tinybert

深度学习入门课、资深课、特色课、学术案例、产业实践案例、深度学习知识百科及面试题库The course, case and knowledge of Deep Learning and AI

Build and train state-of-the-art natural language processing models using BERT

Federated Learning performance evaluation with Flower and PyTorch, comparing Full Fine-Tuning and LoRA Fine-Tuning in local and distributed environments.

Sentiment analysis of Reddit posts and comments.

Code of the paper "Emotion Recognition on Edge Devices: Training and Deployment " by Pandelea et al.

To understand the gender/social bias Bert KD model, more specifically TinyBert.

Private, browser-based sentiment analysis using a lightweight TinyBERT model and Transformers.js without backend inference.

A comprehensive comparison of traditional machine learning and transformer-based models for fake news detection, developed as a Bachelor's graduation project at An-Najah National University.

Software TinyBERT Development and Deployment

Detect duplicate questions using NLP techniques including TF-IDF + Logistic Regression, Siamese LSTM, and TinyBERT Transformers. Built using the Quora Question Pairs dataset.

Public evaluation of OnKith's on-device PII span-detection model (TinyBERT-4, 67-label BIO). Entity-level F1, FP32 vs INT8 quantisation trade-off, and a probe showing part of the EMAIL score is domain memorisation. Notebook with committed outputs. No weights.

Twitter bot detection using TinyBERT + DNN — 93.25% accuracy with LIME explainability