Not working in ML Agents 1.0.0
firdiar-mita opened this issue · comments
unable to start training with ml agent
i was using this command
py mlagents-learn.py ./config/AutoV1_19.yaml --run-id AutoV1_40 --deterministic --force
i found solution,
in my case, i need to specify --base-port=5004
btw im changing the start process method
def start_process(self, i: int) -> None:
n: int = self.run_count
self.run_count += 1
args: List[str] = self.args.get_process_args(n, i, self.config.config_paths[n])
print(args);
if platform.system() == 'Windows':
#self.slots[i] = subprocess.Popen(args, creationflags=subprocess.CREATE_NEW_CONSOLE)
startupinfo = subprocess.STARTUPINFO()
startupinfo.dwFlags |= subprocess.STARTF_USESHOWWINDOW
self.slots[i] = subprocess.Popen(args, startupinfo=startupinfo)
elif platform.system() == 'Linux':
self.slots[i] = subprocess.Popen('gnome-terminal -x ' + ' '.join(args), shell=True)
run_id: str = self.args.get_run_id(n)
self.short_run_ids[i] = run_id
self.verbose_run_ids[i] = self.config.verbose_run_id_combos[n]
log(f'{run_id} started.')
this way, no new window appeared, so we can see the error just like ordinaryy mlagents-learn did