Brief-details: Ziya-BLIP2-14B-Visual-v1 is a bilingual (Chinese/English) visual-language model that combines BLIP2 architecture with LLaMA for sophisticated visual question-answering and dialogue capabilities.
BRIEF DETAILS: ChatLaw-13B is a specialized Chinese-English legal LLM based on Ziya-LLaMA-13B-v1, trained on extensive legal datasets for consultation and reasoning tasks.
Brief-details: LLaMA 7B model optimized for Transformers 4.29, featuring 32 attention heads and 32 layers. Trained on diverse datasets for research purposes.
Brief-details: MPT-7B-Storywriter-GGML is a 6.7B parameter story-focused LLM optimized for CPU inference, with 65k+ token context and GGML quantization support
Brief-details: A specialized image generation model fine-tuned on PVC figure images, offering high-quality anime-style outputs with support for Danbooru tags and multiple rendering styles.
Brief-details: LayoutLM-based document QA model specialized for invoice processing with 128M params. Supports non-consecutive token extraction and multi-modal analysis.
Brief-details: A 4-bit quantized Guanaco model optimized for consumer hardware, requiring <6GB memory, featuring multilingual capabilities and GPTQ optimization techniques
BRIEF DETAILS: A fine-tuned Stable Diffusion model specialized in generating fashion and clothing images, based on OpenJourney, with support for detailed fashion-specific prompts.
Brief-details: Core ML optimized version of Stable Diffusion v1.5 for Apple Silicon, offering efficient text-to-image generation with original and split_einsum variants.
Brief Details: MGP-STR base is a 148M parameter Vision Transformer-based model for scene text recognition, using multi-granularity prediction and A^3 modules.
Brief-details: A powerful 11B parameter Polish language model with strong performance in both Polish and English tasks, featuring sophisticated instruction tuning and achieving state-of-the-art results on multiple benchmarks.
Brief Details: A specialized text-to-image model for generating kawaii-style animal icons, built on Stable Diffusion with creative commons licensing and optimized for cute, minimal designs.
Brief Details: OCRonos - An 8B parameter LLM specialized in correcting OCR errors across 5 languages, trained on diverse digitized texts including cultural and financial documents.
Brief-details: MaskFormer model with Swin backbone for semantic segmentation, trained on ADE20k dataset. Treats all segmentation tasks as instance segmentation through mask and label prediction.
Brief Details: 7B-parameter Polish language model fine-tuned for instruction following, offering strong performance in RAG tasks with 39.28 average score on benchmarks
Brief-details: Multimodal vision-language model combining LLaMA 3 with vision capabilities via SigLIP projection, offering 8.48B parameters for image understanding and Q&A tasks.
Brief-details: Orca-2-7B-GGUF is Microsoft's reasoning-focused 7B parameter LLM, quantized by TheBloke for efficient CPU/GPU inference with multiple compression options
Brief-details: GPT-4o is a Hugging Face-compatible tokenizer implementation adapted from OpenAI's tiktoken, offering seamless integration with Transformers and Transformers.js libraries.
Brief-details: Uncensored 70B parameter LLaMA2 chat model, GPTQ-quantized for efficient deployment. Features multiple quantization options and straightforward responses without filters.
Brief Details: A high-performance text-to-image model combining PixArt-α architecture with LCM for ultra-fast image generation (4 steps), capable of 1024px outputs.
Brief Details: DeepSeek Math 7B-RL is a specialized mathematical reasoning model with 6.91B parameters, optimized through reinforcement learning for step-by-step problem solving.
