Sasha Abramowitz

icon company InstaDeep icon location Cape Town icon location Reinforcement learning research engineer at InstaDeep

Results 83 comments of Sasha Abramowitz

docs: fix bin_pack state doc

Closed in favour of #250

Proposal for Code Structure Improvement Using `jax.lax.cond`

Hi, thanks for the suggestion! Agreed it isn't as clean as the others, but the reason it needs to be done this way is because the `transition` and `termination` function...

Proposal for Code Structure Improvement Using `jax.lax.cond`

Ye that seems like a reasonable solution

Proposal for Code Structure Improvement Using `jax.lax.cond`

@helpingstar would you be interested in making a PR to fix this?

Proposal for Code Structure Improvement Using `jax.lax.cond`

No rush honestly, whenever you have time! I think @clement-bonnet suggested fix should work well

[FEATURE] GNN support for the MARL frameworks

Hi @Jaroan thanks for the issue! This should definitely be possible and is an interesting line of work, however it's not on our roadmap in the near future. But this...

[FEATURE] GNN support for the MARL frameworks

Closing this because it is quite stale, feel free to open it up again if you have any questions :)

[FEATURE] GNN support for the MARL frameworks

Hi @jselvaraaj sure I'll reopen this issue. Let us know if you need any help

[BUG] Actor updates too frequently in SAC with Policy Update Delay

Hi @LeoHink great find, this is indeed a bug! Would you be able to put up a PR for this, your solution looks good to me :smile:

chore: changed the assertions to make sure the num_updates is a multiple of num_evaluations

A suggestion here that would remove the need for the assert and make mava easier to configure is to change the variable to evaluation frequency and then store num_evaluations in...