This repository was archived by the owner on Nov 22, 2022. It is now read-only.
Make tensorboard robust to NaN and Inf in model params #1206
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
facebook-github-bot
added
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D19048506 |
…rch#1206) Summary: Pull Request resolved: facebookresearch#1206 PyText FBLearner training runs with `tensorboard` will fail if the model has any `NaN` or `Inf` parameters during training with the following error: `ValueError: The histogram is empty, please file a bug report.` NaN/Inf params show up quite often in hyperparam sweep runs if a bad initial value of hyperparam is chosen. When this happens, the whole sweep can fail. Some examples just from the past two days: f156398891 f156398480 f156398264 f156399047 f156399125 This diff: - Catches tensorboard exceptions, and prints an error to stderr. Note that `TensordboardChannel` already catches some exceptions while exporting the model - Adds a unit-test with a model that has `NaN` and `Inf` parameters, to verify no Tensorboard errors Reviewed By: psuzhanhy Differential Revision: D19048506 fbshipit-source-id: 1449423d9c2502e0ade5d327f4cc597c26edeb64
kmalik22
force-pushed
the
export-D19048506
branch
from
fb2cc50 to
d8ae397
Compare
|
This pull request was exported from Phabricator. Differential Revision: D19048506 |
|
This pull request has been merged in 631e07f. |
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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Outdated suggestions cannot be applied.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.
Summary:
PyText FBLearner training runs with
tensorboardwill fail if the model has anyNaNorInfparameters during training with the following error:ValueError: The histogram is empty, please file a bug report.NaN/Inf params show up quite often in hyperparam sweep runs if a bad initial value of hyperparam is chosen.
When this happens, the whole sweep can fail.
Some examples just from the past two days:
f156398891
f156398480
f156398264
f156399047
f156399125
This diff:
TensordboardChannelalready catches some exceptions while exporting the modelNaNandInfparameters, to verify no Tensorboard errorsReviewed By: psuzhanhy
Differential Revision: D19048506