sigcgans / conditional-sig-wasserstein-gans Goto Github PK

Hi,

I am puzzled about the range of the concatenation in the rolling_window function.
The range is created with length

x.shape[1] - x_lag

Which means the last sample added will be

 x[:, x.shape[1] - x_lag - 1:x.shape[1] - 1]

In doing this we are not including the last sample we can build from the data which would be

x[:, x.shape[1] - x_lag:x.shape[1]]

Therefore I would say that we need the range over:

range(x.shape[1] - x_lag + 1)

Hi, I am a student and would like to use your code for research purposes, but I can't get the code to work.

When running "%run train.py -use_cuda -total_steps 1000" I get an error: "AssertionError: Torch not compiled with CUDA enabled". Also when running without cuda: "%run train.py -total_steps 1000" I get the same error.

I installed the packages from requirements.yml in a conda environment.

File "evaluate.py", line 55, in get_algo_config
key += '_' + experiment_dir.split('/')[2]
IndexError: list index out of range

How to generate time series data and inverse signiture transform?

In create_summary() in plot.py, line 154, the variables x_fake_future has a different shape than x_real because x_fake only contains the future whereas x_real also contains the past. This means the charts do technically not compare the same thing.

Hello, I am really enjoying your great work as mathematics student. Also, thank you for sharing your codebase.

I want to try your model for my problem, and actually I am having trouble getting data
The below is the message I face when running the train code instructed as readme.md

requests.exceptions.connectionerror: httpsconnectionpool(host='realized.oxford-man.ox.ac.uk', port=443): max retries exceeded with url: /images/oxfordmanrealizedvolatilityindices.zip (caused by newconnectionerror('<urllib3.connection.verifiedhttpsconnection object at 0x7fedde54b8e0>: failed to establish a new connection: [errno -2] name or service not known'))

also, I can not access the link provided in the readme.md for getting the dataset.

Can you help me get the training data in your codebase published?

When running

python train.py -use_cuda -total_steps 1000

I get

ImportError: /home/jordi/anaconda3/envs/siglib/lib/python3.8/site-packages/signatory/_impl.cpython-38-x86_64-linux-gnu.so: undefined symbol: _ZNSt15__exception_ptr13exception_ptr10_M_releaseEv

any suggestions?

Hi,

I have noticed that regardless of whatever seed is passed to the function evaluate_generator, the torch seed is set to 0 on line 72.

File "C:\Users\User.conda\envs\gym2\lib\site-packages\signatory\impl.py", line 19, in
from . import _impl
ImportError: DLL load failed: ModuleNotFoundError

Anyone knows if it is possible to set the batch size>1?
Also, what is the reason for checking this condition?
"assert x_real.shape[0] == 1"