pytorch-a3c

pytorch-a3c copied to clipboard

Fixes issue #66 ( env wrappers need to implement observation() )

briantimar

As you mentioned that A2C is strongly suggested except for specific reason. So if I need to run it in distributed processing, (actually only for collecting data in real time),...

QiXuanWang

The `while True:` of https://github.com/ikostrikov/pytorch-a3c/blob/master/train.py#L35 cannot be break, because the only `break` statement is in https://github.com/ikostrikov/pytorch-a3c/blob/master/train.py#L79 which is used to break for-loop: https://github.com/ikostrikov/pytorch-a3c/blob/master/train.py#L50 How to terminate that forever while-loop in...

machanic

Multi-processing or multi-threading

1

Hi, In the [original paper](https://arxiv.org/pdf/1602.01783.pdf), it mentions that it uses multi threads. But I see in your code, you are using multi-process. As far as I know, these two methods...

lingzhang0319

I'm sorry to ask a simple question. I don't know the difference between the 'Pong-v4' and 'PongDeterministic-v4'. And why you use the latter environment to test your algorithm instead of...

HuiSiqi

how to under ensure ensure_shared_grads?

1

I am kind of confused of the ensure_shared_grads here https://github.com/ikostrikov/pytorch-a3c/blob/master/train.py#L13. Here, the `grad` is synced only when it is `None`. I think we need to set `shared_param._grad = param.grad` all...

luochao1024

Did lstm cell really make more sense in A3C?

1

Very sorry to ask you a simple question, thanks a lot.

WonderSeven

Hi I make some small changes to clear all warnings corresponds to torch and gym old versions. I also add tensorboard to tester agent in order to monitor learning process...

mohamad-hasan-sohan-ajini

When using no-shared = False, the process is blocked

10

Hi,Today, i run the code, and found that when no-shared=False, the process will be blocked. Do you have any suggesstions to fix that? THANKS!

keithyin

Zbranch

7

Hi, added simple logging with tensorboard logger. (no dependencies on tensorflow) If you want to keep it simple and minimal it's ok to reject :) training time here is around...

scientist1642