Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
fix two issues trying to run the commands from dev branch
Discrete
has shape()
andnp.prod(()) = 1.0
so need to cast to int when building networkMultiProcessing
however it does look like dev branch is slower (35k vs 50k sps)
1.0
dev
For the
vector.py
change my understanding isMultiProcessing
spawnnum_workers
workers 'sees' a portion of the buffer(num_workers, agents_per_worker, *obs_shape)
concretely let's say this is(6, 8, 160)
(agents_per_worker, *obs_shape) = (8, 160)
. Before this change the are only getting a slice of(1, 160)
becausedriver_env.num_agents
is 1.driver_env
is like the first env; in the worker process we useSerial
so it's like one of the envs inSerial
.(1, 160)
instead of(8, 160)
, then inSerial._assign_buffers
when we assign parts of the buffer to the worker we will end up assigning empty slices and will see errors downstream