BRIEF-DETAILS: AI-powered speech enhancement tool with dual-module system for denoising and quality improvement, trained on 44.1kHz speech data.
BRIEF DETAILS: A 3B parameter code completion model by Stability AI, trained on diverse programming languages with 16k context window and strong completion capabilities (pass@1: 20.18%)
Brief-details: MS-Image2Video is a state-of-the-art image-to-video generation model with 3.7B parameters, capable of producing high-quality 720P videos with enhanced temporal consistency and texture quality
Brief Details: MuseV is a cutting-edge text-to-video generation model enabling infinite-length virtual human videos with high fidelity and visual conditioning, supporting multiple generation modes.
BRIEF-DETAILS: Experimental ControlNet variant offering lightweight control for SDXL with multiple preprocessing options including blur, canny, and depth maps. Apache 2.0 licensed.
BRIEF-DETAILS: Uncensored 7B parameter LLaMA2 chat model optimized for CPU/GPU inference via GGML format, offering various quantization options from 2-8 bits with different performance tradeoffs.
Brief-details: OpenLLaMA 7B v2 is an Apache 2.0 licensed reproduction of Meta's LLaMA, trained on 1T tokens using Falcon, StarCoder & RedPajama datasets
Brief Details: A 15B parameter code generation model optimized for GGML format, achieving 57.3% pass@1 on HumanEval, with multiple quantization options (4-8 bit).
Brief Details: A powerful 73.2B parameter multimodal LLM built on Qwen2-72B, capable of advanced vision-language tasks with uncensored reasoning and strong benchmark performance.
BRIEF-DETAILS: Enhanced 7B parameter Alpaca variant optimized for llama.cpp with GGML quantization, featuring improved text generation and conversation capabilities
Brief Details: Beta version of Waifu Diffusion 1.5 - An anime-style text-to-image model with improved generation capabilities and custom embeddings for prompt optimization.
Brief Details: TableGPT2-7B is a 7.62B parameter model specialized for tabular data analysis, built on Qwen2.5-7B with support for Chinese/English and extensive table comprehension capabilities.
Brief-details: WD1.5-beta based text-to-image model specializing in high-quality anime-style character generation with enhanced detail and composition control.
Brief-details: A powerful 12.1B parameter language model by Stability AI, trained on 2T tokens across 7 languages, optimized for text generation with Flash Attention 2 support.
Brief Details: DALL-E 3 XL LoRA v2 is a text-to-image diffusion model built on Fluently-XL-v2, offering DALL-E 3-like capabilities with custom LoRA weights
Brief-details: UNA-Cybertron-7B-v2: Advanced 7B parameter Mistral-based LLM achieving top leaderboard performance (69.67%). Excels in mathematics, reasoning, and logic tasks.
Brief-details: ChatMusician (6.74B params) - Specialized LLM for music understanding and generation, built on LLaMA2, using ABC notation for musical tasks and composition.
Brief-details: A 9.24B parameter Gemma-based model fine-tuned using Self-Play Preference Optimization, achieving 53.27% win rate on AlpacaEval, optimized for instruction-following tasks.
BRIEF DETAILS: Text editing AI model (783M params) fine-tuned from FLAN-T5-large for grammar correction, coherence, simplification, and style adaptation.
Brief-details: Manticore-13B: Advanced LLaMA-based model fine-tuned on diverse datasets including ShareGPT and WizardLM, optimized for text generation and instruction following.
Brief-details: AudioGen-medium is a 1.5B parameter text-to-audio generation model by Facebook, capable of synthesizing general audio from text descriptions at 16kHz using EnCodec tokenization.
