BRIEF-DETAILS: Neural machine translation model for Chinese to Vietnamese translation, achieving BLEU score of 20.0 and chrF of 0.385. Built by Helsinki-NLP using transformer-align architecture.
Brief Details: Japanese to Hungarian neural MT model by Helsinki-NLP, trained on OPUS data. BLEU: 12.2, chrF: 0.364. Uses transformer-align architecture.
Brief-details: A French-to-Spanish neural machine translation model based on the transformer-align architecture, achieving BLEU scores of 31.6-53.2 across various test sets
Brief-details: English to Swedish neural machine translation model by Helsinki-NLP, achieving 60.1 BLEU score on Tatoeba test set, based on transformer architecture
BRIEF DETAILS: English to Albanian neural machine translation model by Helsinki-NLP, achieving 46.5 BLEU score on Tatoeba test set using transformer architecture
Brief-details: Helsinki-NLP's English-to-Romanian neural MT model based on transformer architecture, achieving BLEU 30.8 on newsdev2016 benchmark
Brief-details: Multilingual translation model supporting English to 300+ languages, trained on OPUS data. Features SentencePiece tokenization and achieves strong BLEU scores for major languages.
Brief-details: English to Indo-Iranian languages translation model supporting 30+ target languages, based on transformer architecture with BLEU 13.7 and chrF 0.392.
Brief-details: English to Haitian Creole neural machine translation model based on OPUS dataset, achieving BLEU 38.3 on JW300 and 45.2 on Tatoeba test sets.
Brief-details: A MSCOCO-finetuned version of CoCa-ViT-L-14, built on LAION-2B dataset, combining vision-language capabilities for enhanced image understanding and description generation
Brief-details: A specialized FLAN-T5 Large model fine-tuned on QuaRel dataset. As indicated by name, this appears to be a development version not intended for production use.
Brief-details: MLP-Mixer vision model (59.9M params) pretrained on ImageNet-21k and fine-tuned on ImageNet-1k, achieving efficient image classification at 224x224 resolution
Brief Details: Multilingual CLIP model combining ViT-B/32 vision encoder with XLM-RoBERTa text encoder, trained on LAION-5B dataset for zero-shot image classification and retrieval.
Brief-details: A compact experimental model by katuni4ka hosted on HuggingFace, likely a smaller variant of the ExaOne architecture focused on efficient processing.
Brief-details: A 1B parameter Llama model optimized for instruction-following tasks, quantized to 4-bit (Q4) format using GGUF for efficient deployment via llama.cpp framework.
Brief-details: A specialized ELECTRA-based model likely focused on language understanding tasks, developed by crystina-z and hosted on HuggingFace.
Brief Details: BART-based tech keyword extraction model, fine-tuned for identifying technical terms, tools, and company names from text. Loss: 0.8795.
Brief Details: An 8B parameter LLaMA-based model optimized for Traditional Chinese & Taiwan-specific content, featuring enhanced instruction following & educational capabilities.
BRIEF DETAILS: Japanese Stable Diffusion XL - A specialized image generation model by Stability AI optimized for Japanese-style content creation, building on SDXL architecture.
Brief Details: A comprehensive collection of LoRA models for various art styles and clothing, organized specifically for Chinese users with translated names and trigger tags.
Brief Details: First chat model using state-space architecture instead of transformers. Based on Mamba-2.8B, uses Zephyr prompt format for efficient dialogue.
