COCO数据集测试速度异常
- Dominant language
- Python
- Stars
- 306
- Forks
- 96
- PR merge metrics
- No merged PRs in 30d
Description
## 背景
**COCO数据集测试速度异常**
## 任务描述
按照首页readme安装megengine和Models，从 [hub.py](https://github.com/MegEngine/Models/blob/11013f6b3b252df71098b27714e39d9802d2520d/official/vision/detection/configs/atss_res50_coco_3x_800size.py#L18) 下载 预训练权重，直接在coco数据集上进行测试。代码未做任何改动，采用8卡测试，最终结果mAP为42.6（正常），但是耗时将近6m30s（不正常），正常8卡时间应该在1min左右
python：3.6
megengine版本：v1.4.0
## 目标
希望多卡测试速度正常，请确认问题并帮忙修复
Contributor guide
No contributing guide indexed for this repository
Research direction
Reproduce the COCO evaluation with official/vision/detection/configs/atss_res50_coco_3x_800size.py and the stated MegEngine v1.4.0, Python 3.6, and eight-card setup. Compare the nearly 6m30s runtime with the expected roughly one minute while confirming mAP remains 42.6; done means multi-card testing reaches the expected speed.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- distributed-systems, machine-learning, performance
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 38/100