Tropical loss enhance pftmask cnpratio badvars - #61
Merged
daliwang merged 13 commits intoFeb 18, 2026
Merged
Conversation
… CNP model to use absolute values for non-negativity enforcement.
…guration - Introduced command-line arguments for PFT1D activation functions and sparsity weights. - Updated model configuration to handle activation overrides and sparsity weights. - Enhanced CNP model output processing to apply specified activation functions. - Adjusted trainer logic to incorporate per-variable sparsity weights for loss calculation.
- Introduced new command-line arguments for tail-aware loss types, weights, and parameters in `train_cnp_model.py`. - Updated `TrainingConfig` to include additional settings for tail-aware variables and loss functions. - Enhanced the model trainer to support new tail-aware loss calculations, including log1p Huber and quantile losses. - Added a new `commands.txt` file with example commands for training and validation processes. - Updated documentation to reflect changes in restart file paths and usage.
- Introduced `CNP_IO_updated9_cnpratio_reduced.txt` for dataset paths and variable definitions. - Updated `commands.txt` with additional training commands for the new dataset configuration. - Enhanced `train_cnp_model.py` to support tail-aware configurations and save data configurations for inference. - Updated `training_config_unified.json` with new tail-aware settings and variable weights. - Added documentation on CNP stoichiometric relationships and performance comparisons between reduced and complete variable lists. - Introduced new scripts for validation and performance improvement suggestions.
…tings - Introduced `CNP_model_config_tokenization_default.txt` for default CNP model parameters. - Added `large_model_tokenization_config_v02.txt` for a larger, tokenization-focused model configuration. - Included detailed comments on model architecture and usage instructions in both configuration files. - Created documentation for tropical NPOOL/PPOOL observations and suggestions to improve model performance.
- Add CNP_RATIO_ENFORCEMENT_USAGE.md, CNP_DERIVATION_*.md, DERIVATION_RESULTS_INTERPRETATION.md - Update CNP_pipeline_runbook.md: step 5 inference with --derive-np-from-c, new section linking all CNP stoichiometry/derivation docs Co-authored-by: Cursor <cursoragent@cursor.com>
- Added new command-line arguments `--split-seed` and `--train-split` to `train_cnp_model.py` for customizable data splitting. - Updated training logic to utilize the new arguments, allowing for flexible train/validation splits. - Introduced `plot_training_loss.py` script to visualize training and validation loss from CSV files. - Modified `commands.txt` to include additional training commands and updated the runbook for improved clarity on validation processes. - Ensured consistency in loss calculations during training and validation by applying the PFT presence mask where applicable.
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
1 participant
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.
bring up the unified configuration and model tuning option to improve the tropical region prediction