search

devendrachaplot / DeepRL-Grounding

Train an RL agent to execute natural language instructions in a 3D Environment (PyTorch)
https://sites.google.com/view/gated-attention/home
MIT License
237 stars 37 forks source link

Hardware specifications #4

Closed a7b23 closed 6 years ago

a7b23 commented 6 years ago

I have been running the code on 32 threads since 3 days but still haven't been able to get avg Accuracy of more than 0.5 for the 'easy' difficulty. Can you specify the hardware that was used for getting the result mentioned in the paper?

devendrachaplot commented 6 years ago

The experiments in the paper were run on a DGX-1 with two Intel(R) Xeon(R) CPU E5-2698 v4 @ 2.20GHz CPUs. It takes about 35K iterations per thread with 32 threads to reach about 0.5 accuracy on 'easy' environment. What is the number of iterations for you on each thread after 3 days?

a7b23 commented 6 years ago

It had reached only 15k steps and that is why I guess the performance is low. Thanks for specifying the specifications, will try on a high end machine.

rshivansh commented 6 years ago

@devendrachaplot I have been training it on easy difficulty level its 3 days now and avg accuracy has still not crossed 0.35. My system is 48 core cpu.