tesslerc / malmo_rl Goto Github PK
View Code? Open in Web Editor NEWLicense: MIT License
License: MIT License
Tested as well with the unaltered codebase and the training example given in the readme. Over time, memory consumed grows to using up all available ram (nearly 16GBs), at which point either: - Python throws an error and stops the execution, or - Malmo becomes unreachable, throwing a specific error message every step (I can't recall which one, will edit when it occurs again).
I believe this wasn't the case for earlier versions of the codebase as I was previously able to run for days on end without issue.
I'm curious if you recognise this behaviour, and what part of the code might be responsible.
Another thing I just noticed (specifically in my subskill_pickup
domain) is that relatively often, the agent seems to strafe to the right, even when only issuing move 1
commands. This seems very strange. Might it have to do with Malmo not getting time to properly finish executing the command?
Have you encountered something like this?
On the latest version, I get the following error:
> python main.py dqn single_room --double_dqn --normalize_reward --malmo_ports 10001
2018-01-16 15:57:36 INFO: To view results, run 'python -m visdom.server'
2018-01-16 15:57:36 INFO: then head over to http://localhost:8097
Traceback (most recent call last):
File "main.py", line 79, in <module>
True)
File "C:\Github_Repos\tessler_Malmo\utilities\helpers.py", line 52, in play_full_episode
action = policy.get_action(states, is_train)
File "C:\Github_Repos\tessler_Malmo\policies\dqn.py", line 125, in get_action
self.previous_states = np.expand_dims(self.previous_states, dim=0)
TypeError: expand_dims() got an unexpected keyword argument 'dim'
The relevant commit is 9866be5.
Did you mean to use 'axis' in place of dim?
Helpers.py
throws an exception at line 52 when used with the interactive
agent because that agent is missing the learn_start
parameter:
PS C:\Github_Repos\tessler_Malmo> python main.py interactive single_room
INFO: To view results, run 'python -m visdom.server'
INFO: then head over to http://localhost:8097
INFO: Starting Malmo:
INFO: Malmo loaded!
Traceback (most recent call last):
File "main.py", line 55, in <module>
viz)
File "C:\Github_Repos\tessler_Malmo\utilities\helpers.py", line 52, in play_full_episode
if step > params.learn_start and is_train:
AttributeError: 'Namespace' object has no attribute 'learn_start'
PS C:\Github_Repos\tessler_Malmo>
A simple check if that parameter exists might suffice, but I'm not sure if you would maybe rather want to check for / catch that error in some other way.
After executing the agent with the standard parameters as such: python main.py dqn single_room --double_dqn --normalize_reward
, the world is built as normal and the agent starts acting in the world.
However, the minecraft client seems visually very slow, laggy or missing frames. The visdom window in the browser however shows more visual frames being captured than that the client shows.
After ~10 seconds of acting, both the minecraft client and the terminal seem to hang indefinitely and are unresponsive. The terminal does not produce any further logged or printed messages, nor does the visdom window produce any frames. After forcibly closing the minecraft client, the terminal continues reporting the following exceptions every few seconds:
--- Logging error ---
Traceback (most recent call last):
File "C:\Github_Repos\tessler_Malmo\agents\agent.py", line 116, in _load_mission_from_xml
self.agent_host.startMission(mission, self.client_pool, mission_record, 0, str(self.experiment_id))
agents.malmo_dependencies.MalmoPython.MissionException: Failed to find an available client for this mission - tried all the clients in the supplied client pool.
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "C:\Program Files\Anaconda3\lib\logging\__init__.py", line 992, in emit
msg = self.format(record)
File "C:\Program Files\Anaconda3\lib\logging\__init__.py", line 838, in format
return fmt.format(record)
File "C:\Program Files\Anaconda3\lib\logging\__init__.py", line 575, in format
record.message = record.getMessage()
File "C:\Program Files\Anaconda3\lib\logging\__init__.py", line 338, in getMessage
msg = msg % self.args
TypeError: not all arguments converted during string formatting
Call stack:
File "C:\Program Files\Anaconda3\lib\threading.py", line 884, in _bootstrap
self._bootstrap_inner()
File "C:\Program Files\Anaconda3\lib\threading.py", line 916, in _bootstrap_inner
self.run()
File "C:\Program Files\Anaconda3\lib\threading.py", line 864, in run
self._target(*self._args, **self._kwargs)
File "C:\Github_Repos\tessler_Malmo\utilities\parallel_agents_wrapper.py", line 54, in agent_perform_action
reward, terminal, state, terminal_due_to_timeout = agent.perform_action(action)
File "C:\Github_Repos\tessler_Malmo\agents\agent.py", line 142, in perform_action
self._restart_world()
File "C:\Github_Repos\tessler_Malmo\agents\single_room.py", line 31, in _restart_world
self._load_mission_from_xml(mission_xml)
File "C:\Github_Repos\tessler_Malmo\agents\agent.py", line 121, in _load_mission_from_xml
logging.critical('_load_mission_from_xml, Error starting mission', e)
Message: '_load_mission_from_xml, Error starting mission'
Arguments: (MissionException('Failed to find an available client for this mission - tried all the clients in the supplied client pool.',),)
But I expect that all these exceptions are due to the client no longer running.
I cannot forcibly interrupt the process in the terminal other than closing the window.
For completeness, you can see the full behaviour recorded here: https://youtu.be/tzDqEF1s2wk . Around 1:20, the world finished building and the agent starts interacting. Around 1:30, the windows become unresponsive and start to hang indefinitely.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
๐ Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. ๐๐๐
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google โค๏ธ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.