Model Card for Bert-base-cased Paraphrase Classification

Model Details

Model Description

The bert-base-cased-paraphrase-classification model is a fine-tuned version of the BERT (Bidirectional Encoder Representations from Transformers) architecture specifically designed for paraphrase classification. It uses the cased variant of BERT as the base model. This model has been fine-tuned for identifying whether two input sentences are paraphrases of each other.

Developed by: Rushil Jariwala
Model type: Transformer-based neural network
Language(s) (NLP): English
License: Apache 2.0
Finetuned from model: BERT-base-cased

Model Sources

Repository: Hugging Face Model Hub

Uses

Direct Use

This model can directly classify whether two sentences are paraphrases of each other.

Downstream Use

When fine-tuned on a specific task or integrated into a larger application, this model can assist in tasks requiring paraphrase identification.

Out-of-Scope Use

This model may not perform optimally on sentences with highly domain-specific vocabulary not seen during training, and it is limited to the English language.

Bias, Risks, and Limitations

This model's performance may vary based on the similarity of sentences to those in the training data. It may exhibit biases based on the dataset used for training.

Recommendations

Users should consider domain-specific fine-tuning for optimal performance in specific applications. Additionally, careful evaluation and validation are recommended for critical applications.

How to Get Started with the Model

Use the following Python code to get started with the model:

from transformers import pipeline

pipe = pipeline("text-classification", model="rushilJariwala/bert-base-cased-paraphrase-classification")

sequences = [
    "I've been waiting for a HuggingFace course my whole life.",
    "This course is amazing!",
]

result = pipe(sequences)
print(result)


#### Preprocessing
The text was tokenized using BERT's cased tokenizer with truncation and padding.

#### Training Hyperparameters

- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
- Batch Size: 8
- Learning Rate: 5e-5
- Optimizer: AdamW
- Number of Epochs: 3

#### Testing Data

The model was evaluated on the MRPC validation set.
#### Metrics

Accuracy: 86.27%

#### Summary

The model achieved an accuracy of 86.27% on the MRPC validation set.