index < sizes[i] && "index out of bounds"` failed

[windHat]

Hello!When run
python main.py --task kamed --super_rate 0.5 --model vrbot --train_batch_size 16 --test_batch_size 32 --device 0 --worker_num 5 ,
I got an error :
src/ATen/native/cuda/IndexKernel.cu:60: lambda ->auto::operator()(int)->auto: block: [244,0,0], thread: [127,0,0] Assertion index >= -sizes[i] && index < sizes[i] && "index out of bounds" failed.
And I try other task still got the same error.I do not know how to fix this?

[lddsdu]

Hi！ Please check if the code could run through when setting the device to torch.device("cpu:0"), if it could, please check the version of Pytorch. windHat ***@***.***> 于2021年8月13日周五 上午7:55写道：

…

Hello!When run python main.py --task kamed --super_rate 0.5 --model vrbot --train_batch_size 16 --test_batch_size 32 --device 0 --worker_num 5 , I got an error : src/ATen/native/cuda/IndexKernel.cu:60: lambda <http://int>->auto::operator()(int)->auto: block: [244,0,0], thread: [127,0,0] Assertion index >= -sizes[i] && index < sizes[i] && "index out of bounds" failed. And I try other task still got the same error.I do not know how to fix this? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AI5KSRVTZZ2BIC6N5GC4Q6LT4RNOVANCNFSM5CCOHOCQ> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&utm_campaign=notification-email> .

[windHat]

OK!I will try it with cpu.Actually,my torch version ==1.4.0

[windHat]

When I changed some params,python main.py --task kamed --super_rate 0.5 --model vrbot --train_batch_size 8 --test_batch_size 16 --device 2 --worker_num 5
I got another error :

How to fix it ?

[windHat]

When I run with cpu.I got the same error!

[lddsdu]

When I changed some params,python main.py --task kamed --super_rate 0.5 --model vrbot --train_batch_size 8 --test_batch_size 16 --device 2 --worker_num 5
I got another error :

How to fix it ?

kamed is a dataset without any annotation, please set super_rate to 0

[lddsdu]

When I run with cpu.I got the same error!

I rechecked the code again, and it could run through in both CPU and GPU along with all datasets, so please email me the log information to ensure I could locate the error quickly.

[windHat]

OK,.Thansk!I will try this again! Now I had run task medgg successfully,but did not generate ckpt file.Do I have to set param --force_ckpt_dump？

[lddsdu]

The project will not dump the model weight to the disk if the metrics are lower than the previous validation turn when not set "--force_ckpt_dump", this setting ensures the model dumps the weight to disk in each validation turn. windHat ***@***.***> 于2021年8月18日周三 上午10:15写道：

…

OK,.Thansk!I will try this again! Now I had run task medgg successfully,but did not generate ckpt file.Do I have to set param --force_ckpt_dump？ — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#1 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AI5KSRW2VRK5CGOS2A7WWQDT5MJTNANCNFSM5CCOHOCQ> . Triage notifications on the go with GitHub Mobile for iOS <https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675> or Android <https://play.google.com/store/apps/details?id=com.github.android&utm_campaign=notification-email> .

[windHat]

I see it,Thanks! if I want to define more relation in graph，what else do I need to change before I change the graph.json file?

[windHat]

Whether the graph_loader.py file is redundant?I didn't see the model using it.

[windHat]

Whether the graph_loader.py file is redundant?I didn't see the model using it.