jacobandreas / psketch Goto Github PK

May I ask what algorithm did you use when you train the joint policy and independent policies? ('Joint' and 'Indep' in Figure 4 in your paper) Moreover, what algorithms are implemented in models 'attentive.py' and 'reflex.py'?

Hi,

Sorry i cannot find any cliff enironment in experiments repo as stated in your paper. Can you tell me how to run those cliff experiments?

Regards

May I ask Are the positions of subgoals fixed in all experiment environments, at multitask and zero-shot settings?
At first after reading the paper only, I thought that subpolicy pi is only given s_i (state), so the model can't treat variable subgoal position environments. I read the code and guess the positions are fixed in the code too.

Sorry for my question at this late time(2021).
Thank you in advance :)

Hi Jacob,

I recently started to follow your works.

I just run main.py but it says "No module named reflex". Last time i checked, there is a reflex.py in folder models.

Or could you please show me how to execute the whole program to get the craft experiment result?

Thx

hi Jacob,

i was running your experiment light_modular and i found my gpu-utli(nividia-smi) is always zero. Does the same situation happen to you? Or is it true that your algorithm doesn't have gpu part(just cpu)? sorry this question may sound a little bit stupid.

Hi,
Do you have a pytorch version for this method?
best regards

May I ask which part of the source code shows the results of adaptation (learn the form of a suitable sketch) and zero-shot? All I see is just the training process.
Thanks for the response!

I'd like to see the agent acting in the crafting world(in a GUI way), how to make it? Thanks.

I've tried different versions of Python and Tensorflow but always get this error:

File "/usr/local/lib/python2.7/site-packages/tensorflow_core/init.py", line 40, in
from tensorflow.python.tools import module_util as _module_util
ImportError: No module named tools
What versions am I supposed to use?

Hi, I tried to run your code on my laptop and it works fine. The problem is that I cannot reproduce the result on light-joint-feat-plan task. I changed the experiment name to "test" in comfig.yaml ran the main.py directly. The final result is around 0.3. Are there any parameters that I need to change in order to reproduce the result?