Brief Details: A specialized translation model converting Modern Chinese to Classical Chinese (文言文), trained on 900K+ sentence pairs with encoder-decoder architecture and Transformer-based implementation.
Brief-details: A specialized translation model that converts Classical Chinese (文言文) to Modern Chinese, trained on 900k+ sentence pairs with 50% punctuation removal during training.
Brief Details: A specialized NLP model for adding punctuation to Classical Chinese texts, leveraging NER techniques. Created by raynardj with 1,127 downloads and 22 likes, particularly useful for ancient Chinese text processing.
Brief Details: Portuguese clinical NER model specialized in disease entity detection, trained on SemClinBr corpus using BioBERTpt architecture. Part of a larger clinical NLP project.
Brief-details: A Portuguese BERT model specialized for clinical NLP tasks, trained on Brazilian hospital records. Optimized for clinical named entity recognition and biomedical text analysis.
Brief-details: T5-base Portuguese QA model fine-tuned on SQuAD v1.1, achieving 79.3 F1 score. Optimized for question-answering tasks in Portuguese with 67.4% exact match accuracy.
BRIEF DETAILS: Portuguese GPT-2 small model (124M params) trained on Wikipedia data, achieving 37.99% accuracy and 23.76 perplexity. MIT licensed, supports text generation tasks.
BRIEF DETAILS: Portuguese ByT5 small model fine-tuned for QA tasks on SQuAD v1.1. Tokenizer-free architecture optimized for Portuguese question-answering, particularly effective on noisy text.
Brief Details: Portuguese BERT-based QA model fine-tuned on SQUAD v1.1, achieving 82.50 F1 score. Specialized for Portuguese language question answering tasks.
Brief-details: A fine-tuned wav2vec2 model specialized for Turkish speech recognition, achieving 48% WER, built on XLSR-53 architecture with Apache 2.0 license
Brief Details: Turkish text summarization model based on mT5-small (300M params), fine-tuned on MLSUM dataset. Optimized for news article compression.
Brief Details: Vietnamese accent prediction model based on XLM-RoBERTa, achieving 97% accuracy for inserting diacritical marks in Vietnamese text.
Brief-details: Turkish question-answering model based on mT5-small, fine-tuned on TQUAD dataset. 300M parameters, MIT license, optimized for Turkish Q&A tasks.
BRIEF-DETAILS: A transformer-based model for image generation and feature extraction, trained on ImageNet-21k with 32x32 resolution capabilities and color-cluster tokenization
Brief Details: AlephBERT - State-of-the-art Hebrew BERT model trained on 87M+ sentences from OSCAR, Wikipedia & Twitter. Apache-2.0 licensed.
Brief-details: ImageGPT-large is a transformer-based vision model trained on ImageNet-21k for pixel prediction and image generation at 32x32 resolution
Brief-details: SegFormer B5 model fine-tuned for semantic segmentation on Cityscapes dataset at 1024x1024 resolution. Features hierarchical Transformer encoder and MLP decode head.
Brief-details: NVIDIA's Megatron-BERT variant with 345M parameters, featuring 24 layers and 16 attention heads. Trained on diverse text sources for masked language modeling and next sentence prediction.
Brief Details: NVIDIA's 345M parameter BERT-style transformer with 24 layers, trained on diverse text sources. Supports masked LM and next sentence prediction.
Brief Details: Legal document summarization model based on PEGASUS, fine-tuned on SEC litigation data. Achieves 57.39% ROUGE-1 score with 1024 token limit.
Brief-details: BERT model specialized for legal contracts, pre-trained on 76K US contracts from EDGAR database. Part of LEGAL-BERT family with 110M parameters.
