SABER-LLM

Unofficial mirror of SABER-LLM by the original authors Zhao, Tian and Xie. Model weights and inference configuration files are unchanged. Maintained by mcflurryshuoz; not an official author release.

Original model: https://huggingface.co/zhaoxiaoxian/SABER-LLM
Original code: https://github.com/zxzhao0/SABER-LLM
Paper: https://arxiv.org/abs/2601.18321
Upstream revision: 0b0b1e75900882dcda0f095f97e1ca3207a2a562.
Architecture: Qwen2_5OmniForConditionalGeneration, based on Qwen2.5-Omni-7B.

All four Safetensors shards, the model index, tokenizer, processors and speech dictionary are included. Training state and local download logs are excluded. Follow the upstream inference instructions and original project for ms-swift usage.

The upstream model repository requires acceptance of its access conditions. Refer to the original repository for those conditions. The upstream project's Apache-2.0 license is preserved in LICENSE.

Downloads last month: -

Safetensors

Model size

9B params

Tensor type

BF16

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for mcflurryshuoz/SABER-LLM

Base model

Qwen/Qwen2.5-Omni-7B

Finetuned

(83)

this model

Paper for mcflurryshuoz/SABER-LLM

Integrating Fine-Grained Audio-Visual Evidence for Robust Multimodal Emotion Reasoning

Paper • 2601.18321 • Published Jan 26