shunping commented on code in PR #39164:
URL: https://github.com/apache/beam/pull/39164#discussion_r3500803931


##########
sdks/python/apache_beam/testing/benchmarks/cloudml/requirements.txt:
##########
@@ -15,15 +15,12 @@
 #    limitations under the License.
 #
 
-# Core TFT dependencies with version bounds.
-# Note: To avoid pip ResolutionTooDeep errors, always install using the 
constraints file:
-# pip install -c constraints.txt -r requirements.txt
-dill>=0.3,<0.5
-tfx_bsl>=1.15,<1.17
-tensorflow-transform>=1.15,<1.17
-tensorflow>=2.15,<2.16
-numpy>=1.22.0,<2.0
-tensorflow-metadata>=1.15,<1.16
-pyarrow>=10,<11
-tensorflow-serving-api>=2.15,<2.16
-tf-keras>=2.15,<2.16
+dill==0.4.1
+tfx_bsl==1.21.0
+tensorflow-transform==1.21.0
+tensorflow-metadata==1.21.0
+tensorflow
+numpy
+pyarrow
+tensorflow-serving-api
+tf-keras

Review Comment:
   Pinning these package versions would require us to manually update them 
whenever we upgrade the TFX stack, which introduces a lot of maintenance 
overhead. Instead, we rely on uv or pip to resolve and install the compatible 
versions dynamically.
   
   While this approach could occasionally impact test stability, 
troubleshooting is straightforward. If a failure occurs, we can easily compare 
the installed dependencies of the last successful run against the first failed 
run to pinpoint the changes.



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to