Brief-details: A 70B parameter uncensored LLaMA2 variant fine-tuned on Wizard-Vicuna dataset, offering direct responses without safety filters. Uses QLoRA for efficient training.
Brief-details: A quantized version of Falcon-180B-Chat optimized for efficient deployment, offering 4-bit precision and multiple GPTQ configurations for varying hardware requirements.
Brief-details: Large-scale 70B parameter chat model optimized for code and text generation, built on LLaMA architecture with research focus and CC BY-NC-4.0 license.
Brief Details: Atom-7B is a Chinese-English bilingual LLM with 7B parameters, built on Llama2 architecture with enhanced Chinese language capabilities and 4K context length.
Brief Details: A 300M parameter multimodal GGUF quantized model based on Obsidian-3B, optimized for efficiency with Q6 quantization and MIT license.
Brief-details: A fine-tuned Llama2 7B chat model specialized for code generation, trained on 122k code instructions with qLoRA optimization and commercial use potential
Brief-details: 4-bit quantized version of OpenAssistant's Llama 30B model, offering GPTQ & GGML variants with optimized performance for both GPU and CPU usage.
Brief-details: A 4-bit quantized version of WizardLM-Uncensored-Falcon-7B featuring 1.54B parameters, optimized for GPU inference with AutoGPTQ, offering uncensored text generation capabilities.
BRIEF DETAILS: A 16B parameter plugin-augmented conversational LLM from Fudan University supporting Chinese and English, with integrated tools like search, calculator, and text-to-image capabilities.
Brief-details: LaMini-Flan-T5-248M is a distilled instruction-tuned language model with 248M parameters, based on Flan-T5-base, trained on 2.58M instruction samples.
Brief Details: SakuraMix is a Japanese-focused text-to-image diffusion model that balances background and character quality with built-in VAE support. Latest v4 version improves hand rendering and reduces artifacts.
Brief-details: A ControlNet-based model that enhances Stable Diffusion 2.1 with aesthetic image generation capabilities using Canny edge detection, supporting 640x640 resolution.
Brief-details: A 4-bit quantized version of LLaMA-65B, optimized for efficient inference using GPTQ compression. Requires significant RAM (120GB+) for conversion and A100-GPU compatible.
Brief Details: FRED-T5-1.7B is a Russian language T5-based model with 1.7B parameters, trained on 300GB corpus using 7 denoisers, optimized for text generation.
Brief Details: SamDoesArt-V3 is a specialized text-to-image model focusing on artistic portrait generation, featuring a unique artistic style trigger token and flexible prompt positioning.
Brief-details: AltDiffusion-m9 is a multilingual text-to-image diffusion model supporting 9 languages, built on Stable Diffusion with enhanced cross-lingual capabilities and improved image generation quality.
Brief Details: A 784M parameter Chinese T5 model fine-tuned on 100+ datasets for multiple NLP tasks. Ranked 3rd on ZeroClue benchmark, specialized in text classification, NLI, and generation tasks.
BRIEF DETAILS: AMD-OLMo is a 1.2B parameter language model trained on AMD MI250 GPUs, offering strong performance across multiple benchmarks with SFT and DPO variants available.
Brief-details: A 70B parameter reward model built on Llama-3.1 architecture, specialized in evaluating LLM response quality with state-of-the-art performance across multiple benchmarks
Brief-details: A specialized Stable Diffusion model fine-tuned on emoji datasets, enabling high-quality emoji generation from text descriptions. Features custom scheduler and torch integration.
BEN is an advanced image segmentation model specializing in background removal, achieving state-of-the-art performance with 94M parameters. Features both mask and foreground image generation with impressive accuracy metrics (DICE: 0.8743, IOU: 0.8301).
