Bump python deps (#406)

This patch bumps all the python deps to more recent versions. Tensorflow is not bumped to the latest version as the last tf_agents release is only compatible with v2.15.0. We need to figure out what to do here, but this should still be an improvement that we can build off of.

The lockfile here was regenerated with Python 3.10, so probably will not be compatible with anything older.

This patch also contains changes related to the pytype bump.

Author: boomanaiden154

.github/workflows/main.yml

@@ -46,7 +46,7 @@ jobs:
         task: [Typecheck, Lint, Yapf, Test]
         include:
           - task: Typecheck
-            cmd: pytype -j auto --overriding-parameter-count-checks .
+            cmd: pytype -j auto .
           - task: Lint
             cmd: pylint --rcfile .pylintrc --recursive yes .
           - task: Yapf

Pipfile

@@ -4,19 +4,20 @@ verify_ssl = true
 name = "pypi"
 
 [packages]
-absl-py = "==1.0.0"
+absl-py = "==2.1.0"
 gin-config = "==0.5.0"
-psutil = "==5.9.0"
-tf-agents = "==0.16.0"
-tensorflow = "==2.12.0"
-dm-reverb = "==0.11.0"
+psutil = "==6.1.0"
+tf-agents = "==0.19.0"
+tensorflow = "==2.15.0"
+dm-reverb = "==0.14.0"
 
 [dev-packages]
 pylint = "==3.3.2"
-pytest = "==7.1.2"
-pytype = "==2022.06.06"
+pytest = "==8.3.4"
+pytype = "==2024.10.11"
 yapf = "==0.43.0"
 
 [ci]
-joblib = "==1.1.1"
-shap = "==0.41.0"
+joblib = "==1.4.2"
+shap = "==0.46.0"
+

Pipfile.lock

@@ -32,6 +32,10 @@
 from compiler_opt.rl import policy_saver
 from compiler_opt.es import blackbox_evaluator  # pylint: disable=unused-import
 
+# Pytype cannot pick up the pyi file for tensorflow.summary. Disable the error
+# here as these errors are false positives.
+# pytype: disable=pyi-error
+
 # If less than 40% of requests succeed, skip the step.
 _SKIP_STEP_SUCCESS_RATIO = 0.4
 

compiler_opt/es/blackbox_learner.py

@@ -947,9 +947,7 @@ def gen_trajectories(
     )
     not_done = result_futures
     succeeded_idx = 0
-    succeeded: List[Optional[Tuple[Tuple[int, ProfilingDictValueType,
-                                         ProfilingDictValueType],
-                                   tf.train.SequenceExample]]] = []
+    succeeded = []
 
     for written_files_idx in range(num_output_files):
       written_per_file = 0

compiler_opt/rl/imitation_learning/.generate_bc_trajectories_lib.py.swp

@@ -21,6 +21,10 @@
 from absl import logging
 import gin
 
+# Pytype cannot pick up the pyi file for tensorflow.summary. Disable the error
+# here as these errors are false positives.
+# pytype: disable=pyi-error
+
 # <Internal> Using XM - flags.  # pylint: disable=unused-import
 from compiler_opt.rl import agent_config
 from compiler_opt.rl import data_reader

compiler_opt/rl/imitation_learning/generate_bc_trajectories_lib.py

@@ -181,9 +181,11 @@ def train(self, dataset_iter, monitor_dict, num_iterations: int):
     """Trains policy with data from dataset_iter for num_iterations steps."""
     self._reset_metrics()
     # context management is implemented in decorator
+    # pytype: disable=attribute-error
     # pylint: disable=not-context-manager
     with tf.summary.record_if(lambda: tf.math.equal(
         self._global_step % self._summary_export_interval, 0)):
+      # pytype: enable=attribute-error
       for _ in range(num_iterations):
         # When the data is not enough to fill in a batch, next(dataset_iter)
         # will throw StopIteration exception, logging a warning message instead

compiler_opt/rl/train_bc.py

@@ -41,3 +41,5 @@ filterwarnings =
 
 	# Issue #255
 	ignore:The line search algorithm did not converge
+
+        ignore:Conversion of an array with ndim > 0 to a scalar is deprecated, and will error in future. Ensure you extract a single element from your array before performing this operation.