Closed dancju closed 1 year ago
thanks for your effort have you tested it? in software engineering, we hope that a PR only solves a problem for review. could you pls split this PR to several PRs?
Hi @zhumingpassional,
I upgraded Python to 3.10 and Gym to Gymnasium in order to fix the dependency issue. Environment classes were upgraded because the API for reset()
and step()
was changed. Therefore, I don't think we should split them into multiple PRs.
I was running Stock_NeurIPS2018_3_Backtest.ipynb to test the new dependency and updated env classes. I can but I don't think it's a good idea to split the test into another PR.
After the fix of this PR, you should be able to install dependencies simply with poetry install
and run Stock_NeurIPS2018_3_Backtest.ipynb
with zero error.
Hi @dancju Thanks for your commits. I just tested the updates, and will do few more. If no errors are found, I will merge the PR soon.
it raises errors in the notebook Stock_NeurIPS2018_SB3.ipynb the PR will be merged after multiple tests this PR will be reverted considering of error could you pls revise the PR and submit it again?
pyfolio is no longer maintained since Quantopian shut down in 2020, which is why I removed it from pyproject.toml
. I have put it back and will create a new PR.
Does that mean we need to upgrade python to 3.10 to be able to use Gymnasium instead of Gym?
@lcavalie You can try yourself.
Edit pyproject.toml
. Change from python = "~3.10"
to python = "~3.9"
.
poetry env use 3.9
poetry lock
poetry install
Stock_NeurIPS2018_3_Backtest.ipynb
and plot the backtesting results of all models (A2C, DDPG, PPO, TD3, SAC).The backtesting results of all models did not outperform the benchmark. Did I do something wrong?