windows10 单卡训练出现错误

[QIANXUNZDL123]

您好，我在windows上训练的只有一张卡，训练的报错如下：

Traceback (most recent call last): File "E:\D-FINE\src\nn\backbone\hgnetv2.py", line 498, in init if torch.distributed.get_rank() == 0: File "D:\Anaconda2019\envs\qat\lib\site-packages\torch\distributed\distributed_c10d.py", line 1469, in get_rank default_pg = _get_default_group() File "D:\Anaconda2019\envs\qat\lib\site-packages\torch\distributed\distributed_c10d.py", line 940, in _get_default_group raise RuntimeError( RuntimeError: Default process group has not been initialized, please make sure to call init_process_group.

During handling of the above exception, another exception occurred:

Traceback (most recent call last): File "train.py", line 84, in main(args) File "train.py", line 54, in main solver.fit() File "E:\D-FINE\src\solver\det_solver.py", line 24, in fit self.train() File "E:\D-FINE\src\solver_solver.py", line 81, in train self._setup() File "E:\D-FINE\src\solver_solver.py", line 47, in _setup self.model = cfg.model File "E:\D-FINE\src\core\yaml_config.py", line 38, in model self._model = create(self.yaml_cfg['model'], self.global_cfg) File "E:\D-FINE\src\core\workspace.py", line 146, in create module_kwargs[k] = create(_cfg['_name'], global_cfg) File "E:\D-FINE\src\core\workspace.py", line 180, in create return module(**module_kwargs) File "E:\D-FINE\src\nn\backbone\hgnetv2.py", line 512, in init if torch.distributed.get_rank() == 0: File "D:\Anaconda2019\envs\qat\lib\site-packages\torch\distributed\distributed_c10d.py", line 1469, in get_rank default_pg = _get_default_group() File "D:\Anaconda2019\envs\qat\lib\site-packages\torch\distributed\distributed_c10d.py", line 940, in _get_default_group raise RuntimeError( RuntimeError: Default process group has not been initialized, please make sure to call init_process_group.

训练命令： python train.py -c configs/dfine/dfine_hgnetv2_n_coco.yml 请问我应该如何修改呢？

[lycode1202]

我也遇到了相同的问题，请问您解决了嘛？不知道是不是只支持分布式训练？

[nhfujfgid]

我也遇到了相同的问题，请问您解决了嘛？

[super-song-sir]

我也是遇到了相同的问题，求大佬解答

[QIANXUNZDL123]

我也是遇到了相同的问题，求大佬解答

我也没解决，只能使用预训练模型进行微调了

[ss989462]

现在解决了吗大佬

[nhfujfgid]

没有，放弃了，忙着写大论文了  

6221913077-郑月昌 @.***

 

------------------ 原始邮件 ------------------ 发件人: @.>; 发送时间: 2025年2月21日(星期五) 下午2:01 收件人: @.>; 抄送: @.>; @.>; 主题: Re: [Peterande/D-FINE] windows10 单卡训练出现错误 (Issue #115)

现在解决了吗大佬

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***> ss989462 left a comment (Peterande/D-FINE#115)

现在解决了吗大佬

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: @.***>

[luguoyixiazi]

换pytorch能解决吗？感觉不像是代码本身的问题，应该是平台或者torch的

[ggyda]

我也是遇到了相同的问题，求大佬解答

我也没解决，只能使用预训练模型进行微调了

这个怎么实现呀

[QIANXUNZDL123]

我也是遇到了相同的问题，求大佬解答

我也没解决，只能使用预训练模型进行微调了

这个怎么实现呀

看文档呀，文档里面不是有介绍的吗