BRIEF-DETAILS: BERT-based model for detecting political bias in text, classifying content as left, center, or right-leaning. Built by bucketresearch for automated bias detection.
BRIEF-DETAILS: Specialized AI model for generating Inugami Korone (Hololive VTuber) images, trained on Waifu Diffusion 1.3 with 22 training and 3300 regularization images.
Brief-details: 4-bit quantized version of Qwen2.5-3B-Instruct optimized by Unsloth for faster training and lower memory usage, featuring dynamic quantization for improved accuracy
Brief-details: Qwen2.5-VL-7B-Instruct-AWQ is an advanced vision-language model offering enhanced visual understanding, video analysis, and structured output capabilities with AWQ quantization.
BRIEF DETAILS: Fast and lightweight Russian BERT model for sentence embeddings. 312-dim embeddings, 2048 context size, optimized for speed with good accuracy.
Brief-details: GraphCodeBERT is a Transformer-based model for code understanding that combines sequence data with data-flow graphs, featuring 12 layers and trained on 2.3M functions across 6 programming languages.
Brief-details: ConvNeXt tiny model (28.6M params) pretrained on ImageNet-12k and fine-tuned on ImageNet-1k, achieving 84.2% top-1 accuracy at 224px
BRIEF-DETAILS: BERT large cased model with whole word masking, 336M parameters, fine-tuned on SQuAD dataset. Optimized for question-answering tasks.
Brief-details: A powerful 16B parameter code-centric LLM built on Deepseek-v2-Lite-Base, achieving SOTA performance in code generation and mathematical tasks
Brief-details: DistilBERT model fine-tuned for Named Entity Recognition (NER) using CoNLL-2003 dataset, case-insensitive version optimized for English text analysis.
Brief Details: ChemBERTa-77M-MLM is a 77M parameter BERT-style model pre-trained on SMILES molecular representations for chemical property prediction and analysis.
Brief Details: Meta-Llama 3.1 8B Instruct model optimized to 4-bit quantization for MLX framework, offering efficient inference with reduced memory footprint
BRIEF-DETAILS: A 15M parameter MoE model with 4 experts based on TinyLlama, specialized for storytelling. Features Shakespeare LoRA adapter for creative text generation.
Brief-details: BiomedParse - Microsoft's foundation model for biomedical image analysis across 9 modalities, enabling joint segmentation, detection & recognition with advanced transformer architecture.
Brief Details: A minimal BART model implementation designed specifically for TRL library testing purposes, focusing on core transformer architecture validation.
Brief-details: InternLM3's 8B parameter instruction model optimized for reasoning tasks, trained on 4T tokens with 75% less cost than peers. Excels in math and knowledge tasks.
Brief-details: BERT-based model specialized for mental health analysis, trained on Reddit data using 4 Tesla V100 GPUs for 624k iterations over 8 days
Brief-details: MetricX-24 XXL hybrid model for translation evaluation, supporting both reference-based and reference-free assessment with state-of-the-art performance
Brief-details: French NER model based on CamemBERT, specialized in entity & date recognition. F1 score ~83% on mixed chat/email data. Built on wikiner-fr dataset.
Brief Details: Compact 1.1B parameter LLM based on TinyLLama, fine-tuned with WizardVicuna dataset. Available in multiple GGUF quantizations from 482MB to 1.17GB.
BRIEF-DETAILS: Llama-3.2-3B-Instruct-8bit is an MLX-optimized 8-bit quantized version of Meta's Llama 3.2B instruction model, offering efficient deployment
