make-zarr: turn section images into a PyReconstruct zarr

make-zarr turns a folder of section images (.tif, .tiff, .png, .jpg, .jpeg, .bmp) into a PyReconstruct zarr on khlabgpu1. It uses PyReconstruct's own converter, so the result is the same as Series > Images > Convert to scaled images in PyReconstruct, only much faster and without tying up your computer.

New here? Start with Zarr tools on khlabgpu1: connecting, the job list, copying, and troubleshooting.

1 1. Start it
2 2. While it runs
3 3. When it is done
4 How long it takes

1. Start it

On khlabgpu1, type:

make-zarr

It asks three questions, then shows a summary before anything starts.

The three questions and the Ready to start box.

Step	What to give it	What it checks

Step	What to give it	What it checks
STEP 1 the images	The folder that holds your section images. Press Tab to finish folder names, or paste a path.	How many images, their total size, the first and last file, and anything it will ignore (such as `Thumbs.db`). Answer `y` if it is the right folder.
STEP 2 the series code	The 5-letter series code (see Series Code Assignment), for example `SSVQM`.	The zarr is named after it: `SSVQM-zarr`.
STEP 3 where it goes	Press Enter for your own folder on khlabgpu1's big data drive, `/data/data6/<your username>` (made for you if needed), or type another folder.	That you can write there and that there is room. If a zarr with that name is already there, you must type `REPLACE` before it deletes the old one.

Images on khlabfiles? That is fine: they are read over the network. But keep the destination on khlabgpu1 (/data/data6/...). Writing a zarr onto khlabfiles can take hours; the tool warns you if you pick a folder there.

2. While it runs

The live progress screen. Press q to go back to your jobs; the job keeps running.

Make the zarr. khlabgpu1 builds it on its fast NVMe disk and copies each finished image on to your folder as it goes. Two bars show images made and images copied, with the time left.
Finish copying it to your folder. A last sweep for anything not copied yet. YOUR ZARR IS READY TO USE FROM HERE.
Make the one-file .tar copy. Written beside the zarr; then the copy on the fast disk is deleted.

Close the window whenever you like. Type make-zarr again to see your jobs. If khlabgpu1's fast disk is busy with other people's jobs, yours builds straight in your folder instead: it still works, just slower.

3. When it is done

The Done screen: where the zarr is, how to open it, and how to copy it to your own computer.

Open it in PyReconstruct: Series > Images > Find/change image directory, then pick the zarr.
Copy it to your computer or a drive: press c for step-by-step help, or follow Copying a zarr somewhere else.
In your folder: CODE-zarr, CODE-zarr.tar, CODE-zarr.tar.HOW-TO-COPY.txt and CODE-zarr.RUN-LOG.txt.

How long it takes

It depends on how fast your images can be read and the zarr written, not on khlabgpu1's processors. Real numbers from khlabgpu1:

Series	Zarr ready to use	Everything done (with .tar)

Series	Zarr ready to use	Everything done (with .tar)
284 images, 160 GB, read from khlabfiles	about 50 min	about 1 h 20 min
The same series on a laptop	more than 10 h (estimated)	—

The progress screen shows the time left for your own job. It runs slower when other people are using the same drive at the same time.