Brief Details: RoBERTa-base model fine-tuned for RTE task, achieving 79.4% accuracy. Optimized using TextAttack with 5 epochs and 2e-05 learning rate.
Brief-details: ALBERT-base-v2 model fine-tuned on IMDB dataset achieving 89.24% accuracy, optimized for sentiment classification using TextAttack framework
Brief-details: Russian NER model for toxic content detection, fine-tuned from rubert-tiny-toxicity. Specialized in identifying toxic entities in Russian text.
Brief-details: A template repository for implementing text classification models with Hugging Face Hub's Inference API, featuring standardized pipeline implementation and deployment workflows.
Brief-details: Cross-lingual NLI model based on XLM-RoBERTa, trained on SNLI/MNLI/ANLI/XNLI datasets for zero/few-shot classification across languages
Brief-details: Fine-tuned ViT model for gender and age detection, achieving 0.2771 loss after 8 epochs. Built on google/vit-base-patch16-224-in21k with Adam optimizer.
BRIEF-DETAILS: 8B parameter LLaMA 3.1 model optimized with INT8 quantization, achieving 100%+ benchmark recovery with 50% reduced memory footprint and 2x compute speedup.
Brief Details: Text adventure AI model based on Mistral-24B, specialized in creating dangerous and often fatal scenarios with cyberpunk and fantasy themes.
Brief-details: GGUF quantized version of DeepSeek-R1-Distill-Llama-8B-Uncensored, offering multiple compression options from 3.3GB to 16.2GB with varied quality-size tradeoffs
Brief Details: T5-small model fine-tuned on BookSum dataset for book summarization tasks. Optimized for generating concise summaries from longer book passages.
BRIEF-DETAILS: Persian BERT model fine-tuned for binary sentiment analysis on DeepSentiPers dataset, achieving 92.42% F1-score for positive/negative classification
BRIEF DETAILS: Tiny Swin Transformer variant (28.3M params) optimized through S3 architecture search, designed for 224x224 images with 4.6 GMACs compute efficiency
Brief Details: MeloTTS-Japanese is a high-quality Japanese text-to-speech model by MyShell.ai, offering real-time CPU inference and MIT-licensed for commercial use
Brief-details: Compact transformer model (22M params) with 5.3x speedup vs BERT-base. Achieves 86.6% avg score on GLUE/SQuAD benchmarks using task transfer distillation.
BRIEF-DETAILS: YOLOS small model fine-tuned on 735 license plate images, achieving 47.9% AP. Vision Transformer-based detector optimized for vehicle and plate recognition.
Brief-details: A fine-tuned Marian model specialized for English to Spanish translation, based on Helsinki-NLP's opus-mt-en-fr, achieving 10.28 BLEU score on KDE4 dataset.
Brief Details: A minimalist test implementation of GPT-2 architecture, developed by fxmarty for experimental purposes and remote code testing
Brief-details: FOMC-RoBERTa is a specialized RoBERTa-based model for classifying Federal Reserve communications as hawkish, dovish, or neutral, trained on FOMC documents.
BRIEF DETAILS: LLaMA3-iterative-DPO model with various GGUF quantizations optimized for different hardware configurations. Features multiple compression levels from 8.54GB to 2.01GB.
Brief-details: StyleTTS2 fine-tuned TTS model trained on 672 speakers across AniSpeech, VCTK, and LibriTTS-R datasets. Optimized for expressive zero-shot speech synthesis with diverse accents
Brief Details: DeBERTaV3-based model excelling in zero-shot text classification and information extraction, outperforming BART-large-mnli while being 3x smaller.
