ONNX Model Zoo

AXERA-TECH/Qwen2.5-Omni-3B

By AXERA-TECH · ↓ 0 · ♥ 0 · Updated 2025-08-05

What's in this repo

This page lists the files in AXERA-TECH/Qwen2.5-Omni-3B, including thinker_to_talker_proj_decode.onnx, thinker_to_talker_proj_prefill_512.onnx.

  • 2 ONNX models (.onnx files)
  • 109 files, including 66 .axmodel, 22 .json, 7 .py, 4 .bin, 2 .txt, 2 .npy
  • Last updated 2025-08-05

ONNX files

File Size Download
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/thinker_to_talker_proj_decode.onnx — Download
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/thinker_to_talker_proj_prefill_512.onnx — Download

How to use

These ONNX files can be loaded with ONNX Runtime. To inspect a model graph, upload the file to the Netron viewer. Download links point to the original files on Hugging Face.

README


license: mit language:

  • en
  • zh base_model:
  • Qwen/Qwen2.5-Omni-3B pipeline_tag: image-text-to-text library_name: transformers tags:
  • Qwen2.5-Omni
  • Qwen2.5-Omni-3B
  • Int8
  • Omni

Qwen2.5-Omni-3B

This version of Qwen2.5-Omni-3B has been converted to run on the Axera NPU using w8a16 quantization.

Compatible with Pulsar2 version: 3.4

Convert tools links:

For those who are interested in model conversion, you can try to export axmodel through the original repo : https://huggingface.co/Qwen/Qwen2.5-Omni-3B

Pulsar2 Link, How to Convert LLM from Huggingface to axmodel

AXera NPU HOST LLM Runtime

Support Platform

Audio Process 待补充

Image Process 待补充

Video Process 待补充

The DDR capacity refers to the CMM memory that needs to be consumed. Ensure that the CMM memory allocation on the development board is greater than this value.

How to use

Download all files from this repository to the device

If you using AX650 Board

(base) lihongjie@ai-dev1:~/axera-tech/Qwen2.5-Omni-3B$ tree -L 2
.
├── python
│   ├── modeling_axinfer.py
│   ├── preprocess.py
│   ├── run_axinfer.py
│   ├── utils_axinfer.py
│   ├── utils_lm.py
│   ├── utils.py
│   └── utils_vl.py
├── Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512
│   ├── config.json
│   ├── qwen2_5_omni_talker_p128_l0_together.axmodel
......
│   ├── qwen2_5_omni_talker_post.axmodel
│   ├── special_tokens_map.json
│   ├── spk_dict.pt
│   ├── thinker_to_talker_proj_decode.axmodel
│   ├── thinker_to_talker_proj_prefill_512.axmodel
│   ├── token2wav_bigvgan.axmodel
│   ├── token2wav_dit.axmodel
│   ├── tokenizer_config.json
│   ├── tokenizer.json
│   └── vocab.json
├── Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512
│   ├── config.json
......
│   ├── Qwen2.5-Omni-3B_vision.axmodel
│   ├── qwen2_5_omni_text_p128_l0_together.axmodel
......
│   ├── qwen2_5_omni_text_post.axmodel
├── README.md
└── videos


Demo

运行命令
cd python
python3 run_axinfer.py
output

text

It's a Nord Electro 6 keyboard. It's a really popular one. It has a lot of features like different sounds and effects. Have you played with it much?

audio

Model info

image-text-to-text transformers
transformers onnx safetensors Qwen2.5-Omni Qwen2.5-Omni-3B Int8 Omni image-text-to-text en zh base_model:Qwen/Qwen2.5-Omni-3B base_model:quantized:Qwen/Qwen2.5-Omni-3B
↓ 0 ♥ 0

Created: 2025-08-01

Updated: 2025-08-05

View on Hugging Face

Files (109)

.gitattributes
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/added_tokens.json
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/chat_template.json
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/config.json
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/configuration.json
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/generation_config.json
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/merges.txt
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/model.embed_tokens.weight.bfloat16.bin
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/model.embed_tokens.weight.float32.bin
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/model.embed_tokens.weight.npy
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/model.safetensors.index.json
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/preprocessor_config.json
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l0_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l10_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l11_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l12_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l13_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l14_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l15_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l16_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l17_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l18_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l19_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l1_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l20_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l21_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l22_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l23_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l2_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l3_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l4_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l5_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l6_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l7_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l8_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_p128_l9_together.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/qwen2_5_omni_talker_post.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/special_tokens_map.json
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/spk_dict.pt
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/thinker_to_talker_proj_decode.onnx ONNX
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/thinker_to_talker_proj_prefill_512.onnx ONNX
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/token2wav_bigvgan.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/token2wav_dit.axmodel
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/tokenizer.json
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/tokenizer_config.json
Qwen2.5-Omni-3B-AX650-talker-chunk_prefill_512/vocab.json
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/Qwen2.5-Omni-3B_vision.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/added_tokens.json
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/audio_tower.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/chat_template.json
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/config.json
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/configuration.json
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/generation_config.json
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/merges.txt
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/model.embed_tokens.weight.bfloat16.bin
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/model.embed_tokens.weight.float32.bin
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/model.embed_tokens.weight.npy
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/model.safetensors.index.json
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/preprocessor_config.json
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l0_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l10_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l11_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l12_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l13_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l14_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l15_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l16_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l17_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l18_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l19_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l1_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l20_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l21_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l22_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l23_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l24_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l25_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l26_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l27_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l28_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l29_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l2_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l30_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l31_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l32_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l33_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l34_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l35_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l3_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l4_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l5_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l6_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l7_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l8_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_p128_l9_together.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/qwen2_5_omni_text_post.axmodel
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/special_tokens_map.json
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/tokenizer.json
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/tokenizer_config.json
Qwen2.5-Omni-3B-AX650-thinker-chunk_prefill_512/vocab.json
README.md
python/modeling_axinfer.py
python/preprocess.py
python/run_axinfer.py
python/utils.py
python/utils_axinfer.py
python/utils_lm.py
python/utils_vl.py
videos/1.mp4

Related models