add models and new angle param #19
Open
Conversation
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
Collaborator
dsarrut
commented
Aug 25, 2025
dsarrut
commented
- models : ResNet_v2
- new angle param atan2
Introduced `MultiTask_v3` model for dual-task predictions (acceptance and energy window classification) using a shared ResNet backbone. Added custom loss calculation in `MultiTask_v3_loss` to handle binary and categorical outputs with class balancing support. Refactored existing code for consistency and modularity.
Introduced support for multiple model types (e.g., MultiTask_v3, ResNet_v2, and xgboost) and enhanced predictions with modular functions. Added handling for plane axis initialization and improved geometric projection for detector plane interactions. Refactored functions for better abstraction and future extensibility.
Eliminated unnecessary modules like sys, cm, np, uproot, and ntpath to streamline the script. This improves code readability and reduces potential overhead from unused dependencies.
This update introduces two new CLI options, `--rr` and `--epoch`, allowing users to override the corresponding parameters in the config file. These additions enhance training flexibility and enable quick parameter adjustments directly on the command line.
The function no longer sets a default value for the 'rr' parameter, requiring it to be explicitly provided. This change ensures greater clarity and avoids unintended behavior from hidden defaults.
Replaced hardcoded model configuration with dynamic type selection, supporting multiple architectures (Net_v1, ResNet_v2, MultiTask_v3). Optimized DataLoader setup, adjusted training loop parameters, and introduced early stopping in a new `train_nn_TEST` function for efficient training workflows.
Renamed the `input_layer` variable to `fc1` across all relevant classes and methods for improved naming consistency with other layers. This change enhances code readability and aligns variable naming conventions in the model definitions.
Introduced a new angle parametrization using 'atan2' for better computation clarity. Also corrected the tensor type reference in the training loop to use `fc1.weight.dtype` instead of `input_layer.weight.dtype`.
Introduced configurable angle parameterization (acos/atan2) and updated logic accordingly. Deprecated and removed legacy functions (`arf_from_points_to_image_counts_OLD*`) to streamline the codebase. Refactored function and method names to ensure consistency and clarity in angle computation workflows.
Replaced single quotes with double quotes for consistency. Removed outdated or redundant comments and unnecessary line breaks to improve code readability and maintainability.
The `verbose` parameter was removed from `ReduceLROnPlateau` calls to simplify the scheduler configuration. This change ensures consistent and cleaner logging behavior.
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.