Upgrade Tensorflow version to v2.13.0

Signed-off-by: Yuki Iwai <[email protected]>

Author: tenzen-y

cmd/metricscollector/v1beta1/tfevent-metricscollector/requirements.txt

@@ -2,4 +2,5 @@ psutil==5.9.4
 rfc3339>=6.2
 grpcio>=1.41.1
 googleapis-common-protos==1.6.0
-tensorflow==2.11.0
+tensorflow==2.13.0
+protobuf<=3.20.3

cmd/suggestion/nas/enas/v1beta1/requirements.txt

@@ -1,4 +1,5 @@
 grpcio>=1.41.1
 googleapis-common-protos==1.6.0
 cython>=0.29.24
-tensorflow==2.11.0
+tensorflow==2.13.0
+protobuf<=3.20.3

examples/v1beta1/trial-images/enas-cnn-cifar10/requirements.txt

@@ -1,2 +1,2 @@
 scipy>=1.7.2
-tensorflow==2.11.0
+tensorflow==2.13.0

examples/v1beta1/trial-images/tf-mnist-with-summaries/requirements.txt

@@ -1 +1 @@
-tensorflow==2.11.0
+tensorflow==2.13.0

pkg/metricscollector/v1beta1/tfevent-metricscollector/tfevent_loader.py

@@ -22,10 +22,13 @@
 # https://github.com/kubeflow/katib/blob/master/examples/v1beta1/kubeflow-training-operator/tfjob-mnist-with-summaries.yaml#L16-L22
 
 import tensorflow as tf
-from tensorboard.backend.event_processing.event_accumulator import EventAccumulator
+from tensorboard.backend.event_processing.event_accumulator import EventAccumulator, TensorEvent
+from tensorboard.backend.event_processing.tag_types import TENSORS
+from tensorboard.compat.proto import tensor_pb2
 import os
-from datetime import datetime
 import rfc3339
+from datetime import datetime
+from dataclasses import is_dataclass
 import api_pb2
 from logging import getLogger, StreamHandler, INFO
 from pkg.metricscollector.v1beta1.common import const
@@ -41,27 +44,38 @@ def find_all_files(directory):
             for f in files:
                 yield os.path.join(root, f)
 
+    @staticmethod
+    def new_metric_log(metric_name: str, wall_time: float, tensor: tensor_pb2.TensorProto) -> api_pb2.MetricLog:
+        return api_pb2.MetricLog(
+            time_stamp=rfc3339.rfc3339(datetime.fromtimestamp(wall_time)),
+            metric=api_pb2.Metric(
+                name=metric_name,
+                value=str(tf.make_ndarray(tensor))
+            )
+        )
+
     def parse_summary(self, tfefile):
         metric_logs = []
-        event_accumulator = EventAccumulator(tfefile, size_guidance={'tensors': 0})
+        event_accumulator = EventAccumulator(tfefile, size_guidance={TENSORS: 0})
         event_accumulator.Reload()
-        for tag in event_accumulator.Tags()['tensors']:
+        for tag in event_accumulator.Tags()[TENSORS]:
             for m in self.metric_names:
 
                 tfefile_parent_dir = os.path.dirname(m) if len(m.split("/")) >= 2 else os.path.dirname(tfefile)
                 basedir_name = os.path.dirname(tfefile)
                 if not tag.startswith(m.split("/")[-1]) or not basedir_name.endswith(tfefile_parent_dir):
                     continue
 
-                for wall_time, step, tensor in event_accumulator.Tensors(tag):
-                    ml = api_pb2.MetricLog(
-                        time_stamp=rfc3339.rfc3339(datetime.fromtimestamp(wall_time)),
-                        metric=api_pb2.Metric(
-                            name=m,
-                            value=str(tf.make_ndarray(tensor))
-                        )
-                    )
-                    metric_logs.append(ml)
+                # Since Tensorboard v2.12.0, the 'TensorEvent' typed was changed from namedtuple to dataclass.
+                # REF: https://github.com/tensorflow/tensorboard/commit/1975529d953eff55d279ee036240e4db4cb0d57c
+                if is_dataclass(TensorEvent):
+                    # Tensorboard >= v2.12.0
+                    for tensors in event_accumulator.Tensors(tag):
+                        metric_logs.append(self.new_metric_log(m, tensors.wall_time, tensors.tensor_proto))
+                else:
+                    # Tensorboard < v2.12.0
+                    for wall_time, step, tensor in event_accumulator.Tensors(tag):
+                        metric_logs.append(self.new_metric_log(m, wall_time, tensor))
 
         return metric_logs
 

sdk/python/v1beta1/kubeflow/katib/constants/constants.py

@@ -50,8 +50,8 @@
 
 # Supported base images for the Katib Trials.
 # TODO (andreyvelich): Implement list_base_images function to get each image description.
-BASE_IMAGE_TENSORFLOW = "docker.io/tensorflow/tensorflow:2.11.0"
-BASE_IMAGE_TENSORFLOW_GPU = "docker.io/tensorflow/tensorflow:2.11.0-gpu"
+BASE_IMAGE_TENSORFLOW = "docker.io/tensorflow/tensorflow:2.13.0"
+BASE_IMAGE_TENSORFLOW_GPU = "docker.io/tensorflow/tensorflow:2.13.0-gpu"
 BASE_IMAGE_PYTORCH = "docker.io/pytorch/pytorch:1.12.1-cuda11.3-cudnn8-runtime"
 BASE_IMAGE_MXNET = "docker.io/mxnet/python:1.9.1_native_py3"