PaddlePaddle
CPU预测结果正常，GPU预测图片没框，同时训练时map也一直为0
问题确认 Search before asking
Bug组件 Bug Component
Training, Inference, Installation
Bug描述 Describe the Bug
正常配置完环境，没有代码改动，执行的‘快速体验’中的例程（后面自己的权重也一样没结果）：
用PP-YOLO算法在COCO数据集上预训练模型预测一张图片
python tools/infer.py -c configs/ppyolo/ppyolo_r50vd_dcn_1x_coco.yml -o use_gpu=true weights=https://paddledet.bj.bcebos.com/models/ppyolo_r50vd_dcn_1x_coco.pdparams --infer_img=demo/000000014439.jpg
复现环境 Environment
jetpack5.0.2(AGX Orin) paddlepaddle2.4.1 --直接从文档下载的预编译库 paddledection:2.5/develop
Bug描述确认 Bug description confirmation
- [X] 我确认已经提供了Bug复现步骤、代码改动说明、以及环境信息，确认问题是可以复现的。I confirm that the bug replication steps, code change instructions, and environment information have been provided, and the problem can be reproduced.
是否愿意提交PR？ Are you willing to submit a PR?
- [X] 我愿意提交PR！I'd like to help by submitting a PR!
可以在https://github.com/PaddlePaddle/PaddleDetection/blob/release/2.5/ppdet/modeling/architectures/yolo.py#L119 打印下输出的box值，在cpu和gpu下是否不同
可以在https://github.com/PaddlePaddle/PaddleDetection/blob/release/2.5/ppdet/modeling/architectures/yolo.py#L119 打印下输出的box值，在cpu和gpu下是否不同
您好，我测试了下，不使用gpu的时候预测结果是正常的，有bbox输出个数和位置信息，use_gpu=true时输出和打印结果如下，预测信息都是空的，而且Place怎么是CPU，救命！
(pd25) orin@ubuntu:~/Downloads/PaddleDetection-release-2.5$ python tools/infer.py -c configs/ppyolo/ppyolo_r50vd_dcn_1x_coco.yml -o use_gpu=true weights=https://paddledet.bj.bcebos.com/models/ppyolo_r50vd_dcn_1x_coco.pdparams --infer_img=demo/000000014439.jpg
W0207 21:35:50.670056 14369 gpu_resources.cc:61] Please NOTE: device: 0, GPU Compute Capability: 8.7, Driver API Version: 11.4, Runtime API Version: 11.4
W0207 21:35:50.679394 14369 gpu_resources.cc:91] device: 0, cuDNN Version: 8.4.
[02/07 21:35:54] ppdet.utils.checkpoint INFO: Finish loading model weights: /home/orin/.cache/paddle/weights/ppyolo_r50vd_dcn_1x_coco.pdparams
[02/07 21:35:54] ppdet.data.source.category WARNING: anno_file 'dataset/coco/annotations/instances_val2017.json' is None or not set or not exist, please recheck TrainDataset/EvalDataset/TestDataset.anno_path, otherwise the default categories will be used by metric_type.
[02/07 21:35:54] ppdet.data.source.category WARNING: metric_type: COCO, load default categories of COCO.
**output**: {'bbox': Tensor(shape=[0, 6], dtype=float32, place=Place(cpu), stop_gradient=True,
[]), 'bbox_num': Tensor(shape=[1], dtype=int32, place=Place(cpu), stop_gradient=True,
[0])}
**bbox**: Tensor(shape=[0, 6], dtype=float32, place=Place(cpu), stop_gradient=True,
[])
**bbox_num：** Tensor(shape=[1], dtype=int32, place=Place(cpu), stop_gradient=True,
[0])
@jerrywgz
这里paddle测试gpu没问题，上面预测过程gpu预测就没结果了，而且output打印出来place显示是cpu。。。
>>> paddle.utils.run_check()
Running verify PaddlePaddle program ...
W0208 09:52:06.066910 17815 gpu_resources.cc:61] Please NOTE: device: 0, GPU Compute Capability: 8.7, Driver API Version: 11.4, Runtime API Version: 11.4
W0208 09:52:06.073951 17815 gpu_resources.cc:91] device: 0, cuDNN Version: 8.4.
PaddlePaddle works well on 1 GPU.
PaddlePaddle works well on 1 GPUs.
PaddlePaddle is installed successfully! Let's start deep learning with PaddlePaddle now.
这里打印的cpu place是正常的，你这里使用的paddle版本是从哪里获取的呢，我这边测试2.4版本的预测结果是正常的，https://www.paddlepaddle.org.cn/install/quick?docurl=/documentation/docs/zh/install/pip/linux-pip.html
