John Agapiou
John Agapiou
Fundamentally, there's a API incompatibility here: a gym `reset(seed)` cannot be forwarded to a dmenv `reset()`. So you'll have to rebuild the Substrate/Scenario on reset as you say. ~FYI right...
Sorry update here: it's not currently possible to change the substrate seed. [builder](https://github.com/deepmind/meltingpot/blob/main/meltingpot/python/utils/substrates/builder.py ) takes an `env_seed` that allows this but it is [not currently wired in](https://github.com/deepmind/meltingpot/blob/main/meltingpot/python/utils/substrates/substrate.py).
Amazing, yes please!
FYI: @dimonenka's PR updating rllib: https://github.com/deepmind/meltingpot/pull/153
Not the best answer for long-term storage/sharing (and we've not tested it), but have you tried `pickle`?
Fixed and will be released in v2.4
`api_key` is already model specific: some models don't support that argument. So I think it's OK to solve it the same way: have `device` default to `None` and only forward...
@rstrivedi I'll see what @duenez wants to do, as he's been handling the examples. @gogo16888 as you've been finding and as this issue documents, the examples don't work. See Rakshit's...