Explicit target shape argument in the HCS data module #212
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
edyoshikun
pushed a commit
that referenced
this pull request
Dec 18, 2024
* explicit target shape argument in the HCS data module * update docstring * update test cases
edyoshikun
pushed a commit
that referenced
this pull request
Dec 18, 2024
* explicit target shape argument in the HCS data module * update docstring * update test cases
edyoshikun
pushed a commit
that referenced
this pull request
Dec 19, 2024
* explicit target shape argument in the HCS data module * update docstring * update test cases
edyoshikun
pushed a commit
that referenced
this pull request
Dec 19, 2024
* explicit target shape argument in the HCS data module * update docstring * update test cases
edyoshikun
added a commit
that referenced
this pull request
Dec 21, 2024
* translation: fix validation loss aggregation (#202) * exposing prefetch and persistent worker (#203) * metrics for dynamic, smoothness and docstrings * updated metrics and plots for distance * fixed CI test cases * nexnt loss prototype * fix bug with z_scale_range in hcs datamodule. If the value is an int this does not work. * exclude the negative pair from dataloader and forward pass * adding option using pytorch-metric-learning implementation and modifying previous to match same input args * removing our implementation of NTXentLoss and using pytorch metric * ruff * prototype for phate and umap plot * - proofreading the calculations - removing unecessary calls to ALFI script - simplifying code to re-use functions * methods to rank nearest neighbors in embeddings * example script to plot state change of a single track * test using scaled features * phate embeddings * removing dataframe from the compute_phate adding docstring * adding phate to the prediction writer and moving it as dependency. * changing the phate defaults in the prediction writer. * ruff * fixing bug in phate in predict writer * adding code for measuring the smoothness * cleanup to run on triplet and ntxent * fix plots for smoothnes * nexnt loss prototype * exclude the negative pair from dataloader and forward pass * adding option using pytorch-metric-learning implementation and modifying previous to match same input args * removing our implementation of NTXentLoss and using pytorch metric * ruff * remove blank line diff * remove blank line diff * simplying the engine * explicit target shape argument in the HCS data module * Revert "explicit target shape argument in the HCS data module" This reverts commit 464d4c9. * Explicit target shape argument in the HCS data module (#212) * explicit target shape argument in the HCS data module * update docstring * update test cases * Gradio example (#158) * initial demo * using the predict_step * modifying paths to chkpt and example pngs * updating gradio as the one on Huggingface * adding configurable phate arguments via config * script to recompute phate and overwrite the previous phate data * ruff * solving redundancies * modularizing the smoothness * removing redundant _fit_phate() * ruff --------- Co-authored-by: Ziwen Liu <[email protected]> Co-authored-by: Alishba Imran <[email protected]> Co-authored-by: Ziwen Liu <[email protected]>
edyoshikun
added a commit
that referenced
this pull request
Dec 23, 2024
* translation: fix validation loss aggregation (#202) * exposing prefetch and persistent worker (#203) * metrics for dynamic, smoothness and docstrings * updated metrics and plots for distance * fixed CI test cases * nexnt loss prototype * fix bug with z_scale_range in hcs datamodule. If the value is an int this does not work. * exclude the negative pair from dataloader and forward pass * adding option using pytorch-metric-learning implementation and modifying previous to match same input args * removing our implementation of NTXentLoss and using pytorch metric * ruff * prototype for phate and umap plot * - proofreading the calculations - removing unecessary calls to ALFI script - simplifying code to re-use functions * methods to rank nearest neighbors in embeddings * example script to plot state change of a single track * test using scaled features * phate embeddings * removing dataframe from the compute_phate adding docstring * adding phate to the prediction writer and moving it as dependency. * changing the phate defaults in the prediction writer. * ruff * fixing bug in phate in predict writer * adding code for measuring the smoothness * cleanup to run on triplet and ntxent * fix plots for smoothnes * nexnt loss prototype * exclude the negative pair from dataloader and forward pass * adding option using pytorch-metric-learning implementation and modifying previous to match same input args * removing our implementation of NTXentLoss and using pytorch metric * ruff * remove blank line diff * remove blank line diff * simplying the engine * explicit target shape