First refl1d Code Camp

Refl1d Code Camp

We will be holding a small refl1d code camp at SNS in September 2024.

The refl1d package is firmly established as one of the main community tools for fitting reflectometry data. It is a flexible package that allows us to build complex models for both magnetic and non-magnetic specular reflectometry.

Within the last year, the developers of refl1d have started working on a browser-based application that lets users optimize their models. This application can run both locally on a user’s computer and be hosted as a web application. It offers rich functionality to optimize and visualize existing models. This code camp aims to jump-start that community effort by assembling representatives from different facilities to work together on the application. 

Dates

September 9 to 12 2024

Registration

Information will be available here: https://conference.sns.gov/event/441/

Accommodation

People are staying at various hotels in Oak Ridge

Agenda

September 9, 2024

Time	Event	Lead	Place
8:15am – 8:45am	Event Registration	Abby Forgety	Lobby, 8600
8:45am – 9:00am	Opening Remarks	Mathieu Doucet	C156, 8600
9:00am – 9:30am	Set up computers	Brian Maranville (NIST)	C156, 8600
9:30am – 10:00am	Morning Break – Coffee and Drinks	All	C156, 8600
10:00am – 11:00am	Set up computers (continued)	All	C156, 8600
11:00am – 12:00pm	Refl1D Architecture	Brian Maranville (NIST)	C156, 8600
12:00pm – 12:45pm	Git Tutorial and Lunch	Pete Peterson	C156, 8600
12:45pm – 15:00pm	Go through tasks list and split in teams	Brian, Andrew, and Mat	C156, 8600
15:00pm – 15:20pm	Afternoon Break – Coffee and Drinks	All	C156, 8600
15:20pm – 16:50pm	Housekeeping: how to contribute, PRs, reviews, etc...	Brian Maranville	C156, 8600
16:50-17:00	Closing remarks	Mathieu Doucet	C156, 8600

September 10, 2024

Time	Event	Lead	Place
9:00am – 9:10am	Opening Remarks	Mathieu Doucet	C156, 8600
9:10am – 10:30am	Architecture discussion – others may work on tasks	All	C156, 8600
10:30am – 11:00am	Morning Break: - Group photo in the lobby of Building 8600 - Coffee and Drinks	All	SNS Lobby
11:00am – 12:00pm	Architecture discussion – part II	All	C156, 8600
12:00pm – 13:00pm	Lunch and Brian’s Talk “Recent refl1d development”	Brian Maranville (NIST)	C156, 8600
13:00pm – 13:30pm	STFC update: Epic and Dev team structure	Andrew Caruana (STFC)	C156, 8600
15:00pm – 15:20pm	Afternoon Break – Coffee and Drinks	All	C156, 8600
15:20pm – 16:50pm	Nested sampler discussion – others may work on tasks	All	C156, 8600
16:50-17:00	Closing remarks	Mathieu Doucet	C156, 8600

September 11, 2024

Time	Event	Lead	Place
8:45am – 9:00am	Opening Remarks	Mathieu Doucet	C156, 8600
9:00am – 10:00am	Molgroups modeling package	David Hoogerheide (NIST)	C156, 8600
10:00am – 10:30am	Morning Break – Coffee and Drinks	All	C156, 8600
10:30am – 12:00pm	Curated examples for LLMs – others may work on tasks	All	C156, 8600
12:00pm – 12:45pm	Lunch & LLM presentation	Changwoo Do	C156, 8600
12:45pm – 15:00pm	Architecture – part III	All	C156, 8600
15:00pm – 15:20pm	Afternoon Break – Coffee and Drinks	All	C156, 8600
15:20pm – 16:50pm	Architecture – part IV	All	C156, 8600
16:50-17:00	Closing remarks	Mathieu Doucet	C156, 8600

September 12, 2024

Time	Event	Lead	Place
8:45am – 9:00am	Opening Remarks	Mathieu Doucet	C156, 8600
9:00am – 10:00am	Review PRs – finish work	All	C156, 8600
10:00am – 10:30am	Morning Break – Coffee and Drinks	All	C156, 8600
10:30am – 12:00pm	Review PRs – finish work	All	C156, 8600
12:00pm – 12:45pm	Lunch & SasView development approach	Jeff (NCNR)	C156, 8600
12:45pm – 15:00pm	Wrap-up: update on what was achieved	All	C156, 8600
15:00pm – 15:20pm	Afternoon Break – Coffee and Drinks	All	C156, 8600
15:20pm – 16:50pm	Wrap-up: planning for next steps	All	C156, 8600
16:50-17:00	Closing remarks	Mathieu Doucet	C156, 8600

Work packages

(tentative)

• Release bumps dataclass branch as main
  • manually merge recent changes to master branch into webview (Kienzle)
  • publish to pypi
  • document changes
  • check conda-forge status
• Release webview branch of refl1d as main
  • merge webview and master (Kienzle)
  • (publish to pypi) (Kienzle + Maranville)
  • document changes
  • register new conda-forge feedstock for refl1d
• implement code signing workflow for MacOS (executables)
• Defining an API for post-fit processing of parameter population statistics (generating graphs etc.)
• First pass at implementing a model-selection framework (nested sampler?) (Caruana)
• capturing fit output programmatically (Krzywon)
• Adding features to the model-builder
• Implement half-polarized probe, that sums e.g. ++ and +- theory xs to compare to + summed data (Caruana) [Assigned to Mat]
• document:
  • oversampling: what it does, how to choose the right sampling density
  • different interpretations of roughness at an interface for Nevot-Croce (blending) vs. microsteps (rotation) in magnetic samples
• Document/refactor existing layer types that are not commonly known/used.
• Refactor "export" (Maranville)
• merge cuda library backend (Maranville)
• improvements to file access
  • remember working directory between starts
• Create a new repo for curated examples to be used as a started point or for LLM training (Doucet)
• implement "resume fit" for DREAM (Dave)
• formalize "custom plot" for Fitness/Experiment classes (custom plot panels in GUI - Dave)

Organizers

• Brian Maranville [NIST]
• Andrew Caruana [ISIS]
• Mat Doucet [ORNL - local organizer]