# Zero-shot Inference

Vision encoder based on Mistral Small model, supports image-text generation tasks, compatible with llama.cpp framework

MedGemma-4B-IT is a medical-focused image-to-text generation model developed by Google.

VL-Rethinker-7B-8bit is a multimodal model based on Qwen2.5-VL-7B-Instruct, supporting visual question answering tasks.

This model is a multimodal vision-language model converted from Qwen2.5-VL-7B-Instruct, supporting visual question answering tasks.

Qwen2.5-VL-32B-Instruct is a multimodal vision-language model that supports joint understanding and generation tasks for both images and text.

Qwen2.5-VL-72B-Instruct is a multimodal vision-language model that supports interactive generation tasks involving images and text.

ARPG is an innovative autoregressive image generation framework capable of achieving BERT-style masked modeling through a GPT-like causal architecture.

A citation intent classification model fine-tuned based on Qwen 2.5 14B Instruct, specifically designed for citation intent classification in scientific publications.

Eagle 2 is a high-performance vision-language model family that focuses on transparency in data strategies and training schemes, aiming to drive the open-source community in developing competitive vision-language models.

AiM is an unconditional image generation model based on PyTorch, integrated and pushed to Hugging Face Hub via PytorchModelHubMixin.

MiniCPM-Llama3-V-2_5 is a multimodal visual question answering model based on the Llama3 architecture, supporting both Chinese and English interactions.

A fine-tuned version of Depth Anything V2 for indoor metric depth estimation using the synthetic Hypersim dataset, compatible with the transformers library.

A version fine-tuned for indoor metric depth estimation tasks using the Hypersim synthetic dataset, based on the Depth Anything V2 model

A model fine-tuned from Depth Anything V2 for indoor metric depth estimation tasks, trained on the synthetic dataset Hypersim, compatible with the transformers library.

A fine-tuned version of Depth Anything V2, specifically designed for metric depth estimation in outdoor scenes, trained on the synthetic dataset Virtual KITTI.

Chronos is a family of pre-trained time series forecasting models based on language model architectures, which transform time series into token sequences for training through quantization and scaling.

BLIP-2 is a vision-language model based on OPT-2.7b, which achieves image-to-text generation by freezing the image encoder and large language model while training a query transformer.

BLIP-2 is a vision-language model that combines an image encoder with the large language model Flan T5-xxl for image-to-text tasks.

BLIP-2 is a vision-language model based on Flan T5-xl, pre-trained by freezing the image encoder and large language model, supporting tasks such as image captioning and visual question answering.

FLAN-T5 XL is an instruction-finetuned language model based on the T5 architecture, with significantly improved multilingual and few-shot performance after fine-tuning on 1000+ tasks.

A question-answering model based on the GPT-2 architecture, fine-tuned specifically for the SQuAD2 dataset, capable of answering questions based on given text.

A re-ranking model based on the T5-base architecture, fine-tuned for 100,000 steps on the MS MARCO passage dataset, suitable for document re-ranking tasks in information retrieval.

DeBERTa-v3 model trained on the MultiNLI dataset for natural language inference tasks, excelling in zero-shot classification scenarios.

A zero-shot classification model for Arabic, supporting natural language inference tasks