tensorflow / tensorflow/tensorboard
Inferred logdir for `tf.summary.trace_export` broken
Nobody has claimed this yet.
- Dominant language
- TypeScript
- Stars
- 7.2k
- Forks
- 1.7k
- Avg merge
- 4d 22h
- Merged PRs (30d)
- 1
Description
Feature added in https://github.com/tensorflow/tensorflow/pull/35340,
but I have no evidence that it ever worked (it has no tests, and is
broken in TensorFlow 2.2.0 through head). Simple repro:
import tensorflow as tf
writer = tf.summary.create_file_writer("logs")
with writer.as_default():
tf.summary.trace_on(profiler=True)
tf.summary.trace_export("test", step=0)
The last line fails with:
Traceback (most recent call last):
File "test.py", line 6, in <module>
tf.summary.trace_export("test", step=0)
File "/usr/local/google/home/wchargin/virtualenv/tf-nightly-20201214-py3.8/lib/python3.8/site-packages/tensorflow/python/ops/summary_ops_v2.py", line 1314, in trace_export
_profiler.save(profiler_outdir, _profiler.stop())
File "/usr/local/google/home/wchargin/virtualenv/tf-nightly-20201214-py3.8/lib/python3.8/site-packages/tensorflow/python/util/deprecation.py", line 337, in new_func
return func(*args, **kwargs)
File "/usr/local/google/home/wchargin/virtualenv/tf-nightly-20201214-py3.8/lib/python3.8/site-packages/tensorflow/python/eager/profiler.py", line 147, in save
plugin_dir = os.path.join(
File "/usr/lib/python3.8/posixpath.py", line 76, in join
a = os.fspath(a)
TypeError: expected str, bytes or os.PathLike object, not tensorflow.python.framework.ops.EagerTensor
(The error is slightly different in TensorFlow 2.2.0 on Python 3.5, but
that’s just because of the Python version difference.)
Stepping with pdb shows that logdir on line R1234 of that PR is set
to <tf.Tensor: shape=(), dtype=string, numpy=b'logs'>, which
explains why the later call to os.path.join(logdir, ...) would fail in
the internals of _profiler.save.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Reproduce the failure with the Python example, then inspect trace_export in tensorflow/python/ops/summary_ops_v2.py and the path handling in tensorflow/python/eager/profiler.py. Add coverage for inferred logdir behavior and verify that the example completes without passing an EagerTensor to the filesystem path operations.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python
- Domain
- observability
- Issue type
- Bug
- Difficulty
- 3/5
- Estimated time
- 1-2 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 52/100