Draft
Conversation
…xtraction from Tabpfn
…training - Introduced `data_loader.py` for loading datasets from CSV or generating synthetic data, including preprocessing functions. - Added `evaluation_metrics.py` to provide regression and classification metrics, including RMSE, MAE, accuracy, F1 score, and AUC calculations. - Created `model_training.py` to define baseline models for regression and classification, along with functions for training, predicting, saving, and loading models. Includes cross-validation and evaluation capabilities.
- Modified the `events.out.tfevents` file to reflect recent training changes. - Updated `time_left.tsv` with new iteration data, showing adjusted passed and remaining time for each iteration.
Add local baselining helpers and update training logs
Updating local branch
Here are the files whihc I have been working on. The baselining notebook contains all of my code which was used to perform the experiments, the baselining notebook summary markdown file contains a much simpler output for the findings.
Update baseline utilities and notebook naming
- Rename n_ensemble -> n_estimators in TABPFN_CONFIG (7.x rename) - Remove use_wandb (no longer a TabPFN param) - Switch device from 'cpu' to 'auto' (7.x auto GPU detection) - Fix DATA_PATH -> DATA_DIR import in data_loader_class.py
- Fix hardcoded LOCAL_CSV_PATH -> uses DATA_DIR from baseline_config - Update GLOBAL_MAX_TRAIN comment (no longer API limit, now local model limit) - Replace defunct pre_aux try/except block with clean TabPFNClassifier(n_estimators=8, device='auto')
…tiation - TabPFNClassifier init now conditional: local uses n_estimators/device, client uses random_state only - Full rerun complete on eudirectlapse.csv (23K rows, 10K cap, 80/20 split) - tabpfn_extensions upgraded to fix AutoTabPFNClassifier import
Add scripts/download_datasets.py to fetch 3 additional public insurance
classification datasets programmatically:
- coil2000.csv: COIL 2000 (Dutch insurer), OpenML ID 298, 9,822 rows, 85 features
- ausprivauto0405.csv: Australian vehicle insurance 2004-05, CASdatasets GitHub,
67,856 rows, 6 features, ClaimOcc target (6.8% pos rate)
- freMTPL2freq_binary.csv: French MTPL binarised (50K sample),
50,000 rows, 10 features, ClaimIndicator target (5.0% pos rate)
Add notebooks/baseline_experiments/07_multi_dataset_benchmark.ipynb which runs
TabPFN vs GLM (+ CatBoost, RandomForest, XGBoost) across all 4 insurance datasets
and produces a ROC/PR AUC comparison table and bar chart figure. All models capped
at 10,000 training samples for fair comparison.
…CatBoost experiments
Regressional Testing & analysis
- Implement `run_domain_finetune_stage_a.py` for controlled fine-tuning experiments on insurance datasets. - Create batch scripts for fine-tuning trials: `run_finetune_crossover_batch_3000.sh`, `run_finetune_first_batch.sh`, and `run_finetune_stress_batch_2000.sh`. - Introduce `run_small_finetune_classifier_trial.py` for smoke tests on TabPFN classifier fine-tuning. - Enhance logging and result tracking in fine-tuning scripts.
Initial Fine Tuning Experiment
Adding reproducability
Adding note to clarify use of classfifier
Updating funding request
Contributor
|
Note The number of changes in this pull request is too large for Gemini Code Assist to generate a review. |
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
3 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.
No description provided.