BRIEF-DETAILS: Quantized 7B parameter LLaMA2 model optimized for efficient inference with GPTQ compression, offering multiple quantization options and broad compatibility
Brief Details: Vision Transformer model with 86.6M params, trained on ImageNet-21k and fine-tuned on ImageNet-1k. Optimized for 224x224 images with advanced augmentation.
Brief-details: A RoBERTa-based sentiment analysis model trained on 40k tweets from SemEval 2017, specializing in English tweet classification with POS/NEG/NEU labels
Brief Details: Qwen1.5-0.5B-Chat is a 620M parameter chat model optimized for multilingual support with 32K context length, featuring BF16 precision and no trust_remote_code requirement.
Brief-details: Multilingual text embedding model supporting 94 languages, trained on 1B+ text pairs with strong performance on retrieval and similarity tasks
Brief Details: VideoMAE large - A 343M parameter video transformer model for masked autoencoding, pre-trained on Kinetics-400 for self-supervised learning
Brief Details: Microsoft's TrOCR small model for handwritten text recognition, fine-tuned on IAM dataset. Uses transformer-based architecture with 500K+ downloads.
BRIEF-DETAILS: A Dutch to English translation model by Helsinki-NLP using transformer architecture, achieving 60.9 BLEU score on Tatoeba dataset with Apache 2.0 license.
Brief-details: A 7.62B parameter language model based on Qwen2 architecture, optimized for text generation with F32 tensor precision and high download count (500K+)
Brief Details: MeaningBERT - A 109M parameter BERT-based model for assessing meaning preservation between sentences with high human judgment correlation
Brief Details: A Vision Transformer model leveraging SigLIP (Sigmoid Loss) for zero-shot image classification, trained on WebLI dataset with state-of-the-art performance.
Brief Details: A Korean speech recognition model built on wav2vec2-XLSR architecture, achieving 4.74% WER and 1.78% CER on Zeroth Korean dataset. 317M parameters, Apache 2.0 licensed.
Brief Details: Microsoft's WavLM speech model pre-trained on 94k hours of audio data. Optimized for speech recognition and speaker identification tasks using self-supervised learning.
Brief Details: A Korean NER model based on ELECTRA, with 14.1M params. Achieves 82.32% precision and 84.49% recall on Korean entity recognition tasks.
Brief-details: OpenAssistant's 12B parameter language model fine-tuned for conversation, based on Pythia architecture with Apache 2.0 license and extensive English language capabilities.
Brief-details: SDXL 1.0-refiner is a specialized refinement model for Stable Diffusion XL, designed to enhance image generation quality through a two-stage pipeline process.
Brief Details: Powerful multilingual text classifier supporting 51 languages with high accuracy (98.89% avg). Based on XLM-RoBERTa, trained on MASSIVE dataset.
Brief Details: Lightweight Russian sentiment analysis model (29.2M params) for 3-class text classification. MIT-licensed with strong F1 (0.75) and ROC-AUC (0.90) scores.
Brief Details: Document Image Transformer (DiT) base model - A BERT-like transformer for document image processing, pre-trained on 42M documents using self-supervised learning.
Brief-details: Microsoft's LayoutLMv2 - Advanced multimodal document AI model combining text, layout & image analysis. 523K+ downloads, SOTA results on key benchmarks.
Brief-details: ProtBert is a BERT-based protein language model trained on 217M protein sequences from Uniref100, specializing in protein feature extraction and masked language modeling.
