BRIEF DETAILS: WD 1.5 Beta 3 is a text-to-image diffusion model with 5 variants (Base, Radiance, Ink, Mofu, Illusion), specializing in anime-style image generation.
BRIEF DETAILS: MOSS-003-base is a 16B parameter multilingual LLM pre-trained on 700B tokens, supporting English and Chinese with plugin capabilities for enhanced interactions.
Brief-details: A collection of so-vits-svc-4.0 voice conversion models trained on various character voices, featuring 9 different voice models with detailed training data and loss information. MIT licensed.
Brief-details: Long-T5 model fine-tuned for book summarization, handling 16K token inputs. 248M params, ROUGE-1: 36.41. Ideal for long-form text summarization.
Brief-details: WizardCoder-33B-V1.1 is a state-of-the-art 33B parameter code LLM achieving 79.9% pass@1 on HumanEval, outperforming ChatGPT 3.5 and Gemini Pro.
Brief-details: Epic Diffusion is a versatile Stable Diffusion 1.x-based model optimized for high-quality image generation across multiple styles, with NSFW capabilities and extensive artist influences
Brief Details: Microsoft's CodeReviewer - A specialized T5-based model for automating code review tasks, trained on code changes and review data
Brief-details: A specialized Stable Diffusion model for photorealistic Japanese portraits, merged from 5000+ images with excellent facial detail and natural lighting capabilities.
Brief-details: Collection of Hololive VTuber LoRAs and textual inversions for Stable Diffusion, featuring multi-outfit models for EN, JP and ID talents.
Brief-details: An 8-bit quantized version of GPT-J-6B, optimized for single-GPU training and inference, enabling use on consumer hardware while maintaining performance.
Brief-details: A compact 22.7M parameter sentence transformer model optimized for semantic similarity tasks, offering comparable performance to larger models.
Brief-details: A high-resolution NSFW image generation model based on FLUX architecture, specializing in photorealistic outputs with detailed textures and natural lighting effects.
BRIEF DETAILS: Microsoft's Phi-2: A 2.7B parameter transformer model optimized for research, excelling in QA, chat, and code generation. Trained on 250B tokens.
BRIEF-DETAILS: Massive 120B parameter GGUF-format language model derived from Llama-2, combining Xwin and Euryale models. Offers multiple quantization options for efficient deployment.
Brief-details: MolmoE-1B is a multimodal Mixture-of-Experts LLM with 1.5B active/7.2B total parameters, achieving near GPT-4V performance in vision-language tasks.
Brief-details: LongChat-13B-16k is a research-focused chatbot based on LLaMA-13B, featuring extended context handling up to 16k tokens using rotary embedding techniques.
Brief Details: A Flux Dev 1-based model that creates unique hybrid images combining photographic and illustrative elements, optimized for artistic compositions.
Brief-details: GroundingDINO is an open-set object detection model that combines DINO architecture with grounded pre-training, enabling zero-shot detection capabilities.
Brief-details: A compact 33.4M parameter text embedding model from DAMO Academy, optimized for English text similarity tasks with strong MTEB benchmark performance.
Brief Details: A 7B parameter Python-specialized LLM from Meta's Code Llama family, optimized for code completion and understanding with PyTorch compatibility.
Brief-details: Fine-tuned Llama2 13B model with 8K context window, optimized for long-form conversations using Orca dataset. Features RoPE scaling and multi-dataset training.
