Skip to content

Errors in reproducing the experiments #2

@luigiberducci

Description

@luigiberducci

Hi Daniel,
Thanks for sharing your work!

I wanted to reproduce the experiments you did but there is some error thrown when calling model.predict().
When calling it, you often pass apply_sigmoid=False but this is not an argument to the method.
This throws an error and ends the training loop.

I am not sure if it is some old code, now deprecated, that can be removed.
Could you please check?


To reproduce:
I simply run the script in the experiments directory:

bash experiments/training/cartpole_multiple_seeds.sh

Note: I had to fix the path to cartpole.sh in that script because it was not correct.
I changed from:

for SEED in 1 2 3 4 5
do
    bash experiments/cartpole.sh $SEED $group_name
done

to

for SEED in 1 2 3 4 5
do
    bash experiments/training/cartpole.sh $SEED $group_name
done

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions