Brief Details: A DeBERTa-v3 based text classification model with 184M parameters, fine-tuned using linear learning rate scheduling and Adam optimizer. MIT licensed.
Brief Details: Korean sentence embedding model using difference-based contrastive learning, achieving 77.17% average performance on semantic tasks with 111M parameters
Brief Details: A 24.9B parameter GGUF-optimized conversational AI model based on Llama 3.1 architecture, with significant community adoption (33.4K downloads)
Brief-details: HeBERT is a Hebrew BERT-based model specialized in sentiment analysis, achieving 97% accuracy for polarity classification with comprehensive training on massive Hebrew text corpora.
Brief Details: A powerful 12.7B parameter vision-language model capable of processing multiple images with text, built on Transformers architecture with BF16 precision.
Brief-details: Compact-biobert is a distilled version of BioBERT optimized for biomedical text, featuring 65M parameters across 6 transformer layers with MIT license
Brief-details: Qwen2.5-Coder-7B is a specialized code-focused LLM with 7.62B parameters, supporting 128K context length and optimized for code generation, reasoning, and fixing.
Brief-details: A testing-focused RoBERTa model implementation with minimal architecture (10 layers, 20 heads) designed for CI/CD pipelines and unit testing purposes.
Brief-details: Finnish BERT language model with 125M parameters, trained on 3B tokens of Finnish text. Outperforms multilingual BERT for Finnish NLP tasks including NER and POS tagging.
Brief Details: A photorealistic text-to-image diffusion model optimized for landscapes and general imagery, featuring v1.6 capabilities with creative ML open rail license.
Brief-details: German Named Entity Recognition model using Flair embeddings & LSTM-CRF architecture. Achieves 87.94% F1-score on CoNLL-03, identifies PER/LOC/ORG/MISC entities.
Brief-details: A 3.2B parameter GGUF-formatted Llama model optimized for instruction-following, supporting 8 languages with Q8_0 quantization for efficient deployment
Brief-details: A RoBERTa-large based cross-encoder model specialized in semantic textual similarity, trained on STS benchmark dataset with Apache 2.0 license.
Brief-details: A 12-layer transformer-based regression model with 38.9M parameters, pre-trained on synthetic datasets for tabular prediction tasks
Brief-details: EXAONE-3.0-7.8B-Instruct is a 7.82B parameter LLaMA-based conversational AI model optimized for text generation and inference, featuring F32 tensor architecture.
Brief-details: A merged text-to-image model combining RealLife v2 and Timeless, optimized for photorealistic and artistic outputs with fast generation capabilities
Brief-details: RegNetY-008 is a lightweight image classification model with 6.26M parameters, trained on ImageNet-1k, optimized for efficient processing with 0.81 GMACs.
Brief Details: A compact 135M parameter language model optimized for efficiency. Trained on 2T tokens, supports text generation with BF16 precision and shows strong performance on various NLP tasks.
BRIEF-DETAILS: Qwen2.5-3B-Instruct is a compact yet powerful 3.09B parameter instruction-tuned language model offering 32K context length, multilingual support, and enhanced capabilities in coding and mathematics.
Brief Details: A specialized code embedding model with 161M parameters supporting 30+ programming languages. Features 8192 sequence length and ALiBi architecture.
Brief Details: Qwen2-7B is a powerful 7.6B parameter language model excelling in multilingual tasks, coding, and reasoning with state-of-the-art performance.
