AttributeError: 'module' object has no attribute 'keras'

[nezygis]

System information

• What is the top-level directory of the model you are using: object_detection
• Have I written custom code (as opposed to using a stock example script provided in TensorFlow): no
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Ubuntu 16.04
• TensorFlow installed from (source or binary): binary
• TensorFlow version (use command below): 1.13
• Bazel version (if compiling from source): N/A
• CUDA/cuDNN version:
• GPU model and memory:
• Exact command to reproduce:
  gcloud ml-engine jobs submit training whoamiobject_detectiondate +%s --job-dir=${YOUR_GCS_BUCKET}/train --packages dist/object_detection-0.1.tar.gz,slim/dist/slim-0.1.tar.gz --module-name object_detection.train --region europe-west1

Describe the problem

I tried to run object detection workflow on gcloud ml-engine by following this guide: https://cloud.google.com/blog/products/gcp/training-an-object-detector-using-cloud-machine-learning-engine

Initially the job would fail with no module named object_detection.train so I followed advice I found on stackoverflow and copied models\research\object_detection\legacy\train.py to models\research\object_detection and reran python setup.py sdist. After that I tried to run the job again with the command I listed above and it failed with AttributeError: 'module' object has no attribute 'keras'

Source code / logs

Traceback (most recent call last): File "/usr/lib/python2.7/runpy.py", line 162, in _run_module_as_main "__main__", fname, loader, pkg_name) File "/usr/lib/python2.7/runpy.py", line 72, in _run_code exec code in run_globals File "/root/.local/lib/python2.7/site-packages/object_detection/train.py", line 51, in <module> from object_detection.builders import model_builder File "/root/.local/lib/python2.7/site-packages/object_detection/builders/model_builder.py", line 22, in <module> from object_detection.builders import box_predictor_builder File "/root/.local/lib/python2.7/site-packages/object_detection/builders/box_predictor_builder.py", line 20, in <module> from object_detection.predictors import convolutional_box_predictor File "/root/.local/lib/python2.7/site-packages/object_detection/predictors/convolutional_box_predictor.py", line 19, in <module> from object_detection.core import box_predictor File "/root/.local/lib/python2.7/site-packages/object_detection/core/box_predictor.py", line 137, in <module> class KerasBoxPredictor(tf.keras.Model): AttributeError: 'module' object has no attribute 'keras'

[luboren]

I am having the same problem, did you figure out how to fix it?

[pakashu]

I have the same problem

[pakashu]

System information

• What is the top-level directory of the model you are using: object_detection
• Have I written custom code (as opposed to using a stock example script provided in TensorFlow): no
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Ubuntu 16.04
• TensorFlow installed from (source or binary): binary
• TensorFlow version (use command below): 1.13
• Bazel version (if compiling from source): N/A
• CUDA/cuDNN version:
• GPU model and memory:
• Exact command to reproduce:
  gcloud ml-engine jobs submit training whoamiobject_detectiondate +%s --job-dir=${YOUR_GCS_BUCKET}/train --packages dist/object_detection-0.1.tar.gz,slim/dist/slim-0.1.tar.gz --module-name object_detection.train --region europe-west1

Describe the problem

I tried to run object detection workflow on gcloud ml-engine by following this guide: https://cloud.google.com/blog/products/gcp/training-an-object-detector-using-cloud-machine-learning-engine

Initially the job would fail with no module named object_detection.train so I followed advice I found on stackoverflow and copied models\research\object_detection\legacy\train.py to models\research\object_detection and reran python setup.py sdist. After that I tried to run the job again with the command I listed above and it failed with AttributeError: 'module' object has no attribute 'keras'

Source code / logs

Traceback (most recent call last): File "/usr/lib/python2.7/runpy.py", line 162, in _run_module_as_main "__main__", fname, loader, pkg_name) File "/usr/lib/python2.7/runpy.py", line 72, in _run_code exec code in run_globals File "/root/.local/lib/python2.7/site-packages/object_detection/train.py", line 51, in <module> from object_detection.builders import model_builder File "/root/.local/lib/python2.7/site-packages/object_detection/builders/model_builder.py", line 22, in <module> from object_detection.builders import box_predictor_builder File "/root/.local/lib/python2.7/site-packages/object_detection/builders/box_predictor_builder.py", line 20, in <module> from object_detection.predictors import convolutional_box_predictor File "/root/.local/lib/python2.7/site-packages/object_detection/predictors/convolutional_box_predictor.py", line 19, in <module> from object_detection.core import box_predictor File "/root/.local/lib/python2.7/site-packages/object_detection/core/box_predictor.py", line 137, in <module> class KerasBoxPredictor(tf.keras.Model): AttributeError: 'module' object has no attribute 'keras'

did you find the solution?

[nezygis]

I haven't and since then I started using darknet with yolov3 for my object detection purposes. It was way easier for a beginner like me

[TZVIh]

Hey @nezygis
did you solve this problem?
another question, how did you train darknet in google cloud? did you build VM?
Thanks!

[nezygis]

Hey @nezygis
did you solve this problem?
another question, how did you train darknet in google cloud? did you build VM?
Thanks!

I did manage to run a different model later from a VM I build on google cloud. There might have been a problem with my local configuration/tensorflow version that was causing the issue because when I restarted the process on a different machine it worked. So just try to redo everything using different TF versions etc.

[TZVIh]

Thanks for your response!
why my local configuration of TF will affect the ml engine?
I am sorry for my misunderstanding.

[nezygis]

Look at the guide here: https://cloud.google.com/blog/products/gcp/training-an-object-detector-using-cloud-machine-learning-engine

this step: Running training and evaluation jobs

We are packaging our local TF. I'm assuming its being used.

I have not spent a lot of time using tensorflow so take my answers with a grain of salt. As I've mentioned previously for my task Darknet was sufficient. I do recommend to try re configuring with a clean environment either way.