BRIEF-DETAILS: Facebook's Yoruba text-to-speech model using VITS architecture. Part of Massively Multilingual Speech project. Supports end-to-end speech synthesis with non-deterministic output.
Brief-details: T5-small-based text-to-SQL model that converts natural language queries to SQL, supporting multiple tables and joins. Lightweight and production-ready.
BRIEF-DETAILS: RWKV-7 preview model repository featuring cutting-edge training checkpoints and single-round Q&A states, developed by BlinkDL
Brief Details: DarkBERT - A specialized BERT model requiring academic verification for access, designed for research purposes with strict usage guidelines and institutional requirements.
BRIEF DETAILS: SDXL 0.9 is Stability AI's advanced text-to-image model, representing a significant upgrade in image generation capabilities with enhanced quality and detail compared to previous versions.
BRIEF DETAILS: 4x image upscaler based on Remacri architecture, optimized for Stable Diffusion WebUI integration. Provides 4x upscaling with balanced detail preservation.
Brief-details: Paraformer-large is a non-autoregressive speech recognition model trained on 60,000 hours of Mandarin data, offering 10x faster inference than traditional models while maintaining accuracy
Brief-details: DIPPER is an 11B parameter T5-XXL-based paraphraser that enables controlled text paraphrasing with adjustable lexical and order diversity, specializing in long-form content.
Brief-details: Gemma 2 9B instruction-tuned model optimized for MLX framework, quantized to 4-bit precision for efficient deployment while maintaining performance.
Brief-details: Specialized 7B parameter biological AI model for CRISPR-Cas systems, featuring StripedHyena architecture with 8k context length and efficient sequence processing capabilities.
Brief-details: Optimized 4-bit quantized version of Meta's Llama 3.1 (405B) model, featuring 2.4x faster training and 58% reduced memory usage via Unsloth's optimization techniques.
Brief-details: Polish RoBERTa-based embedding model generating 768D vectors, trained via multilingual knowledge distillation on 60M Polish-English pairs. Optimized for semantic tasks.
BRIEF-DETAILS: NVIDIA's content safety model built on Llama2-7B, trained on 11k safety-annotated prompts to detect 13 unsafe content categories with high accuracy (94.1% AUPRC)
Brief Details: F222 is an AI model created by acheong08, hosted on HuggingFace. Limited public information available suggests specialized functionality.
Brief Details: Dolphin-2.2.1-mistral-7b is an uncensored, Apache-2.0 licensed 7B parameter LLM based on Mistral AI, featuring enhanced conversation abilities and empathy, trained on 4xA100s for 48 hours.
BRIEF-DETAILS: Chinese portrait-focused AI model based on FLUX.1-dev, optimized for Asian features with enhanced skin quality rendering and studio-quality outputs
Brief Details: Pre-processed ControlNet difference modules for efficient transfer control in Stable Diffusion workflows. Compatible with sd-webui-controlnet extension.
Brief-details: UIGEN-T1-Qwen-7b is a 7B parameter UI generation model fine-tuned on Qwen2.5-Coder-7B-Instruct, specializing in HTML/CSS generation through chain-of-thought reasoning.
Brief Details: A collection of image annotation tools and algorithms by lllyasviel, hosted on HuggingFace, designed for computer vision tasks and image processing workflows.
BRIEF DETAILS: Advanced bird classification model using MobileNetV3Large architecture, trained on 500 species with 90k+ images. Achieves 92%+ accuracy on test/validation sets.
BRIEF DETAILS: EfficientNet-B0 is Google's lightweight CNN model optimized for mobile use, featuring compound scaling across dimensions for improved accuracy at 224x224 resolution.
