vik-goel / morel Goto Github PK

It is an awesome work that I am watching all the time! However, how to run the code? I notice that there is a make_dataset.py, so need we first run it before training? Then why need we first save the dataset rather than directly collecting the batch while interacting with the env? Are there any differences? Thanks.

Hi, I was trying to plot generated objected masks during training but I am unable to get similar results as in paper, can you help me in correcting this issue?
My results look like this :

I am using this command to run code : python train_network.py ../hparams/env=Breakout,algorithm=a2c.json

Thank you

Hi,

I was trying to retrain this model locally, while the model trains successfully and I am able to store the index, meta and data files, I am unable to figure out how to visualize the results. I was trying to use the following command to plot the results:

python -m baselines.run --alg=a2c --env=PongNoFrameskip-v4 --num_timesteps=0 --load_path=my-model-72000.data-00000-of-00001 --play

I put the index, meta and data files generated in the root folder along with PongNoFrameskip-v4 folder to avoid path issues.

The issue is that baselines folder does not have any run.py file. I tried copying the openAi baseline run.py file here but that doesn't seem to work. Can you please help me with how to visuzalize the results of the trained model?