-
Notifications
You must be signed in to change notification settings - Fork 74k
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
Printing lines without logging prefix using tf.Print() #12537
Labels
Comments
yoon-hyoung
added a commit
to yoon-hyoung/tensorflow
that referenced
this issue
Aug 24, 2017
*print prefix message by flag value
The prefix comes from LOG(INFO) inside the logging_ops kernel. The fix needs to be made at the C++ level, though I don't have a good recommendation as to how. |
jzju
added a commit
to jzju/tensorflow
that referenced
this issue
Dec 7, 2017
caisq
added a commit
that referenced
this issue
Dec 14, 2017
copybara-service bot
pushed a commit
that referenced
this issue
Jun 4, 2024
…ax rewriter Imported from GitHub PR openxla/xla#12537 This PR fixes : XLA Issue : openxla/xla#11772 corresponding JAX issue : google/jax#20856 Accuracy fix : This PR adds support for softmax axis other than -1, previously oneDNN softmax was always executed with axis = -1 (last dimension) Accuracy issue is observed from JAX 0.4.22 till last release, with current main branch we don't see accuracy issue as the JAX softmax HLO pattern changed (google/jax#20643) and HLO pattern is not re-written to oneDNN softmax. Hence, this PR also adjusts the oneDNN softmax pattern to recognize new HLO pattern and rewrite into oneDNN softmax custom call. Copybara import of the project: -- b983bd61044bf52838661057f570e9e218803549 by Sachin Muradi <sachin.muradi@intel.com>: Add axis support -- 07f7d14aa1f00676b21e8c1f48719fe7e56bcfd2 by Sachin Muradi <sachin.muradi@intel.com>: address more comments + optional broadcast Merging this change closes #12537 FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#12537 from Intel-tensorflow:sachin/softmax-axis de9d6bd36272962faf76077f594649d229bb727a PiperOrigin-RevId: 640140990
copybara-service bot
pushed a commit
that referenced
this issue
Jun 4, 2024
…ax rewriter Imported from GitHub PR openxla/xla#12537 This PR fixes : XLA Issue : openxla/xla#11772 corresponding JAX issue : google/jax#20856 Accuracy fix : This PR adds support for softmax axis other than -1, previously oneDNN softmax was always executed with axis = -1 (last dimension) Accuracy issue is observed from JAX 0.4.22 till last release, with current main branch we don't see accuracy issue as the JAX softmax HLO pattern changed (google/jax#20643) and HLO pattern is not re-written to oneDNN softmax. Hence, this PR also adjusts the oneDNN softmax pattern to recognize new HLO pattern and rewrite into oneDNN softmax custom call. Copybara import of the project: -- b983bd61044bf52838661057f570e9e218803549 by Sachin Muradi <sachin.muradi@intel.com>: Add axis support -- 07f7d14aa1f00676b21e8c1f48719fe7e56bcfd2 by Sachin Muradi <sachin.muradi@intel.com>: address more comments + optional broadcast Merging this change closes #12537 FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#12537 from Intel-tensorflow:sachin/softmax-axis de9d6bd36272962faf76077f594649d229bb727a PiperOrigin-RevId: 640140990
copybara-service bot
pushed a commit
that referenced
this issue
Jun 4, 2024
…ax rewriter Imported from GitHub PR openxla/xla#12537 This PR fixes : XLA Issue : openxla/xla#11772 corresponding JAX issue : google/jax#20856 Accuracy fix : This PR adds support for softmax axis other than -1, previously oneDNN softmax was always executed with axis = -1 (last dimension) Accuracy issue is observed from JAX 0.4.22 till last release, with current main branch we don't see accuracy issue as the JAX softmax HLO pattern changed (google/jax#20643) and HLO pattern is not re-written to oneDNN softmax. Hence, this PR also adjusts the oneDNN softmax pattern to recognize new HLO pattern and rewrite into oneDNN softmax custom call. Copybara import of the project: -- b983bd61044bf52838661057f570e9e218803549 by Sachin Muradi <sachin.muradi@intel.com>: Add axis support -- 07f7d14aa1f00676b21e8c1f48719fe7e56bcfd2 by Sachin Muradi <sachin.muradi@intel.com>: address more comments + optional broadcast Merging this change closes #12537 Reverts dbf3cd3 FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#12537 from Intel-tensorflow:sachin/softmax-axis de9d6bd36272962faf76077f594649d229bb727a PiperOrigin-RevId: 640140990
copybara-service bot
pushed a commit
that referenced
this issue
Jun 4, 2024
…ax rewriter Imported from GitHub PR openxla/xla#12537 This PR fixes : XLA Issue : openxla/xla#11772 corresponding JAX issue : google/jax#20856 Accuracy fix : This PR adds support for softmax axis other than -1, previously oneDNN softmax was always executed with axis = -1 (last dimension) Accuracy issue is observed from JAX 0.4.22 till last release, with current main branch we don't see accuracy issue as the JAX softmax HLO pattern changed (google/jax#20643) and HLO pattern is not re-written to oneDNN softmax. Hence, this PR also adjusts the oneDNN softmax pattern to recognize new HLO pattern and rewrite into oneDNN softmax custom call. Copybara import of the project: -- b983bd61044bf52838661057f570e9e218803549 by Sachin Muradi <sachin.muradi@intel.com>: Add axis support -- 07f7d14aa1f00676b21e8c1f48719fe7e56bcfd2 by Sachin Muradi <sachin.muradi@intel.com>: address more comments + optional broadcast Merging this change closes #12537 FUTURE_COPYBARA_INTEGRATE_REVIEW=openxla/xla#12537 from Intel-tensorflow:sachin/softmax-axis de9d6bd36272962faf76077f594649d229bb727a PiperOrigin-RevId: 639049433
copybara-service bot
pushed a commit
that referenced
this issue
Jun 4, 2024
…ax rewriter Imported from GitHub PR openxla/xla#12537 This PR fixes : XLA Issue : openxla/xla#11772 corresponding JAX issue : google/jax#20856 Accuracy fix : This PR adds support for softmax axis other than -1, previously oneDNN softmax was always executed with axis = -1 (last dimension) Accuracy issue is observed from JAX 0.4.22 till last release, with current main branch we don't see accuracy issue as the JAX softmax HLO pattern changed (google/jax#20643) and HLO pattern is not re-written to oneDNN softmax. Hence, this PR also adjusts the oneDNN softmax pattern to recognize new HLO pattern and rewrite into oneDNN softmax custom call. Copybara import of the project: -- b983bd61044bf52838661057f570e9e218803549 by Sachin Muradi <sachin.muradi@intel.com>: Add axis support -- 07f7d14aa1f00676b21e8c1f48719fe7e56bcfd2 by Sachin Muradi <sachin.muradi@intel.com>: address more comments + optional broadcast Merging this change closes #12537 PiperOrigin-RevId: 640246750
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Labels
Right now, all messages printed via
tf.Print()
are prefixed by:It would be useful to have a way to print user-friendly output from within the graph. Therefore, suggest adding a parameter for the logging format to
tf.Print()
, or at least adding a flag that disables the logging prefix.The text was updated successfully, but these errors were encountered: