[WIP] Web UI dataset

[VukW]

Adds functionality for creating & managing datasets

Dataset submission

• [x] "add new dataset" button on datasets list
• [ ] "choose benchmark" step
  • [x] PoC (dropdown list)
  • [ ] p1 list benchmarks as cards with detailed info (as on "benchmarks list" page), + highlight chosen one
  • [ ] p3 allow to choose dataprep mlcube instead
• "Fill text fields" step
  • [x] DS name
  • [x] description
    • [ ] p2 check length beforehand? DS submission would fail if description is > 20 characters ?
  • [x] location
    • [ ] p2 check length beforehand?
  • [ ] p3 "Submit as prepared" flag
    • needs to be explained somewhere (like, in a tooltip)
    • Q: if checked, is dataset automatically created as operational?
  • [ ] p1 check how errors are displayed & handled
• "Paths" step
  • [ ] Q: combine with previous step?
  • [x] data path
  • [x] labels path
  • [ ] p2 metadata path (what is it?) (is required if dataset is already prepared)
  • [ ] p2 redesign path picking panel
    • [ ] folders / files difference
    • [ ] one click - chosing folder, double click - go inside?
  • [ ] p2 "go back" button / navigation
• [x] "Verify entered data" step
• [x] dataset submission
• [ ] p1 check how errors are displayed & handled

Submitted dataset displaying (dataset details page)

• [ ] p2 display paths (data path, labels path)

Dataset preparation

• [x] "Prepare" button on dataset detail page
• [x] Preparation run
• [ ] p1 clean log messages from magic bytes
• [ ] p1 distinguish messages to headers + usual lines in the medperf code
• [ ] p1 display header messages properly
• [ ] p1 display log messages without json
• [ ] p2 log lines highlighting?
• [ ] p1 spinner at text headers to underline process is running
• [ ] p1 check how errors are displayed & handled
• [ ] p1 button "back to the dataset" rename
• [ ] link to the report / display report if it exists
• [ ] p1 check how errors are displayed & handled (display exceptions in the log)

Prepared dataset displaying (dataset details page)

• [ ] p2 if report exists, "Allowed automatic report submission" flag
• [ ] p2 if report exists, link / path to the report
• [ ] p1 if dataset is prepared, unlock next button "set operational" (locked if not prepared)

Set operational

• [ ] p0 "Set operational" button on dataset detail page
• [ ] p0 Set operational
• [ ] p0 Disable button if already operational
• [ ] p1 check how errors are displayed & handled

Associate

• [ ] p1 "Associate with the benchmark" button on the dataset detail page (no choice of benchmarks)
• [ ] p3 choice of benchmarks if dataset was created with dataprep mlcube
• [ ] p1 associate
• [ ] p1 check how errors are displayed & handled

Run benchmark

• [ ] p1 "Run benchmark" button on the dataset detail page
• [ ] p1 running benchmark page with logs
• [ ] ??? runs history??
• [ ] ??? displaying result???
• [ ] p1 check how errors are displayed & handled

Submit result

• [ ] p1 "Submit result" button on the dataset detail page
• [ ] p1 Submit result
• [ ] p1 check how errors are displayed & handled

general dataset UI

• [ ] p2 design buttons prepare-operational-... navigation to one line?
• [ ] p1 hide buttons panel if you're not the dataset owner?
• [ ] p3 redesign state displaying (we'd have dev/op floating blocks in the header + set-op button in the footer)

Technical refactoring

• [ ] p1 split routes to different files: dataset/submission.py, dataset/preparation.py

[github-actions[bot]]

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