Views
No views yet
ECHO_block8 student.| Property | Value |
|---|---|
| Stage | RAD (teacher) |
| Block Length | 8 |
| Decoding | Multi-step block diffusion |
| Architecture | EchoForConditionalGeneration (based on Qwen2.5-VL) |
| Hidden Size | 3584 |
| Languages | English, Chinese |
| License | Midea NC |
1git clone https://github.com/midea-ai/ECHO.git
2cd ECHO
3pip install transformers==4.55.41# Multi-step inference with ECHO_Base_block8
2python inference/generate_vl_block.py \
3 --model_dir Midea-AIRC/ECHO_Base_block8 \
4 --image_path /path/to/chest_xray.jpg \
5 --prompt_text "Review this chest X-ray and write a report. Use this format: Findings: {}, Impression: {}." \
6 --remasking_strategy "low_confidence_dynamic" \
7 --block_length 8 \
8 --denoising_steps 81python inference/generate_vl_block.py \
2 --model_dir Midea-AIRC/ECHO_Base_block8 \
3 --image_path /path/to/chest_xray.jpg \
4 --prompt_text "这是一组胸部X光图像，请生成一份医学报告，包括所见和结论。以以下格式返回报告：所见：{} 结论：{}。" \
5 --remasking_strategy "low_confidence_dynamic" \
6 --block_length 8 \
7 --denoising_steps 8| Model | Stage | Description | Link |
|---|---|---|---|
ECHO_Base_block4 | RAD | Multi-step block diffusion (block length 4), teacher for distillation | ECHO_Base_block4 |
ECHO_Base_block8 | RAD | Multi-step block diffusion (block length 8), teacher for distillation | ECHO_Base_block8 |
ECHO_block4 | DCD | Single-step distilled student (block length 4) | ECHO_block4 |
ECHO_block8 | DCD | Single-step distilled student (block length 8) | ECHO_block8 |
1@misc{chen2026echoefficientchestxray,
2 title={ECHO: Efficient Chest X-ray Report Generation with One-step Block Diffusion},
3 author={Lifeng Chen and Tianqi You and Hao Liu and Zhimin Bao and Jile Jiao and Xiao Han and Zhicai Ou and Tao Sun and Xiaofeng Mou and Xiaojie Jin and Yi Xu},
4 year={2026},
5 eprint={2604.09450},
6 archivePrefix={arXiv},
7 primaryClass={cs.LG},
8 url={https://arxiv.org/abs/2604.09450},
9}