search

stes / torch-associative

Associative Domain Adaptation in PyTorch
https://domainadaptation.org
3 stars 9 forks source link

What are the version of python and pytorch?Thank you! #2

Open 4Statistics opened 6 years ago

4Statistics commented 6 years ago

What are the version of python and pytorch?Thank you very much! because I met a lot of errors.

stes commented 6 years ago

Tested on 0.4.0. Please give more details about your system setup and the error message if that does not resolve the issue. In any case, also consider using the salad library directly: https://github.com/domainadaptation/salad

4Statistics commented 6 years ago

I use python3.6 and Pytorch 0.4.1. I ran the program by "Python3.6 train.py". The error is show below.

RuntimeError: CUDA error: out of memory ####### Thank you very much.

stes commented 5 years ago

@4Statistics , try to decrese the batch size, e.g. using

Python3.6 train.py --sourcebatch 20 --targetbatch 200

The default values are 100 and 1000, respectively, see https://github.com/stes/torch-associative/blob/cb66aef1d2296e54e09f6f970b775d5fd677601a/train.py#L30-L31

Let me know if this fixes the issue.

mrsempress commented 5 years ago

I just run your problem, but get an error like: image Did I ignore something?

mrsempress commented 5 years ago

And if set "Python3.6 train.py --sourcebatch 20 --targetbatch 200", the result looks like overfitting.