-
Notifications
You must be signed in to change notification settings - Fork 140
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Resolves #350 -- Adds support for segmenting LinearPipelines #374
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Looks good.
I know this really isn't part of your diff... but I noticed that in Pipeline.build
we call segment_graph.nodes()
a couple of times, maybe we should save this to a local variable.
If we assume the number of nodes in any segment to be smallish and/or if this call is cheap then maybe this isn't important.
Co-authored-by: David Gardner <96306125+dagardner-nv@users.noreply.github.com>
Remove changes that came in from another MR.
Remove changes from another MR
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
lgtm
@gpucibot merge |
LinearPipeline
,add_segment_boundary([data_type], [is_shared_pointer])
, which will create an egress stage and connected to the most recent linear stage, and an ingress source stage in a new segment. Atbuild
, these stages will be connected.Known issue here: [BUG] Extension libraries compiled with pybind11 don't have cpptype visibility via get_internals MRC#176 : Currently specifying the data type has no effect and all underlying objects will utilize the default PythonObject path. This is due to the method we were using to match the
cpptype
of a wrapped python object to the cpptype of an Ingress/EgressPort adapter. Once we're able to add an out of band path for type detection/storage this will be resolve.Pipeline
to be segment aware, and makes the necessary updates to Pipeline internals to handle the changes.visualize(..)