lstm_package

[peckhams]

Created a Python package for the LSTM model (within lstm folder) and tested it within NextGen. Also bug fixes, extra docs and improved Jupyter notebooks.

Additions

• ngen_files is a new folder that contains files that must be present in the ngen repo in order to use this new lstm (or lstm_py) repo.

• Added __init__.py, __main__.py, and run_lstm_with_bmi_v2.py in the lstm package folder. The latter two make it possible to run an LSTM test at an OS prompt with the command: % python -m lstm

• Added README2.md with info on how to create a Python package.

• Added 2 text files to the docs folder.

Removals

• Removed src folder in favor of new lstm package folder to allow "import lstm"

• Removed numerous, unused instances of "from torch import nn"

Changes

• Fixed bugs in lstm/bmi_lstm.py. Mainly in functions: create_scaled_input_tensor(), get_static_attributes(), and initialize_forcings(). This fixed a known problem with the output hydrographs, issue #32.

• Made changes to the 2 Jupyter notebooks to improve speed and readability.

Testing

1. Jupyter notebooks ran successfully.
2. Was able to run this new LSTM repo for all HUC01 catchments on a Mac within NextGen.

Screenshots

Notes

• It might be good to rename the repo to "lstm_py", since we'd like to use "lstm" for the Python package, which is a subfolder of the repo, and it may be less confusing if the repo and package don't have the same name.

• Since NextGen does not yet support virtual environments properly (ngen issue #441), the work-around is to use the export command to set the environment variable PYTHONPATH so that the "lstm" package can be found.

Todos

-

Checklist

• [x] PR has an informative and human-readable title
• [ ] Changes are limited to a single goal (no scope creep)
• [ ] Code can be automatically merged (no conflicts)
• [x] Code follows project standards (link if applicable)
• [x] Passes all existing automated tests
• [x] Any change in functionality is tested
• [x] New functions are documented (with a description, list of inputs, and expected output)
• [ ] Placeholder code is flagged / future todos are captured in comments
• [ ] Visually tested in supported browsers and devices (see checklist below :point_down:)
• [ ] Project documentation has been updated (including the "Unreleased" section of the CHANGELOG)
• [x] Reviewers requested with the Reviewers tool :arrow_right:

Testing checklist

Target Environment support

• [ ] Windows
• [x] Linux
• [x] Browser

Accessibility

• [ ] Keyboard friendly
• [ ] Screen reader friendly

Other

• [ ] Is useable without CSS
• [ ] Is useable without JS
• [ ] Flexible from small to large screens
• [ ] No linting errors or warnings
• [ ] JavaScript tests are passing

[madMatchstick]

Also bug fixes, extra docs and improved Jupyter notebooks.

@peckhams Which bugs? Fix #32 & Fix #29 ?

[madMatchstick]

Created a Python package for the LSTM model (within lstm folder) and tested it within NextGen.

@peckhams Which, if any "standalone", scripts did you test?

[madMatchstick]

We will need to edit README.md, specifically Running BMI LSTM, info, links, etc. @peckhams Can you expand more on standalone mode here?

Also, README2.md is more focused on lstm_bmi as a package, perhaps rename?

[peckhams]

Also bug fixes, extra docs and improved Jupyter notebooks.

@peckhams Which bugs? Fix #32 & Fix #29 ?

Yes, #32 is fixed, and #29 was fixed previously and should still be.

[peckhams]

Created a Python package for the LSTM model (within lstm folder) and tested it within NextGen.

@peckhams Which, if any "standalone", scripts did you test?

Do you mean a script that tests the LSTM model in "standalone mode"? If so, when you run LSTM via: % python -m lstm, it runs entirely outside of NextGen and uses "__main__.py" in the lstm package folder. Most of the testing was done running LSTM inside of NextGen, however.

[peckhams]

We will need to edit README.md, specifically Running BMI LSTM, info, links, etc. @peckhams Can you expand more on standalone mode here?

Also, README2.md is more focused on lstm_bmi as a package, perhaps rename?

Yes, we should update README.md, and should probably rename README2.md to something more descriptive. Let's discuss this in our next LSTM call.

[madMatchstick]

Created a Python package for the LSTM model (within lstm folder) and tested it within NextGen.

@peckhams Which, if any "standalone", scripts did you test?

Do you mean a script that tests the LSTM model in "standalone mode"? If so, when you run LSTM via: % python -m lstm, it runs entirely outside of NextGen and uses "main.py" in the lstm package folder. Most of the testing was done running LSTM inside of NextGen, however.

Thanks for clarifying, I am running into some pathing mishaps with this "standalone" approach. I'll continue to work through and post any major roadblocks here. -thx

[madMatchstick]

@peckhams why was input variable atmosphere_water__time_integral_of_precipitation_mass_flux changed back to atmosphere_water__liquid_equivalent_precipitation_rate?