114 repos across 6 sub-areas
Deep learning models and frameworks for natural language processing across multiple languages, with heavy emphasis on transformer-based architectures (BERT variants, RoBERTa) compatible with PyTorch, TensorFlow, and JAX. The cluster centers on pre-trained language models for non-English languages (Chinese, Russian, and others), along with the supporting infrastructure for fine-tuning, deployment, and serving these models via standard endpoints. Developers exploring this area will find model checkpoints, training pipelines, and cross-framework compatibility layers for building multilingual NLP applications.
Multilingual BERT Models
34 repos
Pre-trained BERT language models fine-tuned for non-English languages including Italian, Swedish, Turkish, and others. These repositories provide transformer-based NLP models optimized for specific language processing tasks, leveraging PyTorch, TensorFlow, and JAX frameworks. The cluster focuses on democratizing access to high-quality multilingual NLP capabilities across diverse language communities.
Chinese Language BERT Models
32 repos
Pre-trained transformer models and utilities optimized for Chinese language processing, building on BERT and RoBERTa architectures. The cluster centers on several major Chinese-specific BERT variants (MacBERT, RoBERTa-wwm, BERT-wwm) alongside supporting libraries for tokenization, fine-tuning, and deployment across PyTorch and TensorFlow. Repositories here focus on adapting transformer models to Chinese linguistic characteristics and providing inference endpoints for production use.
Chinese BERT Language Models
12 repos
Pre-trained BERT and transformer-based models optimized for Chinese NLP tasks, including variants like RoBERTa-WWM, MacBERT, and specialized architectures from the community. These repositories provide masked language modeling and fill-mask capabilities specifically tuned for Chinese text processing, along with PyTorch implementations compatible with the Hugging Face transformers library. Repositories here serve as foundation models for downstream Chinese language understanding tasks.
BERT Language Models & Variants
12 repos
Pre-trained BERT transformer models optimized for various languages and use cases, primarily implemented in PyTorch and JAX. This cluster contains multilingual BERT implementations including IndoBERT for Indonesian, BanglaBERT for Bengali, and KcBERT for Korean, alongside base English BERT models. The repositories focus on providing ready-to-use transformer checkpoints and their integration with modern deep learning frameworks, useful for transfer learning and fine-tuning on downstream NLP tasks.
Transformer Models and Language Models
12 repos
Libraries, frameworks, and pre-trained models for building and deploying transformer-based neural networks, with heavy emphasis on PyTorch and JAX implementations. This cluster centers on large language models and BERT-style architectures, including multilingual variants like Bangla-BERT and IndoBERT alongside larger models like OPT. Repositories here span training infrastructure, model serving endpoints, and the core deep learning frameworks needed to work with modern NLP systems.
Twitter NLP & Text Classification
12 repos
Pre-trained transformer models for analyzing tweets and social media text, covering sentiment, emotion, hate speech, offensive language, irony, and emoji interpretation. These repositories provide fine-tuned RoBERTa checkpoints and compatible inference endpoints built on PyTorch and JAX, enabling practitioners to quickly deploy text classification models optimized for the linguistic patterns and noise characteristics of social media.