fit_models

fit_models takes the features built from build_features and trains the model to recognize the boundaries between Table 11, 4, and 25. Once it’s done, you’ll have fresh model files ready for the detect_table command.

Arguments

usage: gtranslate fit_models --feature_file FEATURE_FILE --tt_file TT_FILE
                             --out_dir OUT_DIR [--cpus CPUS] [--seed SEED]
                             [--split_data] [-h]

required named arguments

--feature_file

path to TSV file containing features for each genome

--tt_file

File indicating the translation table for each genome.

--out_dir

directory to output files

optional arguments

--cpus

number of CPUs to use

Default: 1

--seed

seed for reproducibility

--split_data

Enable data splitting into training and validation sets.

Files output

• label_encoder.pkl.gz

• [model]_multi_class.pkl.gz

• [prefix].log

• [prefix].warnings.log

Example

Input

gtranslate fit_models --feature_file features_test/gtranslate_training.feature_summary.tsv --tt_file 1000genomes/5000_genomes_r226.tsv --out_dir models

Output

[2026-04-10 19:22:07] INFO: gTranslate v0.0.2 [2026-04-10 19:22:07] INFO: gtranslate fit_models –feature_file features_test/gtranslate_training.feature_summary.tsv –tt_file 1000genomes/5000_genomes_r226.tsv –out_dir models [2026-04-10 19:22:07] INFO: Training models based on training data. [2026-04-10 19:22:07] INFO: We use seed: 427 [2026-04-10 19:22:07] INFO: All feature columns present in input. Skipping feature engineering. [2026-04-10 19:22:07] INFO: Merged dataset shape after target filtering: (5000, 12) [2026-04-10 19:22:07] INFO: First few rows of the preprocessed dataframe: [2026-04-10 19:22:07] INFO: Coding_density_4 Coding_density_11 Density_Diff GC Trp_ratio Trp_magnitude Gly_ratio UGG_density 0 84.29031 85.94244 -1.65213 56.58347 -1.84939 10.08464 -3.39509 0.21315 1 86.46786 88.43902 -1.97116 57.48985 -1.46175 10.24775 -3.11130 0.19213 2 86.37359 88.21455 -1.84096 57.46309 -1.46933 10.24881 -3.12142 0.19164 3 83.01639 91.96291 -8.94652 71.41138 -1.36587 9.99150 -3.37393 0.13424 4 84.64863 86.23985 -1.59122 56.69494 -1.90344 10.05767 -3.46512 0.20978 [2026-04-10 19:22:07] INFO: Class counts: [4988 3 9] [2026-04-10 19:22:07] INFO: Class weights mapping: {‘11’: 0.33413525795241916, ‘25’: 555.5555555555555, ‘4’: 185.1851851851852} [2026-04-10 19:22:07] INFO: Using the full training data [2026-04-10 19:22:07] INFO: — Training KNeighbors — [2026-04-10 19:22:07] INFO: Saved and compressed KNeighbors pipeline to models/knn_multi_class.pkl.gz [2026-04-10 19:22:07] INFO: Balanced accuracy for KNeighbors: 0.8889

[2026-04-10 19:22:07] INFO: — Training AdaBoost — [2026-04-10 19:22:07] INFO: Saved and compressed AdaBoost pipeline to models/ada_multi_class.pkl.gz [2026-04-10 19:22:07] INFO: Balanced accuracy for AdaBoost: 1.0000

[2026-04-10 19:22:07] INFO: — Training XGBoost — [2026-04-10 19:22:08] INFO: Saved and compressed XGBoost pipeline to models/xgb_multi_class.pkl.gz [2026-04-10 19:22:08] INFO: Balanced accuracy for XGBoost: 0.8889

[2026-04-10 19:22:08] INFO: — Training MLP — [2026-04-10 19:22:08] INFO: Saved and compressed MLP pipeline to models/mlp_multi_class.pkl.gz [2026-04-10 19:22:08] INFO: Balanced accuracy for MLP: 0.3333

[2026-04-10 19:22:08] INFO: — Training DecisionTree — [2026-04-10 19:22:08] INFO: Saved and compressed DecisionTree pipeline to models/dt_multi_class.pkl.gz [2026-04-10 19:22:08] INFO: Balanced accuracy for DecisionTree: 1.0000

[2026-04-10 19:22:08] INFO: Done.