Brief Details: A large Chinese text embedding model based on LERT architecture, derived from text2vec-base-chinese. Optimized for semantic understanding and vector representations of Chinese text.
Brief-details: A Llama-1B variant model developed by ai-ml-lab, designed for data processing tasks. Limited documentation available but built on the Llama architecture with 1B parameters.
BRIEF-DETAILS: Legacy version of Vicuna-13B language model by eachadea. Note: Superseded by vicuna-13b-1.1 - historical reference implementation.
Brief Details: A specialized AI model by lodestones hosted on HuggingFace, focused on generating furry-themed content using safetensor format for improved safety and efficiency.
BRIEF DETAILS: Single-step diffusion model using dynamic adversarial training, featuring two variants: Realism and Vibrant. Achieves high-fidelity image generation in 1-4 steps.
Brief Details: A powerful math-specialized 7B parameter model built on Mistral-7B, achieving 77.7% accuracy on GSM8K and 28.2% on MATH benchmarks through MetaMathQA dataset training.
Brief-details: Japanese-focused instruction-tuned 7B parameter LLM by Stability AI, designed for Japanese language understanding and generation tasks with controlled output capabilities.
BRIEF-DETAILS: RusEnQA is a 1.3B parameter QA model for Russian and English, based on rugpt3xl, trained on 80B tokens with sparse attention blocks.
Brief Details: ChineseBERT-base is an innovative language model that enhances Chinese text understanding by combining character embeddings with glyph and pinyin information, offering improved semantic comprehension.
Brief Details: Grappa Large JNT by Salesforce - A research-focused language model designed for academic purposes with strong ethical usage guidelines and safety considerations.
Brief Details: XLM-RoBERTa-based Farsi QA model optimized for Persian question-answering tasks. Supports both pipeline and manual approaches with PyTorch/TensorFlow compatibility.
Brief-details: SikuRoBERTa is a specialized pre-trained language model for ancient Chinese text processing, built on BERT architecture using the Siku Quanshu corpus.
Brief Details: T5-small based model for Python code summarization, trained on tokenized functions. Achieves 8.45 BLEU score, suitable for generating documentation.
Brief Details: CodeTrans T5-Large model for Python code summarization, featuring 220M parameters and multi-task training on 13 supervised tasks. Achieves 13.24 BLEU score for Python.
BRIEF-DETAILS: CodeTrans T5-base model specialized in Java API recommendation generation, achieving 70.45 BLEU score. Built by SEBIS for code understanding tasks.
BRIEF-DETAILS: IndoBERT model fine-tuned on translated SQuAD v2.0 for Indonesian Q&A. Achieves 51.61 EM and 69.09 F1 scores. 420MB model size.
Brief Details: Swedish BERT-based NER model achieving 92% F1-score, specialized for Swedish text analysis with 6 entity categories including Location, Organization, Person, Religion, and Title.
Brief Details: Russian speech recognition model based on wav2vec2-XLS-R-1B, achieving 9.71% WER on Common Voice dataset with extensive fine-tuning
Brief-details: Portuguese BERT-based model for paraphrase detection, achieving 78.09% accuracy. Fine-tuned from neuralmind/bert-base-portuguese-cased.
BRIEF-DETAILS: Spanish RoBERTa-large model fine-tuned for question answering, achieving 82.02% F1 score on SQAC dataset, trained on 570GB BNE corpus
Brief Details: Spanish RoBERTa model trained on 570GB of BNE data. Excels at masked language modeling and NLP tasks. Strong performance on classification and NER.
