Training Error with different datasets

[Anonymous]

In utils. Py, if only the file biwi_hotel.txt appears in the training set, the file will be run to train during the training (2,900 pieces of data in total).If you change this file to another file, that is, comment out the file name and change to the name of other TXT files, all the data sets will be trained (a total of 44000 pieces of data).But I still don't see why it should be all training, and only biwi_hotel.txt can train individual files.

[Baran Nama]

I have commented on just for testing purposes, you can uncomment without trouble.

[Anonymous]

I have commented on just for testing purposes, you can uncomment without trouble.

Please forgive my not replying in time. And thanks for your patient reply!

[Anonymous]

I have commented on just for testing purposes, you can uncomment without trouble.

Excuse me, after training, I encountered two more problems so I hope to ask for your help.

(1)When I finished implementing python3 ./vlstm_train.py, and the results are showed as follows. Then I began to implement python3 ./test.py, whose results are showed as follows. So why the mean error and final error reach so extremely small(10^(-9))? Is it caused by my implementation fault?
(2)After running test.py, I began to implement python3 ./visuailize.py, but it turned out to be the error as follows. My configuration environment are:
Ubuntu 18.04; pytorch 0.4.0, cuda 10, GTX 1650, matplotlib 3.1.2. I searched for so long but in vain, so could you please tell me how to tackle it and see correct visualization?

Thanks a lot in advance! Feel sincerely sorry to bother you!

[Baran Nama]

1. It can be a logical error or something else. Please double-check the code blocks that implement equations in the original paper (in the readme).
2. install ffmpeg (matplotlib/matplotlib#8053).

[Anonymous]

1. It can be a logical error or something else. Please double-check the code blocks that implement equations in the original paper (in the readme).

2. install ffmpeg ([matplotlib/matplotlib#8053](https://github.com/matplotlib/matplotlib/issues/8053)).

Please forgive my not replying in time first, cause there was something wrong with my Wi-Fi. Thanks for your suggestion, and I successfully ran visualization.py with installing ffmpeg. HOWEVER, I still could not successfully see the correct results on test.py, which means the ADE and FDE always remained the same, no matter what methods I tried. That is to say, after running train.py, olstm_train.py or vlstm_train.py, I obtained three completely SAME ADEs and FDEs with running test.py, whose errors all reached extremely small... I believe there might be something wrong with the code in test.py, but I haven't find it...
Anyway, thank you very much! And could you please tell me that could you get correct results under your own implementations? Thanks in advance!

[Huishun Zhang]

Hi YihaoChan, I am also working on the social lstm project, can we communicate by email? My email is [e-mail]

[lk-greenbird]

Hi YihaoChan, I am working on the social lstm project recently. But a TypeError occurs when I run the train.py with python3.5 and pytorch0.1.12 on ubuntu16.04. I want to konw weather the problem dues to the verson of pytorch. Can you tell me your environment which can excute the project successfully.Thank you! My email is [e-mail]

[Anonymous]

Hi YihaoChan, I am working on the social lstm project recently. But a TypeError occurs when I run the train.py with python3.5 and pytorch0.1.12 on ubuntu16.04. I want to konw weather the problem dues to the verson of pytorch. Can you tell me your environment which can excute the project successfully.Thank you! My email is [e-mail]

I've sent a reply to your qq mail~

[lee]

@YihaoChan Hello, I am also trying to learn this project recently. Maybe due to my lack of understanding. I can run through the original project data set. When I used my own data to make a dataset similar to the original project, I encountered some problems. I don't know how to solve it, can I talk to you.Thank you！ My email is [e-mail]