As I attempted to start training using the command provided in the readme:

python run.py --adhoc --cfg conf/c02_selfplay/liars_sp.yaml \
    env.num_dice=1 \
    env.num_faces=4 \
    env.subgame_params.use_cfr=true \
    selfplay.cpu_gen_threads=0  \
    selfplay.threads_per_gpu=8

I encountered the following error:

Traceback (most recent call last):
File "run.py", line 25, in
import cfvpy.tasks
File "/home/myname/rebel-main/cfvpy/tasks.py", line 15, in
import cfvpy.selfplay
File "/home/myname/rebel-main/cfvpy/selfplay.py", line 27, in
import cfvpy.rela
ImportError: /home/myname/rebel-main/cfvpy/rela.cpython-37m-x86_64-linux-gnu.so: undefined symbol: _ZN8pybind116detail11type_casterIN2at6TensorEvE4loadENS_6handleEb

I am not sure what really went wrong. My guess is this has something to do with the make process. I appreciate it if anyone could shed some light on this issue.

Additional info:
When I first attempted make, the compiler was complaining about not being able to find py::module_ in /opt/conda/envs/rebel/lib/python3.7/site-packages/torch/include/torch/csrc/utils/pybind.h (line 176)

And according to pybind 11 documentation,

py::module has been renamed py::module_, but a backward compatible typedef has been included

So I changed py::module_::import to py::module::import. I completed make without any error, but perhaps this is not the right thing to do.