Brief-details: Russian BERT-based language model trained on 30GB data, optimized for mask-filling tasks with 178M parameters and BPE tokenization
Brief Details: Wav2Vec2 Swedish ASR model with 315M params. Achieves 8.49% WER on Common Voice. Fine-tuned on radio broadcasts, NST and Common Voice data.
BRIEF DETAILS: T5-based multilingual translation model supporting English, Russian and Chinese pairs with 111M parameters. Efficient for direct translation between language pairs using prefix-based targeting.
Brief-details: Quantized 8B parameter LLaMA-3.1 model optimized for multilingual dialogue, running in INT4 precision requiring only 4GB VRAM for efficient deployment
Brief-details: A FLUX-based LoRA model specializing in Kodachrome-style image generation with warm tones and nostalgic aesthetics, popular with 27.6K+ downloads
Brief-details: FinBERT-FLS is a specialized BERT model for classifying forward-looking statements in financial texts, fine-tuned on 3,500 annotated sentences from corporate reports.
Brief-details: Korean to English translation model by Helsinki-NLP achieving 41.3 BLEU score, using transformer-align architecture with SentencePiece tokenization
Brief Details: A specialized LoRA model for SDXL that transforms images into pixel art style, optimized for both isometric and non-isometric artwork with simplified implementation
Brief-details: RealVisXL V5.0 Lightning is a photorealistic text-to-image SDXL model optimized for fast inference, supporting both SFW/NSFW with recommended low CFG (1.0-2.0) settings
BRIEF DETAILS: A compact question-answering model (33.4M params) fine-tuned on SQuAD 2.0, achieving 76.19% exact match accuracy. Ideal for extractive QA tasks.
Brief-details: RealVisXL V5.0 is a powerful text-to-image SDXL model focused on photorealism, supporting both SFW/NSFW content with advanced rendering capabilities and optimized parameters.
BRIEF DETAILS: T5-based prompt enhancement model (77M params) that expands brief text prompts into detailed descriptions, optimized for text-to-image generation.
Brief Details: A specialized BERT model for Chinese Named Entity Recognition (NER), developed by CKIPLAB. Features traditional Chinese support with 27.8K+ downloads and GPLv3 license.
Brief Details: Persian speech gender recognition model using HuBERT architecture. Achieves 98% accuracy with high precision for both male/female classification.
Brief Details: A fine-tuned DistilRoBERTa model for detecting AI rejection responses, achieving 98.87% accuracy with 82.1M parameters. Optimized for content moderation.
Brief-details: Korean-specific Sentence-BERT model optimized for semantic similarity tasks, achieving 86.28% accuracy on document classification with 768-dimensional embeddings.
Brief Details: An 8B parameter uncensored LLaMA 3.1 variant using abliteration technique, achieving 73.29% on IFEval with strong instruction-following capabilities
Brief-details: Multilingual Named Entity Recognition model using bidirectional transformer architecture, capable of identifying flexible entity types with 209M parameters.
Brief Details: Juggernaut-X-v10 is a powerful SDXL-based text-to-image model optimized for diverse content generation, featuring enhanced prompt adherence and GPT-4 Vision captioning capabilities.
Brief Details: Optimized 12.2B parameter Mistral model quantized to FP8, offering 50% memory reduction while maintaining 99.53% performance of original model.
BRIEF DETAILS: A sentence embedding model based on DistilRoBERTa, maps text to 768-dimensional vectors, optimized for semantic similarity tasks with 82.1M parameters.
