Is this the latest code version ？

[liushenme]

Hi, I'm trying to reproduce your code, but I found some bugs,such as this

(probably my pytorch version problem, I use 0.4.0). Is this the latest version of your code? If there is an update, please update it. Thanks

liu

[qiuqiangkong]

Hi,

Have you tried to print(e.shape) and print(bgn)?

Best wishes,

Qiuqiang

---

From: liushenme notifications@github.com Sent: 14 January 2020 09:53 To: qiuqiangkong/sound_event_detection_dcase2017_task4 sound_event_detection_dcase2017_task4@noreply.github.com Cc: Subscribed subscribed@noreply.github.com Subject: [qiuqiangkong/sound_event_detection_dcase2017_task4] Is this the latest code version ？ (#1)

Hi, I'm trying to reproduce your code, but I found some bugs,such as this

[image]https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fuser-images.githubusercontent.com%2F49808596%2F72333414-b97c3080-36f6-11ea-872d-88d39ae3d171.png&data=02%7C01%7Cq.kong%40surrey.ac.uk%7C9b031196932641c832d508d798d7a344%7C6b902693107440aa9e21d89446a2ebb5%7C0%7C0%7C637145924234554645&sdata=xfenoE%2BJX%2BlxCwUUnHxHIkT80H0NffBTKQWsdn7R0Xw%3D&reserved=0

(probably my pytorch version problem, I use 0.4.0). Is this the latest version of your code? If there is an update, please update it. Thanks

liu

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHubhttps://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fqiuqiangkong%2Fsound_event_detection_dcase2017_task4%2Fissues%2F1%3Femail_source%3Dnotifications%26email_token%3DADFXTSNW25AXLWP63YCSUHDQ5WDSLA5CNFSM4KGQCDGKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4IF77ARQ&data=02%7C01%7Cq.kong%40surrey.ac.uk%7C9b031196932641c832d508d798d7a344%7C6b902693107440aa9e21d89446a2ebb5%7C0%7C0%7C637145924234564640&sdata=L3TV6SV2aQqiGoF95Vl0b0EFDz6mKZYTALZ9VT1Rh6w%3D&reserved=0, or unsubscribehttps://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FADFXTSJDCJ5W6LKI7UWKG5LQ5WDSLANCNFSM4KGQCDGA&data=02%7C01%7Cq.kong%40surrey.ac.uk%7C9b031196932641c832d508d798d7a344%7C6b902693107440aa9e21d89446a2ebb5%7C0%7C0%7C637145924234574633&sdata=%2B8GcxHp1%2BrmocfYwnAZXo5JWt%2BBNQmcr8pxhabLjt2o%3D&reserved=0.

[liushenme]

Hi, Yes, I have solved this problem. I coerced bgn and distance to int and set the distance low to 1 to avoid another bug. Have you never encountered this problem? And I have another question. Are you training with a single GPU? The time I spent training with a single GPU is below. I feel that it takes too long. Do you have any good ways to improve it? Do you need to train 60,000 iterations to achieve the results in your paper? Can you tell me more about your tricks? Oh, I thought of another question, why did you not extract the features in advance and pack them well, but extracted the features during the training process? Sorry I have too many questions. 😭 Thank you.

Best wishes, Liu