Add scripts for graph-level classification #107
Conversation
ChrisCummins
force-pushed
the
feature/classifyapp_81
branch
from
9b10303 to
4b33172
Compare
| --log_dir LOG_DIR Directory(*) to store logfiles and trained models relative to repository dir. | ||
| [default: programl/task/graph_level_classification/logs/unspecified] |
There was a problem hiding this comment.
@ChrisCummins would you like to specify a default log dir here? it is relative to the repository dir.
There was a problem hiding this comment.
For the dataflow experiment I put logs/ inside the root of the dataset, see: https://github.com/ChrisCummins/ProGraML/blob/development/programl/Documentation/DataflowDataset.md#directory-layout
I think it makes sense to do that with the classifyapp/devmap too
|
@ChrisCummins I used to have a |
It is still there, but it won't show up until I hook it into the bazel build system. In the meantime, you can compile the proto files yourself with: $ protoc programl/proto/*.proto --python_out=. |
github.com//issues/81
We write logs into the dataset directory, not in the source tree. github.com//issues/81
ChrisCummins
force-pushed
the
feature/classifyapp_81
branch
from
b919eb9 to
2876ad5
Compare
ChrisCummins
added
Enhancement
Work-in-progress PR which adds an implementation of graph-level classification, as used for the downstream classifyapp and devmap tasks in the arXiv paper.