Sacha Chernyavskiy
Sacha Chernyavskiy
also, might be a duplicate of https://github.com/google-deepmind/open_spiel/issues/1287
Okay, I'll push the latest fixes for your experiments in the next couple hours
maybe to disable the long tests only for github actions and add a flag to the cmake to make them run during the build?
Linux builds don't fail by themselves, as I see mostly macOS failures. Didn't you investigate into recent libraries upgrades (e.g. `abseil`)?
Hey! what is your operating system version? do you use cuda? I'll try to reproduce and look into the problem.
thank you for the clarification, will check and report shortly
Was able to reproduce the issue on `cpu` with ` ./build/examples/alpha_zero_torch_example --game=leduc_poker --path=$HOME/lp`: ```Bash terminate called after throwing an instance of 'c10::Error' terminate called recursively terminate called recursively terminate called...
@Viren6, the problem was a source of a simple bug: during training, there could be batches with size of 1, which is not allowed by the implementation of `BatchNorm2d`. This...
@fuyuan-li , I found out that $\frac{t}{2T}$ is for the Linear CFR specifically, which shouldn't work with non-linear MLPs, so I decided to stick with $\sqrt{t}$ as in the original...
@fuyuan-li, I added `print_nash_convs` argument for `DeepCFRSolver` to print exploitability of each optimisation step can harm performance, but can help with tracking the progress