search

wenwenyu / PICK-pytorch

Code for the paper "PICK: Processing Key Information Extraction from Documents using Improved Graph Learning-Convolutional Networks" (ICPR 2020)
https://arxiv.org/abs/2004.07464
MIT License
553 stars 191 forks source link

Is it possible to train it in a non-distributed way? #29

Open boli2016 opened 3 years ago

boli2016 commented 3 years ago

Hi wenwenyu,

I am really interested in this work. I am just wondering is it possible for me to train it in a non-distributed way?

Thanks a lot!

wenwenyu commented 3 years ago

Hi,

Thanks for your interest in our work. It can be run in a non-distributed way but needs to modified code. I will support this training mode in this repo as soon as possible. I have lots of things to do, please stay tuned and wait patiently.

tengerye commented 3 years ago

@wenwenyu Hi, I have modified the code to make it run in a non-distributed way. I have also added an example as how to train it on a big dataset. I will push it this week.

wenwenyu commented 3 years ago

@tengerye Hi, I'm very glad that you have finished it. This is also an awesome work about adding a training example on big datasets. Pull requests are absolutely welcome.

boli2016 commented 3 years ago

Thank you so much @wenwenyu @tengerye !

tengerye commented 3 years ago

@boli2016 @wenwenyu Hi, apologize. This pull does not include the "non-distributed" way. There are two options here now:

  1. We add "non-distributed" code and leave it different with "distributed" code;

  2. We make python control the distribution instead of torch.distributed.launch. In other words, "distributed" codes works by invoking "non-distributed" codes. I prefer this way.

Dear @wenwenyu , I just send you an email including more details, would you please have a look at your convenience please?

wenwenyu commented 3 years ago

@tengerye Hi, I've received your email and I will reply you at my convenience time.

cloudfool commented 3 years ago

@tengerye Hi, Could you PLZ send me the "non-distributed" training code? TKS weihao1991@126.com

tengerye commented 3 years ago

@cloudfool Hi, sincerely sorry for the delay. I am quite busy recently. But we are working on this already. My plan is to continue to work on it and try to release it during the National Day.

cloudfool commented 3 years ago

TKS，keep tuned

tengerye commented 3 years ago

@cloudfool @boli2016 I have finished the non-distributed training part. Would you please take a shot? You can find instructions at #41 .

sanku015 commented 3 years ago

@tengerye could you share the codes?

cloudfool commented 3 years ago

@tengerye Thanks a lot