Candle-PyTorch Parity Testing

This repository compares the implementations of various NLP models in Candle against PyTorch equivalents.

Project Structure

The project is organized by model/task type:

• bert-embeddings/: BERT embedding generation comparison
• xlm-roberta-finetuned/: compares inference on s-nlp/xlmr_formality_classifier - XLM RoBERTa fine-tuned on the multilingual text classification task of determining formality of sentences

Each model directory contains:

• Subdirectories for each framework implementation (candle/ and pytorch/)
• READMEs with specific implementation details
• Some results and discussion