Brief-details: A BART-based ChatGPT prompt generator fine-tuned on prompts dataset, achieving 2.48 train loss and 2.73 validation loss after 4 epochs. Apache 2.0 licensed.
Brief Details: Portuguese language instruction-tuned LLaMA model using LoRA, trained on translated Alpaca dataset. Efficient fine-tuning using A100 GPU.
Brief-details: GODEL is a Transformer-based encoder-decoder model for goal-directed dialogs, pre-trained on 551M multi-turn dialogs, optimized for grounded response generation.
Brief Details: GLM-4-Voice-9B: A 9.54B parameter end-to-end voice model for Chinese/English speech understanding and generation with real-time capabilities
Brief Details: A melancholic variant of Llama-3-8B-Instruct using orthogonalization to create an intentionally unenthusiastic conversational style. 8B parameters.
Brief Details: Emu3-Chat: An 8.49B parameter multimodal model using next-token prediction for image/text/video tasks. Outperforms SDXL and LLaVA-1.6 in generation and perception.
Brief Details: A 7B parameter AWQ-quantized Mistral model with 128k context window, optimized for long-form text generation and efficient inference at 4-bit precision.
Brief Details: Llama3-Chinese-8B-Instruct is an 8B parameter Chinese language model based on Llama3, optimized for instruction-following and dialogue tasks with FP16 precision.
Brief-details: A 2.8B parameter Mamba architecture model fine-tuned on OpenHermes dataset, optimized for instruction-following and conversational AI tasks using state space modeling approach.
BRIEF DETAILS: InternLM-XComposer2-4KHD is a powerful vision-language model capable of processing 4K resolution images, built on InternLM2 architecture with advanced visual understanding capabilities.
Brief Details: T2I-Adapter for SDXL specialized in sketch-to-image generation. 77M parameters, Apache 2.0 licensed, built on SDXL base model.
Brief Details: RWKV's v5-Eagle-7B-HF is a 7B parameter transformer-based LLM optimized for both CPU and GPU inference, supporting multilingual text generation with HuggingFace integration.
Brief-details: A 4-bit quantized Chinese-English LLaMA 2 model optimized for bilingual conversation, built on 10M instruction-tuning samples with commercial usage rights.
Brief Details: UltraLM-13b is a fine-tuned LLaMA-based chat model trained on UltraChat dataset, optimized for multi-turn conversations with 13B parameters.
Brief Details: A 7B parameter medical LLM fine-tuned from LLaMA, specialized in medical Q&A with extensive training on healthcare datasets including ChatDoctor and WikiDoc.
Brief Details: A fine-tuned Whisper Large model optimized for Chinese (Mandarin) speech recognition, achieving 9.55% CER on Common Voice 11 test set.
Brief Details: A powerful 32B parameter code-specialized LLM with 128K context length, optimized for code generation, reasoning, and fixing. Built on Qwen2.5 architecture.
Brief-details: A Stable Diffusion 2.x embedding model trained on 120 photos for generating photorealistic images with enhanced color representation and photography-like qualities.
Brief-details: StoryMaker is a specialized text-to-image model focusing on maintaining character consistency across multiple scenes, ideal for visual storytelling and sequential image generation.
Brief Details: Optimized speech recognition model supporting 100+ languages, built on Whisper large-v3, converted to CTranslate2 format for faster inference.
Brief-details: Large-scale 70B parameter LLaMA3-based model fine-tuned for conversational AI. Features uncensored responses, coding capabilities, and function calling support.
