Brief Details: Improved VAE autoencoder for Stable Diffusion, fine-tuned on LAION-Aesthetics/LAION-Humans with EMA weights for better image reconstruction
Brief Details: LLM2Vec transforms decoder-only LLMs into powerful text encoders using bidirectional attention and masked token prediction, built on Llama 3.
Brief Details: A text-to-image diffusion model combining InsaneRealistic2 and ChainGirl-Surreality, specialized in surreal and realistic compositions with 21.5K+ downloads
Brief-details: TestMixtral is a 6.61M parameter transformer-based text generation model using F32 tensor type, optimized for inference endpoints with Safetensors support.
Brief-details: A powerful text-to-image diffusion model optimized for ultra-realistic image generation with advanced cyberpunk and portrait capabilities, offering high-detail output with customizable parameters.
Brief Details: KR-ELECTRA-generator is a Korean language ELECTRA model with strong performance on informal texts, featuring 12 layers, 768-dim embeddings, and 30K vocab size.
Brief Details: NuExtract-1.5: A 3.82B parameter multilingual model fine-tuned from Phi-3.5-mini-instruct, specialized in structured information extraction across 6 languages
Brief Details: Phi-3.5 uncensored model with 3.82B params, available in multiple GGUF quantizations from 7.64GB to 1.32GB, optimized for different hardware configs
Brief Details: DeiT base model with 86.6M params for ImageNet classification. Vision transformer architecture with 16x16 patch size and 224x224 input resolution.
Brief-details: BERT-based dense retrieval model with 110M parameters, trained on BEIR corpus and MS MARCO for robust zero-shot performance.
Brief-details: NemoMix-Unleashed-12B-GGUF is a comprehensive GGUF quantized model offering multiple compression variants from 24GB to 4.4GB, optimized for various hardware configurations.
BRIEF DETAILS: A 7B parameter code generation model fine-tuned from CodeQwen1.5-7B using ORPO technique, achieving 86.6% pass@1 on HumanEval benchmarks.
Brief Details: A BERT model with 137M params optimized for 2048 token sequences, featuring rotary position embeddings and SwiGLU activations for enhanced performance.
Brief Details: A large wav2vec2 model trained on 2.9K hours of French speech data, optimized for speech recognition with 317M parameters and Apache 2.0 license.
Brief-details: Massive merged model combining 234 LoRAs with Flux Schnell base for enhanced text-to-image generation. Optimized for Euler/DPM2/Heun samplers at ~4 steps.
Brief-details: ChemGPT-4.7M is a GPT-Neo-based transformer model for molecular generation, trained on PubChem10M dataset to generate SMILES strings for chemical compounds.
Brief-details: 12B parameter GGUF model optimized for inference with multiple quantization options (Q2_K to Q8_0), offering flexible balance between size and quality
Brief-details: Qwen2.5-7B-Instruct-AWQ is a 4-bit quantized instruction-tuned language model with 7.61B parameters, offering enhanced capabilities in coding, math, and multilingual support with 128K context length.
Brief-details: XLM-RoBERTa-based ONNX model for detecting 21 languages with high accuracy. Optimized for efficient language identification using transformer architecture.
Brief-details: 2B parameter instruction-tuned LLM by IBM, optimized for multilingual tasks with strong performance in reasoning and code. Uses GQA and RoPE architecture.
Brief-details: A powerful text-to-image model optimized for ultra-realistic and anime-style image generation, featuring enhanced prompt processing and safety controls with over 22K downloads
