Computation time

[huntforgz]

Hi I am runing your code on Tesla P100 GPU(one),but computation time is much over than 22h.Can you tell me what's the device implementation you have?

[Ovishake]

My apologies for commenting on someone else's code, but this code is pretty well done I guess. DenseNet169 and DenseNet201 takes a lot of time to complete if you are using the entire dataset. Split the dataset into 1% of the whole. The runtime will be 1 hour for the entire Dataset for DenseNet169. I use 2 Tesla K80 gpu for training.

[huntforgz]

I ran all the data on simple TeslaP100 GPU and it took like forever to train the densenet 121.I thought the gpu I used was much stronger than K80.By the way,why you can compile the code while you passed the async=True in target function?async is keyword of pytorch.What’s your pytorch and cuda version ?This is really important for me. Get Outlook for iOS<https://aka.ms/o0ukef>

…

________________________________ From: Avishek Mukherjee <[email protected]> Sent: Saturday, April 6, 2019 10:10:41 PM To: zoogzog/chexnet Cc: ghh; Author Subject: Re: [zoogzog/chexnet] Computation time (#18) My apologies for commenting on someone else's code, but this code is pretty well done I guess. DenseNet169 and DenseNet201 takes a lot of time to complete if you are using the entire dataset. Split the dataset into 1% of the whole. The runtime will be 1 hour for the entire Dataset for DenseNet169. I use 2 Tesla K80 gpu for training. ― You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<#18 (comment)>, or mute the thread<https://github.com/notifications/unsubscribe-auth/APpryacXYvbf5QfpUrM2tWwe6JP2Lgvlks5veVOhgaJpZM4cb0nM>.