This is making the output file (when the stream is a file) a valid MLIR
file.
Details
Details
Diff Detail
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
Comment Actions
In TensorFlow: we emit a new file on every invocation right now, and we print a full module. When I get dump for reproduction the only change I make is opening the file and deleting the first line :)