LLM_project / README.md
ThuyTran102's picture
Update README.md
93335aa verified
metadata
license: apache-2.0
base_model: distilbert-base-uncased-finetuned-sst-2-english
tags:
  - generated_from_trainer
metrics:
  - accuracy
model-index:
  - name: LLM_project
    results: []

LLM_project

This model is a fine-tuned version of distilbert-base-uncased-finetuned-sst-2-english on IMDb reviews dataset. It achieves the following results on the evaluation set:

  • Loss: 0.0852
  • Accuracy: 0.9804

Model description

This model is a fine-tuned version of the DistilBERT model, which is a smaller, faster, and lighter version of BERT (Bidirectional Encoder Representations from Transformers). The base model has been pre-trained on a large corpus of English data in a self-supervised fashion, and fine-tuning was performed using a sentiment analysis dataset. The model is uncased, meaning it does not distinguish between uppercase and lowercase letters.

DistilBERT retains 97% of BERT's language understanding while being 60% faster and 40% smaller, making it highly efficient for various NLP tasks including sentiment analysis, which this model is specifically tuned for.

Intended uses & limitations

Intended Uses:

Sentiment analysis of English text, particularly for binary classification tasks such as identifying positive and negative sentiments. Can be applied to product reviews, social media posts, customer feedback, etc.

Limitations:

The model's performance is highly dependent on the quality and representativeness of the fine-tuning dataset. May not perform well on text data that is very different from the fine-tuning dataset. Limited by the scope of sentiment analysis and may not capture nuanced sentiments or complex emotions. Not suitable for tasks outside binary sentiment classification without further fine-tuning.

Training and evaluation data

The model was evaluated on a separate validation set that was not seen during training. This evaluation set is also designed for sentiment analysis and includes examples that reflect real-world use cases.

Training procedure

Procedure

  1. Data Preprocessing: Text data was tokenized using the DistilBERT tokenizer, which converts text into a format suitable for the model.
  2. Model Fine-Tuning: The pre-trained DistilBERT model was fine-tuned on the training dataset. Fine-tuning involves adjusting the weights of the model to better fit the sentiment analysis task.
  3. Evaluation: After training, the model was evaluated on the validation set to measure its performance in terms of loss and accuracy.

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 2e-05
  • train_batch_size: 16
  • eval_batch_size: 16
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 100
  • num_epochs: 3

Training results

Training Loss Epoch Step Validation Loss Accuracy
0.0743 1.0 1250 0.1208 0.9696
0.145 2.0 2500 0.0852 0.9804
0.0322 3.0 3750 0.1043 0.9822

Framework versions

  • Transformers 4.41.2
  • Pytorch 2.3.1+cpu
  • Datasets 2.20.0
  • Tokenizers 0.19.1