Single-cell representation learning #153
Merged
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
* notes on standard report * Add code for generating figures --------- Co-authored-by: Alishba Imran <[email protected]>
…y and features learned by embeddings (#140) * notes on standard report * add lib of computed features * correlates PCA with computed features * compute for all timepoints * compute correlation * remove cv library usage * remove edge detection * convert to dataframe * for entire well * add std_dev feature * fix patch size --------- Co-authored-by: Soorya Pradeep <[email protected]>
* remove obsolete training and prediction scripts * lint contrastive scripts
* draft projection head per Update the projection head (normalization and size). #139 * reorganize comments in example fit config * configurable stem stride and projection dimensions * update type hint and docstring for ContrastiveEncoder * clarify embedding_dim * use the forward method directly for projected * normalize projections only when fitting the projected features saved during prediction is now *not* normalized * remove unused logger * refactor training code into translation and representation modules * extract image logging functions * use AdamW instead of Adam for contrastive learning * inline single-use argument * fix normalization * fix MLP layer order * fix output dimensions * remove L2 normalization before computing loss * compute rank of features and projections * documentation --------- Co-authored-by: Shalin Mehta <[email protected]>
ziw-liu
added
enhancement
* docstring * move scripts from contrastive_scripts to viscy/scripts * organize files in applications/contrastive_phenotyping * delete unused evaluation code * more cleanup * refactor evaluation metrics for translation task * refactor viscy.evaluation -> viscy.translation.evaluation_metrics and viscy.representation.evaluation * WIP: representation evaluation module * WIP: representation eval - docstrings in numpy format * WIP: more documentation * refactor: feature_extractor moved to viscy.representation.evaluation * lint * bug fix * refactored common computations and dataset * add imbalance-learn dependecy to metrics * refactor classification of embeddings * organize viscy.representation.evaluation * ruff * Soorya's plotting script * WIP: combine two versions of plot_embeddings.py * simplify representation.viscy.evaluation - move LCA to its own module * refactor of viscy.representation.evaluation * refactored and tested PCA and UMAP plots --------- Co-authored-by: Soorya Pradeep <[email protected]>
ziw-liu
changed the title
…et contrastive task (#154) * wip: sample positive and negative samples from another time point * configure time interval in triplet data module * vectorized anchor filtering * conditional augmentation for anchor anchor is augmented if the positive is another time point * example training script for the CTC dataset this is optimized to run on MPS * add example CTC prediction config for MPS
3 tasks
ziw-liu
commented
Sep 11, 2024
* refactor linear probing with lightning * test convenience function * always convert to long before onehot * use onehot only during training * supply trainer through argument to avoid wrapping * only log per epoch * example script for linear probing * add comment about loss curve * fix sample filtering order for select tracks * add script to visualize integrated gradients * plot integrated gradients over time * Use sklearn's logistic regression for linear probing (#169) * use binary logistic regression to initialize the linear layer * plot integrated gradients from a binary classifier * add cmap to 'visual' requirements * move model assembling to lca * rename init argument * disable feature scaling * update test and evaluation scripts to use new API * add docstrings to LCA
* add maplotlib style sheet for figure making * add cell division attribution * add matplotlib style sheet * move attribution computation to lca * tweak contrast limits and text * add captum to optional dependencies * move attribution function to a method of the classifier * add script to show organelle dynamics * add occlusion attribution * more generic save path * add uninfected cell * tweak subplot spacing
* updated files * format fixed for tests * updated scripts * umap dist code * bug fixes and linting * logistic regression script * add infection figure script * Add script for generating infection figure and perform prediction on the June dataset * Format code * Black format evaluation module and fix import in figure_cell_infection script * Refactor scatterplot colors and markers * Calculate model accuracy * Add script for appendix video * formatted code * updated displacement funcs for full embeddings * script for displacement computation * fix style * fix docstring format --------- Co-authored-by: Shalin Mehta <[email protected]> Co-authored-by: Soorya Pradeep <[email protected]> Co-authored-by: Ziwen Liu <[email protected]>
|
To merge this branch and release candidate 0.3.0-rc1, we need to test the following:
We decided that the configs and checkpoints posted for the preprint will continue to depend on release 0.2.0. |
* fix docstrings and type hint for the ContrastiveEncoder * refactor the representation evaluation module into submodules * move shared image logging into utils * fix line end * fix import paths in example notebooks
|
Things I have tested with the current HEAD of this branch:
@mattersoflight I'm still working on #181 which will also introduce user interface changes. Should we do comprehensive release candidate testing after that? Things need to be tested before release: @edyoshikun or @mattersoflight:
