pytorch / ELF

ELF: a platform for game research with AlphaGoZero/AlphaZero reimplementation
Other
3.37k stars 566 forks source link

'_elfgames_go.ContextOptions' object has no attribute 'verbose_comm' #86

Closed Kayoku closed 6 years ago

Kayoku commented 6 years ago

Hi,

I just installed ELF and follow all the install process. Make & Make test passed. I try the section "Running a Go bot" but it fail at the final command.

"./gtp.sh path/to/modelfile.bin --verbose --gpu 0 --num_block 20 --dim 224 --mcts_puct 1.50 --batchsize 16 --mcts_rollout_per_batch 16 --mcts_threads 2 --mcts_rollout_per_thread 8192 --resign_thres 0.05 --mcts_virtual_loss 1"

With this error:

[2018-08-14 11:27:13.476] [rlpytorch.model_loader.load_env] [info] Finished loading env Traceback (most recent call last): File "df_console.py", line 86, in <module> main() File "df_console.py", line 44, in main GC = env["game"].initialize() File "/home/jordan/git/ELF/src_py/elfgames/go/game.py", line 357, in initialize co, GC, opt = self._set_params() File "/home/jordan/git/ELF/src_py/elfgames/go/game.py", line 271, in _set_params self.context_args.initialize(co) File "/home/jordan/git/ELF/src_py/elf/context_utils.py", line 100, in initialize co.verbose_comm = options.verbose_comm AttributeError: '_elfgames_go.ContextOptions' object has no attribute 'verbose_comm'

I don't think it's a specific error about my configuration, so I guess it's recent commit that break your code. Hope it help.

Kayoku commented 6 years ago

Hi,

Ok so. I solved the problem just by removing the line

co.verbose_comm = options.verbose_comm

in the file src_py/elf/context_utils.py.

I found an other bug when I quit the GTP interface (with the "quit" command). I got this:

[2018-08-14 16:36:05.116] [elf::base::Context-3] [info] Prepare to stop ... [2018-08-14 16:36:05.117] [elfgames::go::GoGameSelfPlay-0-16] [warning] Invalid move: x = 0 y = 1 move: ab please try again [2018-08-14 16:36:05.118] [elfgames::go::mcts::MCTSActor-25] [error] model version 1 and required version 1920000 are not consistent terminate called after throwing an instance of 'std::runtime_error' what(): model version 1 and required version 1920000 are not consistent

I'm using the last .bin in Release section, so I don't know what is wrong.

jma127 commented 6 years ago

Thanks for the report. The first issue is fixed in https://github.com/pytorch/ELF/pull/87 . Currently investigating the second issue.

jma127 commented 6 years ago

Ah, never mind this is pre-existing behavior. Would be good to fix but requires some sighandler twiddling.

AFAIK this should not affect your ability to use the GTP interface (before running quit).