Brief Details: A powerful 72B parameter bilingual (English/Chinese) chat model based on Qwen architecture, adapted with LLaMA compatibility and GPL-3.0 license.
Brief-details: Advanced German language Mistral-based LLM (7.24B params) optimized for German text generation and understanding, featuring enhanced conversational abilities
Brief-details: A 7B parameter reward model built on LLaMA architecture, specialized in evaluating text quality in Chinese and English, trained on 40K+ preference samples.
Brief-details: GPT4-X-Alpasta-30b-4bit is a merged model combining GPT4-Alpaca and Open Assistant, optimized for both GPU/GPTQ and CPU/GGML usage with multiple quantization options.
Brief-details: A triple-merged model combining Alpaca with Chain-of-Thought and Storytelling capabilities, optimized for reasoning and narrative generation. Based on LLaMA 30B architecture.
Brief-details: An anime-focused AI model combining Vtuber elements with AbyssOrangeMix2, optimized for artistic generation with special attention to NSFW capabilities and LORA compatibility.
Brief Details: LLaMA-Mesh is an 8.03B parameter model that unifies 3D mesh generation with language models, capable of generating and understanding both text and 3D meshes from textual prompts.
Brief Details: A 4-bit quantized version of Alpaca-30B using GPTQ method, optimized for efficient inference with support for instruction-tuning and text generation.
Brief Details: Athene-V2-Agent: A 72.7B parameter agent model excelling in function calling and tool use, surpassing GPT-4o with enhanced reasoning capabilities.
BRIEF DETAILS: T5-based text-to-SQL model (223M params) that converts natural language questions to SQL queries using table schema information
Brief-details: Alpaca-LoRA-7B is a fine-tuned variant of LLaMA-7B using Stanford Alpaca dataset, optimized for instruction-following tasks with PEFT/LoRA techniques
Brief-details: Zero-DCE model for enhancing low-light images without reference data. Uses deep curve estimation for dynamic range adjustment. TF-Keras implementation, Apache 2.0 licensed.
Brief-details: Pixtral-12B-Base-2409 is a powerful multimodal AI model with 12B parameters, capable of processing both images and text with a 128k sequence length and support for 9 languages.
Brief-details: Extended context Llama-3 8B model (4194k tokens) optimized for instruction-following with NTK-aware interpolation and progressive training
Brief-details: Tencent's 7.57B parameter image captioning model supporting Chinese/English, built on LLaVA architecture. Specialized in detailed image descriptions with high text-image consistency.
Brief-details: 70B parameter Mistral-based quantized model optimized for high performance and French language capability, with varied quantization levels for different VRAM requirements
BRIEF DETAILS: A versatile text-to-image diffusion model featuring enhanced CLIP architecture, supporting wide CFG ranges (3-50) and cross-style generation capabilities with improved prompt interpretation.
Brief-details: DolphinCoder StarCoder2 15B - Advanced coding-focused AI assistant trained on 8 datasets, built for software engineering tasks with uncensored capabilities.
Brief Details: ChemLLM-7B-Chat is a specialized 7.74B parameter LLM focused on chemistry and molecular science, built on InternLM-2, supporting both English and Chinese interactions.
Brief-details: Emo-AffectNet is a PyTorch-based facial emotion recognition model capable of real-time video/image analysis, with MIT license and research backing.
Brief-details: MegaDolphin-120b is a 120B parameter language model built on Dolphin-2.2-70b, featuring enhanced conversational abilities and empathy, with strong performance across multiple benchmarks.
