Brief-details: Multilingual phone recognition model using Wav2Vec2 architecture, supporting 6 languages with impressive accuracy (avg 9.2% PER). Optimized for pathological speech analysis.
Brief Details: Quantized version of Microsoft's Phi-3 medium model with 128k context window, offering multiple compression options from 3.7GB to 55.8GB with varying quality-size tradeoffs.
Brief Details: LLaVA-OneVision is an 8.03B parameter multimodal LLM that combines Qwen2 with vision capabilities for single-image, multi-image, and video tasks.
BRIEF DETAILS: Text-to-image model focusing on photorealistic generation with 43.7K+ downloads. Supports various styles from abstract to realistic, built on StableDiffusion.
Brief Details: ControlNet inpainting model for Stable Diffusion v1.5, enabling precise image editing with masked regions and conditional control over generation.
Brief-details: Large-scale BERT model (335M params) for dense retrieval, pretrained on BEIR corpus and fine-tuned on MS MARCO, focusing on distribution shift handling.
Brief Details: A 10B parameter Gemma-based creative writing model with enhanced prose generation capabilities through Brainstorm 5x technology. Optimized for fiction, storytelling, and vivid descriptions.
Brief-details: CodeBERTa-small-v1 is a 6-layer RoBERTa-like model trained on CodeSearchNet, optimized for code understanding across 6 programming languages with 84M parameters.
Brief Details: Erlangshen-Roberta-330M-Sentiment is a fine-tuned Chinese RoBERTa model with 326M parameters, specialized in sentiment analysis across 8 datasets with 227,347 samples.
Brief Details: OpenAI GPT-1: Pioneer 120M parameter transformer model for language understanding. First of its kind from OpenAI with MIT license and strong zero-shot capabilities.
Brief-details: MedCPT-Article-Encoder is a 109M-parameter transformer model for generating biomedical text embeddings, trained on 255M PubMed query-article pairs
Brief Details: A state-of-the-art depth estimation model with 345M parameters, combining relative and metric depth estimation using DPT framework. MIT licensed.
Brief Details: Japanese DeBERTa V2 base model (122M params) pre-trained on Wikipedia, CC-100, and OSCAR. Features character-level tokenization and whole word masking for advanced NLP tasks.
Brief Details: InstructPLM protein design model with 6.57B parameters, combining ProGen2 and ProteinMPNN architectures for accurate protein sequence generation based on backbone structures
BRIEF-DETAILS: Lightweight text-to-speech model with 878M parameters, capable of generating natural speech with controllable features like gender, speed, and pitch. Apache 2.0 licensed.
Brief Details: DistilBERT model fine-tuned on MNLI dataset for zero-shot classification. 67M parameters, English-language focused, achieving 82% accuracy on MNLI tasks.
Brief Details: Pegasus-large: Google's powerful abstractive summarization transformer model with mixed & stochastic training on C4 and HugeNews datasets, achieving state-of-the-art results.
Brief-details: Neural machine translation model for Finnish to English conversion with strong BLEU scores (53.4 on Tatoeba test), built by Helsinki-NLP using transformer architecture
Brief-details: A Thai-to-English translation model by Helsinki-NLP using transformer-align architecture, achieving 48.1 BLEU score on Tatoeba test set.
Brief Details: Universal Sentence Encoder for Russian (USER-bge-m3) - 359M parameter model for Russian text embeddings with 1024-dimensional vectors, based on BGE-M3.
Brief Details: PhotoMaker is an advanced text-to-image AI model that creates customized photos from face inputs and text prompts, featuring ID embedding and SDXL compatibility.
