OpenELM is a series of efficient language models introduced by Apple, utilizing a hierarchical scaling strategy to optimize parameter allocation, offering pretrained and instruction-tuned models ranging from 270M to 3B parameters.

Natural Language ProcessingTransformers

PHOENIX is a domain-adaptive model based on CLIP/ViT, designed to enhance patent image retrieval capabilities, particularly suitable for retrieving semantically or hierarchically related images rather than exact matches.

Computer VisionSafetensors

ReasonFlux-F1-7B is a hierarchical large language model reasoning model based on thinking template expansion, fine-tuned through template-enhanced reasoning trajectories, demonstrating outstanding performance in multiple reasoning tasks.

Natural Language ProcessingTransformers

ReasonFlux-F1-32B is a hierarchical large language model based on template-enhanced reasoning trajectory fine-tuning, excelling in reasoning tasks.

Natural Language ProcessingTransformers

HiMTok is a hierarchical mask token learning framework fine-tuned on the InternVL2_5-8B large multimodal model, focusing on image segmentation tasks.

MultimodalSafetensors

Hierarchical Distillation (SwD) is a novel framework that accelerates diffusion models (DMs) by progressively increasing spatial resolution during the generation process.

Computer VisionDiffusers

Hierarchical Distillation (SwD) is a novel framework that accelerates diffusion models (DMs) by progressively increasing spatial resolution during the generation process.

Computer VisionDiffusers

ModularStarEncoder-300M is an encoder model fine-tuned on the SynthCoNL dataset based on the ModularStarEncoder-1B pre-trained model. It is specifically designed for code-to-code and text-to-code retrieval tasks. This model uses hierarchical self-distillation technology, allowing users to choose different layer versions according to their computing power.

Natural Language ProcessingTransformers

SwarmFormer is an efficient sequence modeling architecture that optimizes long-sequence processing capabilities through hierarchical attention mechanisms and dynamic clustering techniques.

Natural Language ProcessingTransformers

LLaVA-UHD v2 is an advanced multimodal large language model built around a hierarchical window transformer, capable of capturing different visual granularities through a high-resolution feature pyramid.

MultimodalTransformers

HiCo is a hierarchical controllable diffusion model specifically designed for layout-to-image generation tasks.

Computer VisionDiffusersEnglish

OWSM-CTC is an encoder-only speech foundation model based on hierarchical multitask self-conditioned CTC, supporting multilingual speech recognition, speech translation, and language identification.

Audio ProcessingEspnetOther

The Arabic Triplet Matryoshka V2 Model is an advanced Arabic embedding model based on the sentence-transformers framework, specifically optimized for semantic understanding of Arabic texts. This model is fine-tuned from aubmindlab/bert-base-arabertv02 and uses the matryoshka loss training technique to create hierarchical embeddings at multiple resolutions, capable of mapping sentences and paragraphs to a 768-dimensional dense vector space.

Natural Language ProcessingSentence-transformersArabic

An improved model based on Llama2 7B, supporting hierarchical skip and self-speculative decoding to enhance inference efficiency

Natural Language ProcessingTransformersEnglish

Hiera is an efficient hierarchical vision Transformer model that excels in image and video tasks with fast runtime

Computer VisionTransformersEnglish

Hiera is a hierarchical vision Transformer model that is fast, powerful, and concise, surpassing existing technologies in image and video tasks while being faster.

Computer VisionTransformersEnglish

Hiera is a hierarchical vision Transformer model that is fast, powerful, and concise, surpassing state-of-the-art performance in a wide range of image and video tasks while significantly improving runtime speed.

Computer VisionTransformersEnglish

Hiera is a hierarchical vision Transformer model that is fast, powerful, and concise, excelling in image and video tasks.

Computer VisionTransformersEnglish

Hiera is a hierarchical vision Transformer model that is fast, powerful, and concise. It surpasses state-of-the-art performance in a wide range of image and video tasks while significantly improving runtime speed.

Computer VisionTransformersEnglish

Hiera is a hierarchical vision Transformer model that combines speed, powerful functionality, and minimalist design, significantly surpassing existing technical standards in image and video tasks with outstanding computational efficiency.

Computer VisionTransformersEnglish