|
@ziw-liu I suggest merging #181 (CLI interface) in this branch and then doing the tests you outlined so that everyone builds familiarity with the revised CLI. Since these two PRs make multiple breaking (and welcome) changes to the codebase, I suggest tagging the current head of |
|
@mattersoflight Compared to the latest stable release (v0.2.1), the current HEAD of |
|
I have done one round of testing. The training is underway and the prediction using an earlier model checkpoint was completed. |
* remove obsolete metrics script for translation * move cellpose annotation script * consolidate CLI documentation * remove old CLI help * move translation CLI to its own module * move contrastive CLI to its own module * remove old CLI module * remove global entry script * share trainer class between tasks * move cli from init to main * inherit base CLI class for tasks * improve type hint and docstring * restore global CLI entry point * special case subclass mode for preprocessing * remove separate entry points * add CLI description message * make the setup function private * fix subclass mode detection * remove unused arguments from custom subcommands * use generic path in example * fix docstring style * update virtual staining example configs * update CTC SSL example configs * update infection SSL example configs
ziw-liu
commented
Oct 16, 2024
This was
linked to
issues
Oct 16, 2024
|
As discussed, the HF model is pinned to use <0.3 versions and the gradio code is not exposed to the user, so we don't need to update this for now. The model weights are posted separately and point to the github. |
|
@ziwen, I'm done testing the virtual staining end-to-end. I didn't run any of the representation learning. I really appreciate the new config files structure and CLI. This will work well with any type of custom dataloaders and models. Thank you |
* extract function for computing umap * specific return type for predict step * write umap in prediction * raise log level for umap computation * fix key conversion
* draft readme * direct link dynaCLR schematic * add DynaCLR schemetic figure * add static schematic and link to video --------- Co-authored-by: Ziwen Liu <[email protected]> Co-authored-by: Ziwen Liu <[email protected]>
8 tasks
ziw-liu
added a commit
that referenced
this pull request
Oct 21, 2024
* Merging code related to figures (#146) * notes on standard report * Add code for generating figures --------- Co-authored-by: Alishba Imran <[email protected]> * produce a report of useful visualizations to assess the dimensionality and features learned by embeddings (#140) * notes on standard report * add lib of computed features * correlates PCA with computed features * compute for all timepoints * compute correlation * remove cv library usage * remove edge detection * convert to dataframe * for entire well * add std_dev feature * fix patch size --------- Co-authored-by: Soorya Pradeep <[email protected]> * Remove obsolete scripts for contrastive phenotyping (#150) * remove obsolete training and prediction scripts * lint contrastive scripts * SSL: fix MLP head and remove L2 normalization (#145) * draft projection head per Update the projection head (normalization and size). #139 * reorganize comments in example fit config * configurable stem stride and projection dimensions * update type hint and docstring for ContrastiveEncoder * clarify embedding_dim * use the forward method directly for projected * normalize projections only when fitting the projected features saved during prediction is now *not* normalized * remove unused logger * refactor training code into translation and representation modules * extract image logging functions * use AdamW instead of Adam for contrastive learning * inline single-use argument * fix normalization * fix MLP layer order * fix output dimensions * remove L2 normalization before computing loss * compute rank of features and projections * documentation --------- Co-authored-by: Shalin Mehta <[email protected]> * created and updated classify_feb_embeddings.py * Module and scripts for evaluating representations (#156) * docstring * move scripts from contrastive_scripts to viscy/scripts * organize files in applications/contrastive_phenotyping * delete unused evaluation code * more cleanup * refactor evaluation metrics for translation task * refactor viscy.evaluation -> viscy.translation.evaluation_metrics and viscy.representation.evaluation * WIP: representation evaluation module * WIP: representation eval - docstrings in numpy format * WIP: more documentation * refactor: feature_extractor moved to viscy.representation.evaluation * lint * bug fix * refactored common computations and dataset * add imbalance-learn dependecy to metrics * refactor classification of embeddings * organize viscy.representation.evaluation * ruff * Soorya's plotting script * WIP: combine two versions of plot_embeddings.py * simplify representation.viscy.evaluation - move LCA to its own module * refactor of viscy.representation.evaluation * refactored and tested PCA and UMAP plots --------- Co-authored-by: Soorya Pradeep <[email protected]> * delete duplicate file * lint * fix import paths * rename translation tests * rename translation metrics * Sample positive and negative samples with a time offset for the triplet contrastive task (#154) * wip: sample positive and negative samples from another time point * configure time interval in triplet data module * vectorized anchor filtering * conditional augmentation for anchor anchor is augmented if the positive is another time point * example training script for the CTC dataset this is optimized to run on MPS * add example CTC prediction