fast-reid
fast-reid copied to clipboard
JDAI-CV
SOTA Re-identification Methods and Toolbox
问题描述：项目:FastRetri,backbone:resnet50,训练测试batchsize:32,第一次验证,显存暴涨至7G多，后面训练验证能稳定到4G以内。由于最大显存为7G多，相当于我需要提供的显存要在8G以上。为什么第一次验证时，需要那么大的显存缓冲区空间？可以减小么？
Dear Author, There is a bug in ./fastreid/modeling/backbones/regnet/regnet.py - line 562. It is currently written as: '4000x': 'fastreid/modeling/backbones/regnet/regnety/RegNetX-4.0GF_dds_8gpu.yaml', But '**regnety**' should be rectified as '**regnetx**', like this: '4000x': 'fastreid/modeling/backbones/regnet/regnetx/RegNetX-4.0GF_dds_8gpu.yaml', I...
## Instructions To Reproduce the 🐛 Bug: FastAttr 的每次前传推理每次的结果都不一样，是不是在AttrEvaluator中没有使用model.eval()
## ❓ How to do something using fastreid 请问，行人属性识别工程，提供tensorrt部署细节吗？
Hi everyone I have a question. We are planning to learn fast-reid considering its use in FastMOT. I would like to use the widerface dataset for learning. Is this possible?...
使用的设备是4060 LAPTOP 当我尝试可视化训练的AGW模型时候出现下述报错，显存足够但无法调用（可以完成训练，以及训练中的验证） `torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 1.55 GiB (GPU 0; 8.00 GiB total capacity; 335.96 MiB already allocated; 5.59 GiB free; 1.32 GiB reserved in...
there was a change in Python 3.10 and the Mapping class has been moved to the collections.abc module. So, if you're using Python 3.10 or later, you should import the...
A few issues occur: 1. the docker cannot run on CUDA11, aka, all the Amphere arch GPUs, like 3070, 3080, ... 2. the documented docker run has issues: `/bin/sh` exe...
