedding and encoder layers were frozen to focus on the classification head.

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0003
• train_batch_size: 256
• eval_batch_size: 128
• seed: 0
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 20

Training results

              precision    recall  f1-score   support

           1       0.84      0.46      0.59      8364
           2       0.61      0.76      0.68     19605
           3       0.60      0.62      0.61     16187
           4       0.72      0.50      0.59      4872
           5       0.38      0.08      0.13       118

    accuracy                           0.64     49146
   macro avg       0.63      0.48      0.52     49146
weighted avg       0.66      0.64      0.63     49146

Framework versions

• Transformers 4.40.1
• Pytorch 2.2.0+cu121
• Datasets 2.17.1
• Tokenizers 0.19.1