YOLOX训练报错TypeError: evaluate() got an unexpected keyword argument ‘return_outputs‘

最新推荐文章于 2025-02-06 16:00:45 发布

氯化氢323

最新推荐文章于 2025-02-06 16:00:45 发布

阅读量1.6k

点赞数 1

分类专栏：深度学习踩坑日记文章标签： python 深度学习开发语言

本文链接：https://blog.csdn.net/qq_67954671/article/details/125623875

版权

深度学习踩坑日记专栏收录该内容

4 篇文章

订阅专栏

本文记录了在训练YOLOX模型过程中遇到的评估阶段错误，并提供了详细的错误信息及解决方案。通过更新代码到最新版本，成功解决了评估时出现的TypeError问题。

摘要由CSDN通过智能技术生成

训练YOLOX的时候，每次评估均会报错停止，报错信息如下：

2022-07-02 17:51:57 | INFO | yolox.core.trainer:253 - epoch: 10/300, iter: 10/36, mem: 5854Mb, iter_time: 0.220s, data_time: 0.001s, total_loss: 6.4, i
ou_loss: 2.7, l1_loss: 0.0, conf_loss: 3.1, cls_loss: 0.6, lr: 1.248e-03, size: 608, ETA: 0:53:54
2022-07-02 17:52:00 | INFO | yolox.core.trainer:253 - epoch: 10/300, iter: 20/36, mem: 5854Mb, iter_time: 0.225s, data_time: 0.001s, total_loss: 6.3, i
ou_loss: 2.8, l1_loss: 0.0, conf_loss: 2.9, cls_loss: 0.7, lr: 1.247e-03, size: 480, ETA: 0:53:26
2022-07-02 17:52:02 | INFO | yolox.core.trainer:253 - epoch: 10/300, iter: 30/36, mem: 5854Mb, iter_time: 0.235s, data_time: 0.001s, total_loss: 6.6, i
ou_loss: 2.7, l1_loss: 0.0, conf_loss: 3.3, cls_loss: 0.6, lr: 1.247e-03, size: 672, ETA: 0:53:02
2022-07-02 17:52:04 | INFO | yolox.core.trainer:356 - Save weights to ./YOLOX_outputs\yolox_voc_s
2022-07-02 17:52:04 | INFO | yolox.core.trainer:195 - Training of experiment is done and the best AP is 0.00
2022-07-02 17:52:04 | ERROR | yolox.core.launch:98 - An error has been caught in function 'launch', process 'MainProcess' (14092), thread 'MainThread' (
5720):
Traceback (most recent call last):

File "E:\Study\DeepLearn\DPlearn\YOLO\YOLOX\tools\train.py", line 134, in
launch(
└ <function launch at 0x000001FE9ADB3940>

File "C:\ProgramData\Anaconda3\lib\site-packages\yolox-0.3.0-py3.9.egg\yolox\core\launch.py", line 98, in launch
main_func(*args)
│ └ (╒═══════════════════╤═══════════════════════════════════════════════════════════════════════════════════════════════════════...
└ <function main at 0x000001FEA3077F70>

File "E:\Study\DeepLearn\DPlearn\YOLO\YOLOX\tools\train.py", line 118, in main
trainer.train()
│ └ <function Trainer.train at 0x000001FEA3E20EE0>
└ <yolox.core.trainer.Trainer object at 0x000001FEA3E2BD60>

File "C:\ProgramData\Anaconda3\lib\site-packages\yolox-0.3.0-py3.9.egg\yolox\core\trainer.py", line 76, in train
self.train_in_epoch()
│ └ <function Trainer.train_in_epoch at 0x000001FEA3E295E0>
└ <yolox.core.trainer.Trainer object at 0x000001FEA3E2BD60>

File "C:\ProgramData\Anaconda3\lib\site-packages\yolox-0.3.0-py3.9.egg\yolox\core\trainer.py", line 86, in train_in_epoch
self.after_epoch()
│ └ <function Trainer.after_epoch at 0x000001FEA3E29940>
└ <yolox.core.trainer.Trainer object at 0x000001FEA3E2BD60>

File "C:\ProgramData\Anaconda3\lib\site-packages\yolox-0.3.0-py3.9.egg\yolox\core\trainer.py", line 222, in after_epoch
self.evaluate_and_save_model()
│ └ <function Trainer.evaluate_and_save_model at 0x000001FEA3E29C10>
└ <yolox.core.trainer.Trainer object at 0x000001FEA3E2BD60>

File "C:\ProgramData\Anaconda3\lib\site-packages\yolox-0.3.0-py3.9.egg\yolox\core\trainer.py", line 328, in evaluate_and_save_model
(ap50_95, ap50, summary), predictions = self.exp.eval(
│ │ └ <function Exp.eval at 0x000001FEA3E29550>
│ └ ╒═══════════════════╤═══════════════════════════════════════════════════════════════════════════════════
═════════════════════...
└ <yolox.core.trainer.Trainer object at 0x000001FEA3E2BD60>

File "C:\ProgramData\Anaconda3\lib\site-packages\yolox-0.3.0-py3.9.egg\yolox\exp\yolox_base.py", line 322, in eval
return evaluator.evaluate(model, is_distributed, half, return_outputs=return_outputs)
│ │ │ │ │ └ True
│ │ │ │ └ False
│ │ │ └ False
│ │ └ YOLOX(
│ │ (backbone): YOLOPAFPN(
│ │ (backbone): CSPDarknet(
│ │ (stem): Focus(
│ │ (conv): BaseConv(
│ │ (conv): ...
│ └ <function VOCEvaluator.evaluate at 0x000001FEA3E1B940>
└ <yolox.evaluators.voc_evaluator.VOCEvaluator object at 0x000001FEA7253DF0>

TypeError: evaluate() got an unexpected keyword argument 'return_outputs'

查询无果后，在github提交了Issues

现在这个问题被标记为了bug，已经进行了修复，现在出现了这个问题的解决方案就是重新pull最新的代码下来。