viscy -> VisCy #34
Merged
viscy -> VisCy #34
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Introducing capitalization to highlight vision and single-cell aspects of the pipeline.
ziw-liu
approved these changes
Aug 18, 2023
mattersoflight
added a commit
that referenced
this pull request
Aug 22, 2023
ziw-liu
added a commit
that referenced
this pull request
Aug 23, 2023
* updated intro and paths * updated figures, tested data loader * setup.sh fetches correct dataset * finalized the exercise outline * semi-final exercise * parts 1 and 2 tested, part 3 outline ready * clearer variables, train with larger patch size * fix typo * clarify variable names * trying to log graph * match example size with training * reuse globals * fix reference * log sample images from the first batch * wider model * low LR solution * fix path * seed everything * fix test dataset without masks * metrics solution this needs a new test dataset * fetch test data, compute metrics * byass cellpose import error due to numpy version conflicts * final exercise * moved files * fixed formatting - ready for review * viscy -> VisCy (#34) (#39) Introducing capitalization to highlight vision and single-cell aspects of the pipeline. * trying to log graph * log graph * black --------- Co-authored-by: Shalin Mehta <[email protected]> Co-authored-by: Shalin Mehta <[email protected]>
mattersoflight
added a commit
that referenced
this pull request
Aug 30, 2023
* pixelshuffle decoder * Allow sampling multiple patches from the same stack (#35) * sample multiple patches from one stack * do not use type annotations from future it breaks jsonargparse * fix channel stacking for non-training samples * remove batch size from model the metrics will be automatically reduced by lightning * add flop counting script * 3d ouput head * add datamodule target dims mode * remove unused argument and configure drop path * move architecture argument to model level * DLMBL 2023 excercise (#36) * updated intro and paths * updated figures, tested data loader * setup.sh fetches correct dataset * finalized the exercise outline * semi-final exercise * parts 1 and 2 tested, part 3 outline ready * clearer variables, train with larger patch size * fix typo * clarify variable names * trying to log graph * match example size with training * reuse globals * fix reference * log sample images from the first batch * wider model * low LR solution * fix path * seed everything * fix test dataset without masks * metrics solution this needs a new test dataset * fetch test data, compute metrics * byass cellpose import error due to numpy version conflicts * final exercise * moved files * fixed formatting - ready for review * viscy -> VisCy (#34) (#39) Introducing capitalization to highlight vision and single-cell aspects of the pipeline. * trying to log graph * log graph * black --------- Co-authored-by: Shalin Mehta <[email protected]> Co-authored-by: Shalin Mehta <[email protected]> * fix channel dimension size for example input #40 * fix argument linking * 3D prediction writer sliding windows are blended with uniform average * update network diagram * upgrade flop counting * shallow 3D (2.5D) SSIM metric * ms-ssim * mixed loss * fix arguments * fix inheritance * fix weight checking * squeeze metric * aggregate metrics * optinal clamp to stabilize gradient of MS-SSIM * fix calling * increase epsilon * disable autocast for loss * restore relu for clamping * plot all architectures with network_diagram script --------- Co-authored-by: Shalin Mehta <[email protected]>
ziw-liu
added a commit
that referenced
this pull request
Nov 1, 2023
* pixelshuffle decoder * Allow sampling multiple patches from the same stack (#35) * sample multiple patches from one stack * do not use type annotations from future it breaks jsonargparse * fix channel stacking for non-training samples * remove batch size from model the metrics will be automatically reduced by lightning * add flop counting script * 3d ouput head * add datamodule target dims mode * remove unused argument and configure drop path * move architecture argument to model level * DLMBL 2023 excercise (#36) * updated intro and paths * updated figures, tested data loader * setup.sh fetches correct dataset * finalized the exercise outline * semi-final exercise * parts 1 and 2 tested, part 3 outline ready * clearer variables, train with larger patch size * fix typo * clarify variable names * trying to log graph * match example size with training * reuse globals * fix reference * log sample images from the first batch * wider model * low LR solution * fix path * seed everything * fix test dataset without masks * metrics solution this needs a new test dataset * fetch test data, compute metrics * byass cellpose import error due to numpy version conflicts * final exercise * moved files * fixed formatting - ready for review * viscy -> VisCy (#34) (#39) Introducing capitalization to highlight vision and single-cell aspects of the pipeline. * trying to log graph * log graph * black --------- Co-authored-by: Shalin Mehta <[email protected]> Co-authored-by: Shalin Mehta <[email protected]> * fix channel dimension size for example input #40 * fix argument linking * 3D prediction writer sliding windows are blended with uniform average * update network diagram * upgrade flop counting * shallow 3D (2.5D) SSIM metric * ms-ssim * mixed loss * fix arguments * fix inheritance * fix weight checking * squeeze metric * aggregate metrics * optinal clamp to stabilize gradient of MS-SSIM * fix calling * increase epsilon * disable autocast for loss * shuffle validation data for logging this hurts cache hit rate, but can avoid logging neighboring windows * simplify decoder structure * pop-head * fix head expansion * init conv weights * update diagnostic scripts * fix center slice metrics for 3D output (#51) * Configure the number of image samples logged at each epoch and batch (#49) * log sample size at epoch and batch levels * update example configs * do not shuffle validation dataset * fix upsampling weight initialization * fix merge * fix merge error * fix formatting --------- Co-authored-by: Shalin Mehta <[email protected]> Co-authored-by: Shalin Mehta <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Introducing capitalization to highlight vision and single-cell aspects of the pipeline.