Enable BatchNorm for Keras Model in DDPG Example

[amifunny]

training=True was added to each model call to enable BatchNorm.

[fchollet]

How did the extra whitespace get added? You can remove it.

[amifunny]

Yeah, Many blank lines are added by black.

[fchollet]

black does not add any such blank lines. Let's get rid of them.

• Start from the original file (on the master branch)
• Add the training=True statements and modify the "last modified" date
• Run black ddpg_pendulum.py
• Commit the file

[fchollet]

You can also just remove the added extra space by hand...

Please add the generated files.

[amifunny]

Sorry for the late response. Deleted extra lines manually.

Generating files with cmd -
python autogen.py add_example rl/ddpg_pendulum

gave the following error -
Traceback (most recent call last): File "../scripts/autogen.py", line 918, in <module> working_dir=get_working_dir(sys.argv[3]) if len(sys.argv) == 4 else None, File "../scripts/autogen.py", line 212, in add_example assert path.count(os.path.sep) == 1 AssertionError
Can you add it on your part?

[fchollet]

Thanks, I will refresh the files then.

[fchollet]

FYI batchnorm breaks training entirely so I have removed it.