BRIEF DETAILS: Compact 22.7M parameter sentence embedding model optimized for semantic search, trained on 215M question-answer pairs with 384-dimensional output vectors.
BRIEF DETAILS: BERT-based sentence embedding model with 109M parameters. Maps sentences to 768D vectors. Deprecated due to low quality - newer alternatives recommended.
Brief-details: A specialized latent diffusion model for image inpainting, based on Stable Diffusion v1.5. Enables high-quality image editing and completion with text prompts.
Brief-details: CLIP ViT-B/16 model trained on LAION-2B dataset, achieving 70.2% ImageNet accuracy. Specialized in zero-shot image classification and retrieval.
Brief Details: Popular text-to-image model trained on LAION-2B dataset. Features 595k training steps at 512x512 resolution with improved classifier-free guidance sampling.
Brief Details: astroBERT is a 110M-parameter BERT-based language model specialized for astrophysics research, featuring masked language modeling and named entity recognition capabilities.
Brief-details: FastText language identification model by Facebook, capable of detecting 217 languages with efficient word representation learning and quick CPU-based processing
Brief Details: A RoBERTa-based sentiment analysis model trained on 124M tweets (2018-2021), offering 3-class classification with high accuracy and Twitter-specific preprocessing.
Brief-details: Advanced NLP model from Microsoft with 304M parameters, achieving SOTA on NLU tasks. Features disentangled attention and enhanced mask decoder.
Brief Details: A 1.01B parameter GPT-2 variant optimized for efficient text generation, featuring F32 tensor type and custom optimizations for improved performance.
Brief-details: Microsoft's DeBERTa base model featuring disentangled attention mechanism, achieving SOTA results on NLU tasks with 88.8% MNLI-m accuracy.
BRIEF DETAILS: BART-large-CNN: 406M parameter transformer-based summarization model fine-tuned on CNN Daily Mail dataset. Achieves ROUGE-1: 42.95, ROUGE-2: 20.81.
BRIEF DETAILS: A compact BERT variant (L=4, H=512) optimized for efficiency, part of smaller BERT family models, MIT licensed with 5.5M+ downloads
BRIEF DETAILS: Chronos-T5-Small: A 46M parameter time series forecasting model based on T5 architecture. Enables probabilistic forecasts through token-based sequence modeling.
BRIEF DETAILS: BERT multilingual model supporting 104 languages with 179M parameters. Pre-trained on Wikipedia data using masked language modeling. Apache 2.0 licensed.
Brief-details: A BERT-like transformer model optimized for long documents up to 4,096 tokens, featuring sliding window attention and global attention mechanisms.
Brief Details: OPT-125M is Meta AI's smallest open-source GPT-style language model with 125M parameters, designed for text generation and research accessibility.
Brief-details: Efficient sentence embedding model with 22.7M params, maps text to 384D vectors. Popular choice with 5.9M+ downloads. Apache 2.0 licensed.
Brief Details: ALBERT Base v2 - Lightweight BERT variant with 11.8M params, sharing layer weights. Trained on BookCorpus & Wikipedia for MLM tasks.
Brief-details: T5-small is a compact 60M parameter text-to-text transformer model that can handle multiple NLP tasks like translation, summarization and QA through a unified text-based approach.
Brief Details: Efficient sentence embedding model with 33.4M params, trained on 1B+ sentence pairs. Maps text to 384D vectors for similarity tasks.
