tf-resnet-profiling-multi-gpu-multi-node.ipynb failed CI

Link to the notebook:
https://github.com/aws/amazon-sagemaker-examples/blob/master/sagemaker-debugger/tensorflow_profiling/tf-resnet-profiling-multi-gpu-multi-node.ipynb

Error:

---------------------------------------------------------------------------
Exception encountered at "In [10]":
---------------------------------------------------------------------------
ModuleNotFoundError Traceback (most recent call last)
in
----> 1 from smdebug.profiler.analysis.notebook_utils.training_job import TrainingJob
2 tj = TrainingJob(training_job_name, region)
3 tj.wait_for_sys_profiling_data_to_be_available()

/usr/local/lib/python3.7/site-packages/smdebug/profiler/analysis/notebook_utils/__init__.py in
1 # Local
----> 2 from .metrics_histogram import MetricsHistogram
3 from .step_histogram import StepHistogram
4 from .training_job import TrainingJob

/usr/local/lib/python3.7/site-packages/smdebug/profiler/analysis/notebook_utils/metrics_histogram.py in
4 # Third Party
5 import numpy as np
----> 6 from bokeh.io import o

[...]

ModuleNotFoundError: No module named 'bokeh'

Open tf-resnet-profiling-multi-gpu-multi-node.ipynb and reproduce the CI failure at In [10], starting with the missing bokeh import in smdebug.profiler.analysis.notebook_utils. Trace how the notebook environment installs dependencies; done means the notebook passes CI and the TrainingJob import succeeds.

Written by the indexing model from the issue text.