Brief Details: BLIP-2 vision-language model using Flan T5-XL (3.94B params). Excellent for image captioning, VQA, and visual conversations. MIT licensed.
Brief Details: A comprehensive index repository for NovelAI-related LoRA models, focusing on curating high-quality Stable Diffusion fine-tunes with character-specific implementations.
Brief-details: H2O.ai's 7B parameter LLM based on Falcon, fine-tuned on OpenAssistant dataset. Optimized for conversational AI with Apache 2.0 license.
Brief Details: A specialized Stable Diffusion model fine-tuned on Dahi Puri food images, enabling high-quality food visualization with creative prompts and custom implementations.
BRIEF-DETAILS: A specialized text-to-image Stable Diffusion model focused on ligne claire anime style, featuring clean lines and flat colors with CreativeML OpenRAIL-M license.
Brief Details: VintageHelper is a Stable Diffusion 2.0/2.1 enhancement model that adds analog-style effects, bokeh, and vintage aesthetics to images
Brief-details: Textual inversion embedding for Stable Diffusion 2.x focused on toxic environments and dystopian themes with green-tinted aesthetics and chemical/industrial elements.
Brief-details: YOLOS-small: A 30.7M parameter Vision Transformer for object detection, achieving 36.1 AP on COCO. Built by hustvl with Apache 2.0 license.
Brief-details: BART-based summarization model fine-tuned on SAMSum dialogue dataset, achieving 43.21 ROUGE-1 score. Optimized for conversation summaries.
Brief Details: GAN-based image-to-image model that transforms photos into anime backgrounds in Makoto Shinkai's distinctive style, built with PyTorch under MIT license.
Brief Details: First Chinese instruction-tuned model based on Gemma-2-27B-IT, optimized for bilingual capabilities with 27.2B parameters and 8K context length. Uses ORPO fine-tuning.
Brief-details: SDXL-EcomID is an advanced text-to-image model combining PuLID and InstantID technologies for enhanced ID-based image generation with strong facial consistency and keypoint control.
Brief-details: Russian text summarization model based on mBART architecture, specialized for news articles with 867M parameters. Achieves ROUGE-1 F1 of 32.4 on Gazeta dataset.
Brief-details: Flux.1-lumiere-alpha is an enhanced text-to-image model focused on improved realism while maintaining prompt coherency, built on FLUX.1-Dev.
Brief-details: WALDO30 is a YOLOv8-based object detection model specializing in overhead imagery analysis, capable of identifying 12 distinct object classes from drone to satellite altitudes.
Brief Details: A 1.8B parameter chat-optimized LLM based on Mistral architecture, fine-tuned with SFT+DPO, supporting 8K context length and achieving 48.44% on LLM benchmarks.
Brief Details: Yuan2.0-M32 is a 40B parameter MoE model with 32 experts (2 active), using Attention Router for efficient selection. Achieves state-of-the-art performance with only 3.7B active parameters.
Brief-details: A specialized 34B parameter Yi-based merge focused on 40K+ context and instruct-enhanced storytelling, combining multiple Vicuna-format models for optimal narrative generation.
Brief Details: A 7.37B parameter bilingual (Japanese/English) chat model by Rakuten, built on Mistral architecture with extended vocabulary and state-of-the-art Japanese language understanding capabilities.
Brief-details: A repository for requesting and managing AI model quantizations, maintained by mradermacher. Specializes in various GGUF quantization types and imatrix generation.
Brief-details: A 34B parameter LLM using novel DPO-Positive training, achieving 77.29% average score across benchmarks. Built on Bagel-34B with focus on math and reasoning.
