Brief Details: A specialized MPT variant focused on code generation, featuring random initialization and remote code execution capabilities. Developed by echarlaix on HuggingFace.
Brief Details: 8B parameter Llama model converted to GGUF format for efficient CPU/GPU inference via llama.cpp, featuring DeepSeek's distillation techniques
BRIEF-DETAILS: Fine-tuned 8B parameter LLaMA-3 model trained on diverse datasets for SFT, serving as RLHF research foundation
BRIEF DETAILS: A Vietnamese-focused text-to-speech model supporting 18 languages, fine-tuned from XTTS-v2.0.3. Enables voice cloning with just 6 seconds of audio input.
Brief-details: GoodHands-beta2 is a specialized AI model developed by jlsim, available on HuggingFace, focused on hand-related tasks and computer vision applications.
Brief-details: A voice AI model created by Byzern that specializes in synthesizing speech in the style of Jokowi, with community support features and donation options available.
Brief-details: Textual Inversion embedding model for Stable Diffusion 2.0 (768x768) offering papercut-style image generation, trained on 106 manually captioned images with 8 vectors
BRIEF-DETAILS: Domain-adapted BERT model for German job ads, trained on 4M Swiss job listings. Specialized for job market analysis and recruitment tasks.
Brief-details: Domain-adapted BERT model for German job ads, trained on 4M Swiss job listings. Specialized for job market analysis and recruitment NLP tasks.
BRIEF DETAILS: DQN-based reinforcement learning model for Space Invaders, trained using Stable-Baselines3. Features 4-frame stacking and CNN policy for Atari gameplay optimization.
Brief-details: SAC (Soft Actor-Critic) reinforcement learning model trained on Humanoid-v3 environment using stable-baselines3, optimized for bipedal locomotion tasks.
BRIEF DETAILS: GPT-2-based tweet generator fine-tuned on Grindr & GAY TIMES tweets (5,318 curated tweets). Created by huggingtweets for social media content generation.
Brief-details: Chinese speech model pre-trained on 10k hours of WenetSpeech data, using HuBERT architecture for speech representation learning. Requires fine-tuning for ASR tasks.
Brief-details: BCMS-BERTIC model fine-tuned for ternary sentiment classification (Negative/Neutral/Positive) for Bosnian, Croatian, Montenegrin & Serbian, achieving 0.79 F1 score
Brief-details: Multilingual CLIP model extending OpenAI's text encoders to multiple languages, achieving 91.6% R@10 for English and strong performance across 10+ languages
Brief Details: Chinese biomedical language model specialized in healthcare, featuring 12-layer transformer architecture with 768 hidden dimensions and 12 attention heads. Pre-trained on medical dialogues and records.
BRIEF-DETAILS: PatentSBERTa_V2 is a specialized SBERT-based model for patent analysis, mapping text to 768-dimensional vectors for semantic search and classification.
Brief-details: T5-based model for generating natural language questions from table data, trained on WikiSQL dataset. Specializes in table-based QA generation.
Brief-details: Multilingual text summarization model fine-tuned on Persian (Farsi) content, achieving 45.12 ROUGE-1 score with strong summarization capabilities
BRIEF DETAILS: DeBERTa-v3 model fine-tuned for financial sentiment analysis, achieving 89.13% accuracy on financial news classification with strong precision and recall metrics.
Brief-details: ConvNeXt V2 large model with 198M params, trained with FCMAE and fine-tuned on ImageNet-22k/1k. Achieves 87.26% top-1 accuracy.
