BRIEF DETAILS: Multimodal Speech LLM combining Llama3.1-8B-Instruct and Whisper-medium for speech/text processing. Achieves 4.45% WER on LibriSpeech with ~50-100 tokens/sec generation.
BRIEF DETAILS: ClimateBERT model fine-tuned for classifying climate-related text into commitment/action vs non-commitment paragraphs. Built on distilroberta-base architecture.
Brief-details: A minimal test model designed for TRL library unit testing, implementing LlavaForConditionalGeneration architecture with compact specifications.
Brief Details: CodeT5+ 220M - Encoder-decoder LLM for code tasks. Supports 9 programming languages. Built by Salesforce with span denoising and CLM pretraining.
Brief-details: XLSR-53-based speech recognition model fine-tuned for English with punctuation support. Built on Wav2Vec2-Large architecture for 16kHz audio processing.
Brief-details: A compact Mixture-of-Experts (MoE) model created by katuni4ka, hosted on HuggingFace. Designed for efficient processing with distributed expert networks.
BRIEF-DETAILS: Optimized 4-bit quantized version of Meta's Llama 3.2 (3B) with Unsloth's Dynamic Quantization, offering 2.4x faster training and 58% less memory usage
BRIEF-DETAILS: A comprehensive LoRA model collection for Genshin Impact characters, featuring 70+ carefully curated and labeled models with Chinese annotations and trigger tags.
Brief Details: A specialized diffusion model trained on Pepe Larraz's artistic style, designed to generate superhero-style comic artwork using the "comicmay artsyle" token.
BRIEF DETAILS: xLSTM-7B: A 7B parameter LSTM-based language model pre-trained on 2.3T tokens, offering competitive performance on various benchmarks with efficient implementation options.
Brief Details: CodeGemma-2b is Google's 2B parameter code generation model requiring Hugging Face login and license acceptance for access
Brief Details: Komodo-7B-Base: A 7B parameter LLM built on Llama-2, specialized for Indonesian and 11 regional languages, with expanded vocabulary and pre-training capabilities.
BRIEF-DETAILS: CodeGPT-small-py is Microsoft's compact Python-focused code generation model, designed for autocompletion and code synthesis tasks
Brief-details: 42-eicar-street by mcpotato is a specialized model hosted on HuggingFace, focusing on security testing and validation functionalities.
Brief Details: NER model fine-tuned on Yoruba language data, achieving 78.22% F1 score. Based on XLM-RoBERTa, specialized for African language NER tasks.
BRIEF-DETAILS: Korean sentiment analysis model for text classification. Evaluates positive/negative sentiment in Korean text. Built using Hugging Face Transformers.
Brief Details: DialoGPT-small variant fine-tuned on Rick Sanchez (Rick and Morty) dialogues, enabling Rick-style conversational responses and interactions.
BRIEF-DETAILS: MatSciBERT - A BERT-based language model specialized for materials science research, trained on papers covering alloys, glasses, cement & concrete from ScienceDirect
Brief-details: Fine-tuned Wav2Vec2-Large-XLSR-53 model for Persian speech recognition, achieving 31.92% WER on Common Voice test set, optimized for 16kHz audio.
Brief-details: A Persian/English text summarization model based on mT5-small architecture, achieving ROUGE-1 F-scores of 43.57% on validation and 43.40% on test sets
Brief Details: MT5-based model specialized in Persian/English article summarization. Strong ROUGE scores (45%+ R1, 27%+ R2) for content summary generation.