argument in the HCS data module * Revert "explicit target shape argument in the HCS data module" This reverts commit 464d4c9. * Explicit target shape argument in the HCS data module (#212) * explicit target shape argument in the HCS data module * update docstring * update test cases * Gradio example (#158) * initial demo * using the predict_step * modifying paths to chkpt and example pngs * updating gradio as the one on Huggingface * adding configurable phate arguments via config * script to recompute phate and overwrite the previous phate data * ruff * solving redundancies * modularizing the smoothness * removing redundant _fit_phate() * ruff --------- Co-authored-by: Ziwen Liu <[email protected]> Co-authored-by: Alishba Imran <[email protected]> Co-authored-by: Ziwen Liu <[email protected]>
mattersoflight
pushed a commit
that referenced
this pull request
Dec 23, 2024
* nexnt loss prototype * fix bug with z_scale_range in hcs datamodule. If the value is an int this does not work. * exclude the negative pair from dataloader and forward pass * adding option using pytorch-metric-learning implementation and modifying previous to match same input args * removing our implementation of NTXentLoss and using pytorch metric * ruff * remove blank line diff * remove blank line diff * simplying the engine * PHATE (#210) * translation: fix validation loss aggregation (#202) * exposing prefetch and persistent worker (#203) * metrics for dynamic, smoothness and docstrings * updated metrics and plots for distance * fixed CI test cases * nexnt loss prototype * fix bug with z_scale_range in hcs datamodule. If the value is an int this does not work. * exclude the negative pair from dataloader and forward pass * adding option using pytorch-metric-learning implementation and modifying previous to match same input args * removing our implementation of NTXentLoss and using pytorch metric * ruff * prototype for phate and umap plot * - proofreading the calculations - removing unecessary calls to ALFI script - simplifying code to re-use functions * methods to rank nearest neighbors in embeddings * example script to plot state change of a single track * test using scaled features * phate embeddings * removing dataframe from the compute_phate adding docstring * adding phate to the prediction writer and moving it as dependency. * changing the phate defaults in the prediction writer. * ruff * fixing bug in phate in predict writer * adding code for measuring the smoothness * cleanup to run on triplet and ntxent * fix plots for smoothnes * nexnt loss prototype * exclude the negative pair from dataloader and forward pass * adding option using pytorch-metric-learning implementation and modifying previous to match same input args * removing our implementation of NTXentLoss and using pytorch metric * ruff * remove blank line diff * remove blank line diff * simplying the engine * explicit target shape argument in the HCS data module * Revert "explicit target shape argument in the HCS data module" This reverts commit 464d4c9. * Explicit target shape argument in the HCS data module (#212) * explicit target shape argument in the HCS data module * update docstring * update test cases * Gradio example (#158) * initial demo * using the predict_step * modifying paths to chkpt and example pngs * updating gradio as the one on Huggingface * adding configurable phate arguments via config * script to recompute phate and overwrite the previous phate data * ruff * solving redundancies * modularizing the smoothness * removing redundant _fit_phate() * ruff --------- Co-authored-by: Ziwen Liu <[email protected]> Co-authored-by: Alishba Imran <[email protected]> Co-authored-by: Ziwen Liu <[email protected]> * renaming cross_dissimilairy with pairwaise_distance_matrix --------- Co-authored-by: Ziwen Liu <[email protected]> Co-authored-by: Alishba Imran <[email protected]> Co-authored-by: Ziwen Liu <[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.
Replace the
architectureargument with explicittarge_2dto avoid confusion. Also changed the default toFalsesince 2.5D is no longer used in production.Previously, if
architectureis not specified, it will default to '2.5D' and produce incorrect results for 3D models, where the output would have 1 Z-window worth of extra blank slices relative to the input. This was not a problem before #182, when this argument was hard-linked by the CLI with the model configuration.@tayllatheodoro @Soorya19Pradeep @ieivanov @talonchandler:
This parameter used to be automatically configured (correctly) for
viscy<=0.2.1, which is all the stable versions on PyPI. If you recently did 3D predictions with a pre-release installation, please check the results to make sure that the configuration is correct (look for blank slices at the start of z-stacks). Specifically for @tayllatheodoro I noticed that some of the mantis predictions was affected by this.