AttributeError: module 'portalocker' has no attribute 'Lock' #4
Comments
|
It seems that there may be wrong with 'Loading checkpoint from detectron2://ImageNetPretrained/MSRA/R-50.pkl |
|
Could you check the version of >>> import portalocker
>>> portalocker.__version__
>>> portalocker.Lock |
|
Sorry, there must be something wrong with my |
|
Try to re-install |
|
first |
|
How to close DDP？ |
|
I think this is the last bug before I can train YOLOF... |
|
Comment out the lines with BTW, when you train with only one GPU, you should adjust the learning rate and batch size. Refer to this response. |
|
Support training with one GPU in this commit. |
|
Thanks ！ |
|
真的很抱歉，作者，由于我的水平太低，这又出现了新的bug。请问这个应该怎么改？ |
|
Could you give more details about what command you use? |
|
command : yaml: MODEL: |
|
Can you try this setting? |
|
I left lab and I will try tomorrow. thanks a lot from bottom of my heart ！
|
|
Good Morning! When I tried your setting, it stiill remains the same bug as: [03/26 17:38:53 d2.engine.train_loop]: Starting training from iteration 0 /opt/conda/conda-bld/pytorch_1607370169888/work/aten/src/ATen/native/cuda/IndexKernel.cu:84: operator(): block: [0,0,0], thread: [11,0,0] Assertion ERROR [03/26 17:39:00 d2.engine.train_loop]: Exception during training: |
|
Sorry, the Try to warm up more iterations, e.g., |
|
Thanks. It stil does not work... |
|
Could you upload your training log file? |
|
[03/26 21:44:54] detectron2 INFO: Rank of current process: 0. World size: 2 sys.platform linux PyTorch built with:
[03/26 21:44:55] detectron2 INFO: Command line arguments: Namespace(config_file='./configs/yolof_R_50_C5_1x.yaml', dist_url='tcp://127.0.0.1:50159', eval_only=False, machine_rank=0, num_gpus=2, num_machines=1, opts=['OUTPUT_DIR', '/hdd2/wh/cw/train/yolof/R_50_C5_1x/'], resume=False) [03/26 21:44:55] detectron2 INFO: Running with full config: |
|
[03/26 21:45:54 d2.engine.train_loop]: Starting training from iteration 0 Traceback (most recent call last): -- Process 1 terminated with the following error: |
|
command : yaml: OUTPUT_DIR: '/hdd2/wh/cw/train/yolof/R_50_C5_1x' |
|
作者您好，我粗浅的认为是跟数据有关的错误，可能是发生了数组越界等？ 但我的数据集是coco2017，不应该有这个错误才对，其余的我也没有更改了，我原本的detectron2也重新升级了一下。 |
|
看错误是在取index的时候越界了，但是很奇怪，我之前跑过这么多次都没有遇到过这个问题。log文件一眼看过去也并没有找到明显不对的地方，感觉没啥道理。。。你这个是每次一跑，必然会出现这个错误嘛？ |
|
|
|
|
1.gt_class.size() torch.Size([38000]) 2.gt_class.size() torch.Size([42000]) 3.gt_class.size() torch.Size([38000]) /opt/conda/conda-bld/pytorch_1595629408163/work/aten/src/ATen/native/cuda/IndexKernel.cu:84: operator(): block: [0,0,0], thread: [9,0,0] Assertion |
|
感觉这里不应该有错，建议用 看看到底哪里出错了。 |
|
还是不行，可能只能用其他机器试一下， 我不确定我的cuda 9.0是否会对这个有影响 。 |
|
作者您好 我换了一台cuda 10.1的然后重复了我之前的操作，（基本就是直接安装了,数据集也是直接从原来的服务器上传输的）。 |
|
另外有一个小地方，我觉得您可以考虑修改下 。 |
Thanks for sharing your great work. I am sorry that I have a bug when I use
python ./tools/train_net.py --num-gpus 1 --config-file ./configs/yolof_R_50_C5_1x.yamlBug log below as :
[03/26 07:38:03 d2.data.build]: Using training sampler TrainingSampler
[03/26 07:38:03 d2.data.common]: Serializing 117266 elements to byte tensors and concatenating them all ...
[03/26 07:38:10 d2.data.common]: Serialized dataset takes 451.21 MiB
[03/26 07:38:15 fvcore.common.checkpoint]: Loading checkpoint from detectron2://ImageNetPretrained/MSRA/R-50.pkl
Traceback (most recent call last):
File "./tools/train_net.py", line 234, in
args=(args,),
File "/home/cw/miniconda3/envs/yolof/lib/python3.6/site-packages/detectron2/engine/launch.py", line 62, in launch
main_func(*args)
File "./tools/train_net.py", line 215, in main
trainer.resume_or_load(resume=args.resume)
File "/home/cw/miniconda3/envs/yolof/lib/python3.6/site-packages/detectron2/engine/defaults.py", line 353, in resume_or_load
checkpoint = self.checkpointer.resume_or_load(self.cfg.MODEL.WEIGHTS, resume=resume)
File "/home/cw/miniconda3/envs/yolof/lib/python3.6/site-packages/fvcore/common/checkpoint.py", line 215, in resume_or_load
return self.load(path, checkpointables=[])
File "/home/cw/miniconda3/envs/yolof/lib/python3.6/site-packages/fvcore/common/checkpoint.py", line 140, in load
path = self.path_manager.get_local_path(path)
File "/home/cw/miniconda3/envs/yolof/lib/python3.6/site-packages/iopath/common/file_io.py", line 1100, in get_local_path
path, force=force, **kwargs
File "/home/cw/miniconda3/envs/yolof/lib/python3.6/site-packages/detectron2/utils/file_io.py", line 29, in _get_local_path
return PathManager.get_local_path(self.S3_DETECTRON2_PREFIX + name, **kwargs)
File "/home/cw/miniconda3/envs/yolof/lib/python3.6/site-packages/iopath/common/file_io.py", line 1100, in get_local_path
path, force=force, **kwargs
File "/home/cw/miniconda3/envs/yolof/lib/python3.6/site-packages/iopath/common/file_io.py", line 755, in _get_local_path
with file_lock(cached):
File "/home/cw/miniconda3/envs/yolof/lib/python3.6/site-packages/iopath/common/file_io.py", line 82, in file_lock
return portalocker.Lock(path + ".lock", timeout=3600) # type: ignore
AttributeError: module 'portalocker' has no attribute 'Lock'
I woule be grateful if you could give me some advice. Thanks.
The text was updated successfully, but these errors were encountered: