Notebook readme#2
Open
RevathyVenukuttan wants to merge 26 commits into
Open
Conversation
An unreleased dependency (from which only a single module, ConfigFile, is used here) is included using git+ syntax.
This initial version is exported from the environment that we know works. Many of the version specs can probably be relaxed to at least allow minor version upgrades.
Also factors the non TF requirements into their own reqs file so both the Dockerfile and a standard install can use a reqs file.
The image is pushed to ghcr.io as the registry. Closes #6.
Also includes a few other minor changes and fixes. Co-authored-by: revathy95 <66286555+RevathyVenukuttan@users.noreply.github.com>
This is to mirror the architecture of DeepSTARR, which uses a Flatten step with subsequent Dense layers, but does not put a dropout layer between these.
This is mainly useful for testing a given configuration file for its contents and syntax.
Adds making dropout after flatten optional
Adds library size normalization for custom loss If the input count file has library sizes for DNA and RNA replicates, activating the custom loss function in the configuration will now use them for correcting the RNA/DNA count ratios accordingly. Note that library sizes won't be used if not present as additional columns, and not unless custom loss (`UseCustomLoss`) is activated. Code ported over from https://github.com/bmajoros/BlueSTARR/blob/main/BlueSTARR-NLL.py --------- Co-authored-by: Bill Majoros <bmajoros@duke.edu>
Corrects library size ratio to using average of DNA library sizes rather than sum. (This corrects the shift in predicted values to the right.) Also simplifies the conditional for using library size correction for theta or not.
Changes from SinePositionEncoding to RotaryEncoding, and changes the transformer layer to use TransformerEncoder from keras_nlp, rather than MultiHeadAttention. Also upgrades the recommended Python version to 3.11 (from 3.10).
|
See Duke-IGVF/BlueSTARR#25 where this was "moved" and then merged. |
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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.
Added the Jupyter notebook for generating counts.txt.gz file and updated the README with link to the notebook for data pre-processing.