config for MPS * add fig for mitosis * add script to save image patches * add save patches as npy * save figure at 300dpi * Linear probing (#160) * refactor linear probing with lightning * test convenience function * always convert to long before onehot * use onehot only during training * supply trainer through argument to avoid wrapping * only log per epoch * example script for linear probing * add comment about loss curve * fix sample filtering order for select tracks * add script to visualize integrated gradients * plot integrated gradients over time * Use sklearn's logistic regression for linear probing (#169) * use binary logistic regression to initialize the linear layer * plot integrated gradients from a binary classifier * add cmap to 'visual' requirements * move model assembling to lca * rename init argument * disable feature scaling * update test and evaluation scripts to use new API * add docstrings to LCA * Tweak attribution visualization (#170) * add maplotlib style sheet for figure making * add cell division attribution * add matplotlib style sheet * move attribution computation to lca * tweak contrast limits and text * add captum to optional dependencies * move attribution function to a method of the classifier * add script to show organelle dynamics * add occlusion attribution * more generic save path * add uninfected cell * tweak subplot spacing * UMAP line plot to assess temporal smoothness in features space (#176) * add maplotlib style sheet for figure making * add cell division attribution * add matplotlib style sheet * move attribution computation to lca * tweak contrast limits and text * add captum to optional dependencies * move attribution function to a method of the classifier * add script to show organelle dynamics * add occlusion attribution * more generic save path * add uninfected cell * tweak subplot spacing * lower case titles * reduce UMAP components to 2 and add indices * add script to make the bridge gaps figure * fixed import error * formatted with black * reduce to single arrow on plot * remove reduntant script * Fixes on correlation of PCA and UMAP components to computed_feature script (#159) * reduce initial patch size * add radial profiling * add function descriptions * add umap correlation * add def comments * change umap for all data * add script for 1 chan * add p-value analysis * add PCA analysis * remove duplicate script * Refactor and format code * Format code * Removed umap correlation * note for future refactor --------- Co-authored-by: Ziwen Liu <[email protected]> * updated eval module & cosine sim figures (#168) * updated files * format fixed for tests * updated scripts * umap dist code * bug fixes and linting * logistic regression script * add infection figure script * Add script for generating infection figure and perform prediction on the June dataset * Format code * Black format evaluation module and fix import in figure_cell_infection script * Refactor scatterplot colors and markers * Calculate model accuracy * Add script for appendix video * formatted code * updated displacement funcs for full embeddings * script for displacement computation * fix style * fix docstring format --------- Co-authored-by: Shalin Mehta <[email protected]> Co-authored-by: Soorya Pradeep <[email protected]> Co-authored-by: Ziwen Liu <[email protected]> * Fixup representation (#180) * fix docstrings and type hint for the ContrastiveEncoder * refactor the representation evaluation module into submodules * move shared image logging into utils * fix line end * fix import paths in example notebooks * Unified CLI entry point (#182) * remove obsolete metrics script for translation * move cellpose annotation script * consolidate CLI documentation * remove old CLI help * move translation CLI to its own module * move contrastive CLI to its own module * remove old CLI module * remove global entry script * share trainer class between tasks * move cli from init to main * inherit base CLI class for tasks * improve type hint and docstring * restore global CLI entry point * special case subclass mode for preprocessing * remove separate entry points * add CLI description message * make the setup function private * fix subclass mode detection * remove unused arguments from custom subcommands * use generic path in example * fix docstring style * update virtual staining example configs * update CTC SSL example configs * update infection SSL example configs * Remove outdated comment * updating the dlmbl notebooks * updating dependendencies to allow viscy>0.2 in examples * updating phase contrast demo notebook. * updating references to main * Store UMAP embeddings in SSL predictions (#184) * extract function for computing umap * specific return type for predict step * write umap in prediction * raise log level for umap computation * fix key conversion * Add representation section to readme (#186) * draft readme * direct link dynaCLR schematic * add DynaCLR schemetic figure * add static schematic and link to video --------- Co-authored-by: Ziwen Liu <[email protected]> Co-authored-by: Ziwen Liu <[email protected]> * fix link syntax in readme --------- Co-authored-by: Shalin Mehta <[email protected]> Co-authored-by: Alishba Imran <[email protected]> Co-authored-by: Soorya Pradeep <[email protected]> Co-authored-by: Alishba Imran <[email protected]> Co-authored-by: Soorya19Pradeep <[email protected]> Co-authored-by: Eduardo Hirata-Miyasaki <[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.
Accumulated changes for single-cell representation learning.
@edyoshikun this PR include breaking API changes for image translation (#145).
Pending before merging this to
main: