autogluon.tabular 1.5.1b20260105__py3-none-any.whl → 1.5.1b20260117__py3-none-any.whl

autogluon/tabular/configs/zeroshot/zeroshot_portfolio_gpu_2025_12_18.py

@@ -1,2 +1,422 @@
 # State-of-the-art for datasets < 100k samples. Requires a GPU with at least 20 GB VRAM.
-hyperparameter_portfolio_zeroshot_gpu_2025_12_18 = {'TABDPT': [{'ag_args': {'name_suffix': '_c1', 'priority': -3}, 'ag_args_ensemble': {'model_random_seed': 0, 'vary_seed_across_folds': False}}, {'ag_args': {'name_suffix': '_r20', 'priority': -5}, 'ag_args_ensemble': {'model_random_seed': 0, 'vary_seed_across_folds': False}, 'clip_sigma': 8, 'feature_reduction': 'subsample', 'missing_indicators': False, 'normalizer': 'quantile-uniform', 'permute_classes': False, 'temperature': 0.5}, {'ag_args': {'name_suffix': '_r1', 'priority': -7}, 'ag_args_ensemble': {'model_random_seed': 0, 'vary_seed_across_folds': False}, 'clip_sigma': 16, 'feature_reduction': 'subsample', 'missing_indicators': False, 'normalizer': 'log1p', 'permute_classes': False, 'temperature': 0.5}, {'ag_args': {'name_suffix': '_r15', 'priority': -9}, 'ag_args_ensemble': {'model_random_seed': 0, 'vary_seed_across_folds': False}, 'clip_sigma': 16, 'feature_reduction': 'subsample', 'missing_indicators': False, 'normalizer': 'standard', 'permute_classes': True, 'temperature': 0.7}, {'ag_args': {'name_suffix': '_r22', 'priority': -11}, 'ag_args_ensemble': {'model_random_seed': 0, 'vary_seed_across_folds': False}, 'clip_sigma': 8, 'feature_reduction': 'pca', 'missing_indicators': True, 'normalizer': 'robust', 'permute_classes': False, 'temperature': 0.5}], 'TABICL': [{'ag_args': {'name_suffix': '_c1', 'priority': -4}, 'ag_args_ensemble': {'refit_folds': True}}], 'MITRA': [{'ag_args': {'name_suffix': '_c1', 'priority': -12}, 'ag_args_ensemble': {'model_random_seed': 0, 'vary_seed_across_folds': True}}], 'TABM': [{'ag_args': {'name_suffix': '_r99', 'priority': -13}, 'amp': False, 'arch_type': 'tabm-mini', 'batch_size': 'auto', 'd_block': 880, 'd_embedding': 24, 'dropout': 0.10792355695428629, 'gradient_clipping_norm': 1.0, 'lr': 0.0013641856391615784, 'n_blocks': 5, 'num_emb_n_bins': 16, 'num_emb_type': 'pwl', 'patience': 16, 'share_training_batches': False, 'tabm_k': 32, 'weight_decay': 0.0}, {'ag_args': {'name_suffix': '_r124', 'priority': -17}, 'amp': False, 'arch_type': 'tabm-mini', 'batch_size': 'auto', 'd_block': 208, 'd_embedding': 16, 'dropout': 0.0, 'gradient_clipping_norm': 1.0, 'lr': 0.00042152744054701374, 'n_blocks': 2, 'num_emb_n_bins': 109, 'num_emb_type': 'pwl', 'patience': 16, 'share_training_batches': False, 'tabm_k': 32, 'weight_decay': 0.00014007839435474664}, {'ag_args': {'name_suffix': '_r69', 'priority': -21}, 'amp': False, 'arch_type': 'tabm-mini', 'batch_size': 'auto', 'd_block': 848, 'd_embedding': 28, 'dropout': 0.40215621636031007, 'gradient_clipping_norm': 1.0, 'lr': 0.0010413640454559532, 'n_blocks': 3, 'num_emb_n_bins': 18, 'num_emb_type': 'pwl', 'patience': 16, 'share_training_batches': False, 'tabm_k': 32, 'weight_decay': 0.0}, {'ag_args': {'name_suffix': '_r184', 'priority': -24}, 'amp': False, 'arch_type': 'tabm-mini', 'batch_size': 'auto', 'd_block': 864, 'd_embedding': 24, 'dropout': 0.0, 'gradient_clipping_norm': 1.0, 'lr': 0.0019256819924656217, 'n_blocks': 3, 'num_emb_n_bins': 3, 'num_emb_type': 'pwl', 'patience': 16, 'share_training_batches': False, 'tabm_k': 32, 'weight_decay': 0.0}, {'ag_args': {'name_suffix': '_r34', 'priority': -26}, 'amp': False, 'arch_type': 'tabm-mini', 'batch_size': 'auto', 'd_block': 896, 'd_embedding': 8, 'dropout': 0.0, 'gradient_clipping_norm': 1.0, 'lr': 0.002459175026451607, 'n_blocks': 4, 'num_emb_n_bins': 104, 'num_emb_type': 'pwl', 'patience': 16, 'share_training_batches': False, 'tabm_k': 32, 'weight_decay': 0.0006299584388562901}], 'GBM_PREP': [{'ag.prep_params': [[[['ArithmeticFeatureGenerator', {}]], [['CategoricalInteractionFeatureGenerator', {'passthrough': True}], ['OOFTargetEncodingFeatureGenerator', {}]]]], 'ag.prep_params.passthrough_types': {'invalid_raw_types': ['category', 'object']}, 'ag_args': {'name_suffix': '_r13', 'priority': -14}, 'ag_args_ensemble': {'model_random_seed': 0, 'vary_seed_across_folds': True}, 'bagging_fraction': 0.9923026236907, 'bagging_freq': 1, 'cat_l2': 0.014290368488, 'cat_smooth': 1.8662939903973, 'extra_trees': True, 'feature_fraction': 0.5533919718605, 'lambda_l1': 0.914411672958, 'lambda_l2': 1.90439560009, 'learning_rate': 0.0193225778401, 'max_cat_to_onehot': 18, 'min_data_in_leaf': 28, 'min_data_per_group': 54, 'num_leaves': 64}, {'ag.prep_params': [[[['ArithmeticFeatureGenerator', {}]], [['CategoricalInteractionFeatureGenerator', {'passthrough': True}], ['OOFTargetEncodingFeatureGenerator', {}]]]], 'ag.prep_params.passthrough_types': {'invalid_raw_types': ['category', 'object']}, 'ag_args': {'name_suffix': '_r41', 'priority': -16}, 'ag_args_ensemble': {'model_random_seed': 0, 'vary_seed_across_folds': True}, 'bagging_fraction': 0.7215411996558, 'bagging_freq': 1, 'cat_l2': 1.887369154362, 'cat_smooth': 0.0278693980873, 'extra_trees': True, 'feature_fraction': 0.4247583287144, 'lambda_l1': 0.1129800247772, 'lambda_l2': 0.2623265718536, 'learning_rate': 0.0074201920651, 'max_cat_to_onehot': 9, 'min_data_in_leaf': 15, 'min_data_per_group': 10, 'num_leaves': 8}, {'ag.prep_params': [[[['ArithmeticFeatureGenerator', {}]], [['CategoricalInteractionFeatureGenerator', {'passthrough': True}], ['OOFTargetEncodingFeatureGenerator', {}]]]], 'ag.prep_params.passthrough_types': {'invalid_raw_types': ['category', 'object']}, 'ag_args': {'name_suffix': '_r31', 'priority': -18}, 'ag_args_ensemble': {'model_random_seed': 0, 'vary_seed_across_folds': True}, 'bagging_fraction': 0.9591526242875, 'bagging_freq': 1, 'cat_l2': 1.8962346412823, 'cat_smooth': 0.0215219089995, 'extra_trees': False, 'feature_fraction': 0.5791844062459, 'lambda_l1': 0.938461750637, 'lambda_l2': 0.9899852075056, 'learning_rate': 0.0397613094741, 'max_cat_to_onehot': 27, 'min_data_in_leaf': 1, 'min_data_per_group': 39, 'num_leaves': 16}, {'ag.prep_params': [], 'ag_args': {'name_suffix': '_r21', 'priority': -20}, 'ag_args_ensemble': {'model_random_seed': 0, 'vary_seed_across_folds': True}, 'bagging_fraction': 0.7111549514262, 'bagging_freq': 1, 'cat_l2': 0.8679131150136, 'cat_smooth': 48.7244965504817, 'extra_trees': False, 'feature_fraction': 0.425140839263, 'lambda_l1': 0.5140528525242, 'lambda_l2': 0.5134051978198, 'learning_rate': 0.0134375321277, 'max_cat_to_onehot': 16, 'min_data_in_leaf': 2, 'min_data_per_group': 32, 'num_leaves': 20}, {'ag.prep_params': [[[['ArithmeticFeatureGenerator', {}]], [['CategoricalInteractionFeatureGenerator', {'passthrough': True}], ['OOFTargetEncodingFeatureGenerator', {}]]]], 'ag.prep_params.passthrough_types': {'invalid_raw_types': ['category', 'object']}, 'ag_args': {'name_suffix': '_r17', 'priority': -23}, 'ag_args_ensemble': {'model_random_seed': 0, 'vary_seed_across_folds': True}, 'bagging_fraction': 0.9277474245702, 'bagging_freq': 1, 'cat_l2': 0.0731876168104, 'cat_smooth': 0.1369210915339, 'extra_trees': False, 'feature_fraction': 0.6680440910385, 'lambda_l1': 0.0125057410295, 'lambda_l2': 0.7157181359874, 'learning_rate': 0.0351342879995, 'max_cat_to_onehot': 20, 'min_data_in_leaf': 1, 'min_data_per_group': 2, 'num_leaves': 64}], 'CAT': [{'ag_args': {'name_suffix': '_c1', 'priority': -15}}], 'GBM': [{'ag_args': {'name_suffix': '_r73', 'priority': -19}, 'bagging_fraction': 0.7295548973583, 'bagging_freq': 1, 'cat_l2': 1.8025485263237, 'cat_smooth': 59.6178463268351, 'extra_trees': False, 'feature_fraction': 0.8242607305914, 'lambda_l1': 0.7265522905459, 'lambda_l2': 0.3492160682092, 'learning_rate': 0.0068803786367, 'max_cat_to_onehot': 16, 'min_data_in_leaf': 1, 'min_data_per_group': 10, 'num_leaves': 24}, {'ag_args': {'name_suffix': '_r37', 'priority': -22}, 'bagging_fraction': 0.8096374561947, 'bagging_freq': 1, 'cat_l2': 1.6385754694703, 'cat_smooth': 16.1922506671724, 'extra_trees': True, 'feature_fraction': 0.885927003286, 'lambda_l1': 0.0430386950502, 'lambda_l2': 0.2507506811761, 'learning_rate': 0.0079622660542, 'max_cat_to_onehot': 23, 'min_data_in_leaf': 7, 'min_data_per_group': 49, 'num_leaves': 6}, {'ag_args': {'name_suffix': '_r162', 'priority': -25}, 'bagging_fraction': 0.7552878818396, 'bagging_freq': 1, 'cat_l2': 0.0081083103544, 'cat_smooth': 75.7373446363438, 'extra_trees': False, 'feature_fraction': 0.6171258454584, 'lambda_l1': 0.1071522383181, 'lambda_l2': 1.7882554584069, 'learning_rate': 0.0229328987255, 'max_cat_to_onehot': 24, 'min_data_in_leaf': 23, 'min_data_per_group': 2, 'num_leaves': 125}, {'ag_args': {'name_suffix': '_r57', 'priority': -27}, 'bagging_fraction': 0.8515739264605, 'bagging_freq': 1, 'cat_l2': 0.2263901847144, 'cat_smooth': 1.7397457971767, 'extra_trees': True, 'feature_fraction': 0.6284015946887, 'lambda_l1': 0.6935431676756, 'lambda_l2': 1.7605230133162, 'learning_rate': 0.0294830579218, 'max_cat_to_onehot': 52, 'min_data_in_leaf': 8, 'min_data_per_group': 3, 'num_leaves': 43}, {'ag_args': {'name_suffix': '_r33', 'priority': -28}, 'bagging_fraction': 0.9625293420216, 'bagging_freq': 1, 'cat_l2': 0.1236875455555, 'cat_smooth': 68.8584757332856, 'extra_trees': False, 'feature_fraction': 0.6189215809382, 'lambda_l1': 0.1641757352921, 'lambda_l2': 0.6937755557881, 'learning_rate': 0.0154031028561, 'max_cat_to_onehot': 17, 'min_data_in_leaf': 1, 'min_data_per_group': 30, 'num_leaves': 68}], 'REALTABPFN-V2': [{'ag_args': {'name_suffix': '_r13', 'priority': -1}, 'ag_args_ensemble': {'model_random_seed': 104, 'vary_seed_across_folds': True}, 'balance_probabilities': False, 'inference_config/OUTLIER_REMOVAL_STD': 6, 'inference_config/POLYNOMIAL_FEATURES': 'no', 'inference_config/REGRESSION_Y_PREPROCESS_TRANSFORMS': [None, 'safepower'], 'preprocessing/append_original': False, 'preprocessing/categoricals': 'numeric', 'preprocessing/global': None, 'preprocessing/scaling': ['squashing_scaler_default', 'quantile_uni_coarse'], 'softmax_temperature': 1.0, 'zip_model_path': ['tabpfn-v2-classifier-finetuned-zk73skhh.ckpt', 'tabpfn-v2-regressor-v2_default.ckpt']}, {'ag_args': {'name_suffix': '_r106', 'priority': -2}, 'ag_args_ensemble': {'model_random_seed': 848, 'vary_seed_across_folds': True}, 'balance_probabilities': False, 'inference_config/OUTLIER_REMOVAL_STD': 6, 'inference_config/POLYNOMIAL_FEATURES': 'no', 'inference_config/REGRESSION_Y_PREPROCESS_TRANSFORMS': [None], 'preprocessing/append_original': True, 'preprocessing/categoricals': 'numeric', 'preprocessing/global': 'svd_quarter_components', 'preprocessing/scaling': ['quantile_uni_coarse'], 'softmax_temperature': 0.8, 'zip_model_path': ['tabpfn-v2-classifier-finetuned-zk73skhh.ckpt', 'tabpfn-v2-regressor-v2_default.ckpt']}, {'ag_args': {'name_suffix': '_r11', 'priority': -6}, 'ag_args_ensemble': {'model_random_seed': 88, 'vary_seed_across_folds': True}, 'balance_probabilities': True, 'inference_config/OUTLIER_REMOVAL_STD': 6, 'inference_config/POLYNOMIAL_FEATURES': 25, 'inference_config/REGRESSION_Y_PREPROCESS_TRANSFORMS': [None], 'preprocessing/append_original': True, 'preprocessing/categoricals': 'onehot', 'preprocessing/global': 'svd_quarter_components', 'preprocessing/scaling': ['safepower', 'quantile_uni'], 'softmax_temperature': 0.7, 'zip_model_path': ['tabpfn-v2-classifier-finetuned-zk73skhh.ckpt', 'tabpfn-v2-regressor-v2_default.ckpt']}, {'ag_args': {'name_suffix': '_c1', 'priority': -8}, 'ag_args_ensemble': {'model_random_seed': 0, 'vary_seed_across_folds': True}, 'zip_model_path': ['tabpfn-v2-classifier-finetuned-zk73skhh.ckpt', 'tabpfn-v2-regressor-v2_default.ckpt']}, {'ag_args': {'name_suffix': '_r196', 'priority': -10}, 'ag_args_ensemble': {'model_random_seed': 1568, 'vary_seed_across_folds': True}, 'balance_probabilities': False, 'inference_config/OUTLIER_REMOVAL_STD': 12, 'inference_config/POLYNOMIAL_FEATURES': 'no', 'inference_config/REGRESSION_Y_PREPROCESS_TRANSFORMS': ['kdi_alpha_1.0'], 'preprocessing/append_original': False, 'preprocessing/categoricals': 'numeric', 'preprocessing/global': None, 'preprocessing/scaling': ['squashing_scaler_default'], 'softmax_temperature': 1.25, 'zip_model_path': ['tabpfn-v2-classifier-finetuned-zk73skhh.ckpt', 'tabpfn-v2-regressor-v2_default.ckpt']}]}
+hyperparameter_portfolio_zeroshot_gpu_2025_12_18 = {
+    "TABDPT": [
+        {
+            "ag_args": {"name_suffix": "_c1", "priority": -3},
+            "ag_args_ensemble": {"model_random_seed": 0, "vary_seed_across_folds": False},
+        },
+        {
+            "ag_args": {"name_suffix": "_r20", "priority": -5},
+            "ag_args_ensemble": {"model_random_seed": 0, "vary_seed_across_folds": False},
+            "clip_sigma": 8,
+            "feature_reduction": "subsample",
+            "missing_indicators": False,
+            "normalizer": "quantile-uniform",
+            "permute_classes": False,
+            "temperature": 0.5,
+        },
+        {
+            "ag_args": {"name_suffix": "_r1", "priority": -7},
+            "ag_args_ensemble": {"model_random_seed": 0, "vary_seed_across_folds": False},
+            "clip_sigma": 16,
+            "feature_reduction": "subsample",
+            "missing_indicators": False,
+            "normalizer": "log1p",
+            "permute_classes": False,
+            "temperature": 0.5,
+        },
+        {
+            "ag_args": {"name_suffix": "_r15", "priority": -9},
+            "ag_args_ensemble": {"model_random_seed": 0, "vary_seed_across_folds": False},
+            "clip_sigma": 16,
+            "feature_reduction": "subsample",
+            "missing_indicators": False,
+            "normalizer": "standard",
+            "permute_classes": True,
+            "temperature": 0.7,
+        },
+        {
+            "ag_args": {"name_suffix": "_r22", "priority": -11},
+            "ag_args_ensemble": {"model_random_seed": 0, "vary_seed_across_folds": False},
+            "clip_sigma": 8,
+            "feature_reduction": "pca",
+            "missing_indicators": True,
+            "normalizer": "robust",
+            "permute_classes": False,
+            "temperature": 0.5,
+        },
+    ],
+    "TABICL": [{"ag_args": {"name_suffix": "_c1", "priority": -4}, "ag_args_ensemble": {"refit_folds": True}}],
+    "MITRA": [
+        {
+            "ag_args": {"name_suffix": "_c1", "priority": -12},
+            "ag_args_ensemble": {"model_random_seed": 0, "vary_seed_across_folds": True},
+        }
+    ],
+    "TABM": [
+        {
+            "ag_args": {"name_suffix": "_r99", "priority": -13},
+            "amp": False,
+            "arch_type": "tabm-mini",
+            "batch_size": "auto",
+            "d_block": 880,
+            "d_embedding": 24,
+            "dropout": 0.10792355695428629,
+            "gradient_clipping_norm": 1.0,
+            "lr": 0.0013641856391615784,
+            "n_blocks": 5,
+            "num_emb_n_bins": 16,
+            "num_emb_type": "pwl",
+            "patience": 16,
+            "share_training_batches": False,
+            "tabm_k": 32,
+            "weight_decay": 0.0,
+        },
+        {
+            "ag_args": {"name_suffix": "_r124", "priority": -17},
+            "amp": False,
+            "arch_type": "tabm-mini",
+            "batch_size": "auto",
+            "d_block": 208,
+            "d_embedding": 16,
+            "dropout": 0.0,
+            "gradient_clipping_norm": 1.0,
+            "lr": 0.00042152744054701374,
+            "n_blocks": 2,
+            "num_emb_n_bins": 109,
+            "num_emb_type": "pwl",
+            "patience": 16,
+            "share_training_batches": False,
+            "tabm_k": 32,
+            "weight_decay": 0.00014007839435474664,
+        },
+        {
+            "ag_args": {"name_suffix": "_r69", "priority": -21},
+            "amp": False,
+            "arch_type": "tabm-mini",
+            "batch_size": "auto",
+            "d_block": 848,
+            "d_embedding": 28,
+            "dropout": 0.40215621636031007,
+            "gradient_clipping_norm": 1.0,
+            "lr": 0.0010413640454559532,
+            "n_blocks": 3,
+            "num_emb_n_bins": 18,
+            "num_emb_type": "pwl",
+            "patience": 16,
+            "share_training_batches": False,
+            "tabm_k": 32,
+            "weight_decay": 0.0,
+        },
+        {
+            "ag_args": {"name_suffix": "_r184", "priority": -24},
+            "amp": False,
+            "arch_type": "tabm-mini",
+            "batch_size": "auto",
+            "d_block": 864,
+            "d_embedding": 24,
+            "dropout": 0.0,
+            "gradient_clipping_norm": 1.0,
+            "lr": 0.0019256819924656217,
+            "n_blocks": 3,
+            "num_emb_n_bins": 3,
+            "num_emb_type": "pwl",
+            "patience": 16,
+            "share_training_batches": False,
+            "tabm_k": 32,
+            "weight_decay": 0.0,
+        },
+        {
+            "ag_args": {"name_suffix": "_r34", "priority": -26},
+            "amp": False,
+            "arch_type": "tabm-mini",
+            "batch_size": "auto",
+            "d_block": 896,
+            "d_embedding": 8,
+            "dropout": 0.0,
+            "gradient_clipping_norm": 1.0,
+            "lr": 0.002459175026451607,
+            "n_blocks": 4,
+            "num_emb_n_bins": 104,
+            "num_emb_type": "pwl",
+            "patience": 16,
+            "share_training_batches": False,
+            "tabm_k": 32,
+            "weight_decay": 0.0006299584388562901,
+        },
+    ],
+    "GBM_PREP": [
+        {
+            "ag.prep_params": [
+                [
+                    [["ArithmeticFeatureGenerator", {}]],
+                    [
+                        ["CategoricalInteractionFeatureGenerator", {"passthrough": True}],
+                        ["OOFTargetEncodingFeatureGenerator", {}],
+                    ],
+                ]
+            ],
+            "ag.prep_params.passthrough_types": {"invalid_raw_types": ["category", "object"]},
+            "ag_args": {"name_suffix": "_r13", "priority": -14},
+            "ag_args_ensemble": {"model_random_seed": 0, "vary_seed_across_folds": True},
+            "bagging_fraction": 0.9923026236907,
+            "bagging_freq": 1,
+            "cat_l2": 0.014290368488,
+            "cat_smooth": 1.8662939903973,
+            "extra_trees": True,
+            "feature_fraction": 0.5533919718605,
+            "lambda_l1": 0.914411672958,
+            "lambda_l2": 1.90439560009,
+            "learning_rate": 0.0193225778401,
+            "max_cat_to_onehot": 18,
+            "min_data_in_leaf": 28,
+            "min_data_per_group": 54,
+            "num_leaves": 64,
+        },
+        {
+            "ag.prep_params": [
+                [
+                    [["ArithmeticFeatureGenerator", {}]],
+                    [
+                        ["CategoricalInteractionFeatureGenerator", {"passthrough": True}],
+                        ["OOFTargetEncodingFeatureGenerator", {}],
+                    ],
+                ]
+            ],
+            "ag.prep_params.passthrough_types": {"invalid_raw_types": ["category", "object"]},
+            "ag_args": {"name_suffix": "_r41", "priority": -16},
+            "ag_args_ensemble": {"model_random_seed": 0, "vary_seed_across_folds": True},
+            "bagging_fraction": 0.7215411996558,
+            "bagging_freq": 1,
+            "cat_l2": 1.887369154362,
+            "cat_smooth": 0.0278693980873,
+            "extra_trees": True,
+            "feature_fraction": 0.4247583287144,
+            "lambda_l1": 0.1129800247772,
+            "lambda_l2": 0.2623265718536,
+            "learning_rate": 0.0074201920651,
+            "max_cat_to_onehot": 9,
+            "min_data_in_leaf": 15,
+            "min_data_per_group": 10,
+            "num_leaves": 8,
+        },
+        {
+            "ag.prep_params": [
+                [
+                    [["ArithmeticFeatureGenerator", {}]],
+                    [
+                        ["CategoricalInteractionFeatureGenerator", {"passthrough": True}],
+                        ["OOFTargetEncodingFeatureGenerator", {}],
+                    ],
+                ]
+            ],
+            "ag.prep_params.passthrough_types": {"invalid_raw_types": ["category", "object"]},
+            "ag_args": {"name_suffix": "_r31", "priority": -18},
+            "ag_args_ensemble": {"model_random_seed": 0, "vary_seed_across_folds": True},
+            "bagging_fraction": 0.9591526242875,
+            "bagging_freq": 1,
+            "cat_l2": 1.8962346412823,
+            "cat_smooth": 0.0215219089995,
+            "extra_trees": False,
+            "feature_fraction": 0.5791844062459,
+            "lambda_l1": 0.938461750637,
+            "lambda_l2": 0.9899852075056,
+            "learning_rate": 0.0397613094741,
+            "max_cat_to_onehot": 27,
+            "min_data_in_leaf": 1,
+            "min_data_per_group": 39,
+            "num_leaves": 16,
+        },
+        {
+            "ag.prep_params": [],
+            "ag_args": {"name_suffix": "_r21", "priority": -20},
+            "ag_args_ensemble": {"model_random_seed": 0, "vary_seed_across_folds": True},
+            "bagging_fraction": 0.7111549514262,
+            "bagging_freq": 1,
+            "cat_l2": 0.8679131150136,
+            "cat_smooth": 48.7244965504817,
+            "extra_trees": False,
+            "feature_fraction": 0.425140839263,
+            "lambda_l1": 0.5140528525242,
+            "lambda_l2": 0.5134051978198,
+            "learning_rate": 0.0134375321277,
+            "max_cat_to_onehot": 16,
+            "min_data_in_leaf": 2,
+            "min_data_per_group": 32,
+            "num_leaves": 20,
+        },
+        {
+            "ag.prep_params": [
+                [
+                    [["ArithmeticFeatureGenerator", {}]],
+                    [
+                        ["CategoricalInteractionFeatureGenerator", {"passthrough": True}],
+                        ["OOFTargetEncodingFeatureGenerator", {}],
+                    ],
+                ]
+            ],
+            "ag.prep_params.passthrough_types": {"invalid_raw_types": ["category", "object"]},
+            "ag_args": {"name_suffix": "_r17", "priority": -23},
+            "ag_args_ensemble": {"model_random_seed": 0, "vary_seed_across_folds": True},
+            "bagging_fraction": 0.9277474245702,
+            "bagging_freq": 1,
+            "cat_l2": 0.0731876168104,
+            "cat_smooth": 0.1369210915339,
+            "extra_trees": False,
+            "feature_fraction": 0.6680440910385,
+            "lambda_l1": 0.0125057410295,
+            "lambda_l2": 0.7157181359874,
+            "learning_rate": 0.0351342879995,
+            "max_cat_to_onehot": 20,
+            "min_data_in_leaf": 1,
+            "min_data_per_group": 2,
+            "num_leaves": 64,
+        },
+    ],
+    "CAT": [{"ag_args": {"name_suffix": "_c1", "priority": -15}}],
+    "GBM": [
+        {
+            "ag_args": {"name_suffix": "_r73", "priority": -19},
+            "bagging_fraction": 0.7295548973583,
+            "bagging_freq": 1,
+            "cat_l2": 1.8025485263237,
+            "cat_smooth": 59.6178463268351,
+            "extra_trees": False,
+            "feature_fraction": 0.8242607305914,
+            "lambda_l1": 0.7265522905459,
+            "lambda_l2": 0.3492160682092,
+            "learning_rate": 0.0068803786367,
+            "max_cat_to_onehot": 16,
+            "min_data_in_leaf": 1,
+            "min_data_per_group": 10,
+            "num_leaves": 24,
+        },
+        {
+            "ag_args": {"name_suffix": "_r37", "priority": -22},
+            "bagging_fraction": 0.8096374561947,
+            "bagging_freq": 1,
+            "cat_l2": 1.6385754694703,
+            "cat_smooth": 16.1922506671724,
+            "extra_trees": True,
+            "feature_fraction": 0.885927003286,
+            "lambda_l1": 0.0430386950502,
+            "lambda_l2": 0.2507506811761,
+            "learning_rate": 0.0079622660542,
+            "max_cat_to_onehot": 23,
+            "min_data_in_leaf": 7,
+            "min_data_per_group": 49,
+            "num_leaves": 6,
+        },
+        {
+            "ag_args": {"name_suffix": "_r162", "priority": -25},
+            "bagging_fraction": 0.7552878818396,
+            "bagging_freq": 1,
+            "cat_l2": 0.0081083103544,
+            "cat_smooth": 75.7373446363438,
+            "extra_trees": False,
+            "feature_fraction": 0.6171258454584,
+            "lambda_l1": 0.1071522383181,
+            "lambda_l2": 1.7882554584069,
+            "learning_rate": 0.0229328987255,
+            "max_cat_to_onehot": 24,
+            "min_data_in_leaf": 23,
+            "min_data_per_group": 2,
+            "num_leaves": 125,
+        },
+        {
+            "ag_args": {"name_suffix": "_r57", "priority": -27},
+            "bagging_fraction": 0.8515739264605,
+            "bagging_freq": 1,
+            "cat_l2": 0.2263901847144,
+            "cat_smooth": 1.7397457971767,
+            "extra_trees": True,
+            "feature_fraction": 0.6284015946887,
+            "lambda_l1": 0.6935431676756,
+            "lambda_l2": 1.7605230133162,
+            "learning_rate": 0.0294830579218,
+            "max_cat_to_onehot": 52,
+            "min_data_in_leaf": 8,
+            "min_data_per_group": 3,
+            "num_leaves": 43,
+        },
+        {
+            "ag_args": {"name_suffix": "_r33", "priority": -28},
+            "bagging_fraction": 0.9625293420216,
+            "bagging_freq": 1,
+            "cat_l2": 0.1236875455555,
+            "cat_smooth": 68.8584757332856,
+            "extra_trees": False,
+            "feature_fraction": 0.6189215809382,
+            "lambda_l1": 0.1641757352921,
+            "lambda_l2": 0.6937755557881,
+            "learning_rate": 0.0154031028561,
+            "max_cat_to_onehot": 17,
+            "min_data_in_leaf": 1,
+            "min_data_per_group": 30,
+            "num_leaves": 68,
+        },
+    ],
+    "REALTABPFN-V2": [
+        {
+            "ag_args": {"name_suffix": "_r13", "priority": -1},
+            "ag_args_ensemble": {"model_random_seed": 104, "vary_seed_across_folds": True},
+            "balance_probabilities": False,
+            "inference_config/OUTLIER_REMOVAL_STD": 6,
+            "inference_config/POLYNOMIAL_FEATURES": "no",
+            "inference_config/REGRESSION_Y_PREPROCESS_TRANSFORMS": [None, "safepower"],
+            "preprocessing/append_original": False,
+            "preprocessing/categoricals": "numeric",
+            "preprocessing/global": None,
+            "preprocessing/scaling": ["squashing_scaler_default", "quantile_uni_coarse"],
+            "softmax_temperature": 1.0,
+            "zip_model_path": ["tabpfn-v2-classifier-finetuned-zk73skhh.ckpt", "tabpfn-v2-regressor-v2_default.ckpt"],
+        },
+        {
+            "ag_args": {"name_suffix": "_r106", "priority": -2},
+            "ag_args_ensemble": {"model_random_seed": 848, "vary_seed_across_folds": True},
+            "balance_probabilities": False,
+            "inference_config/OUTLIER_REMOVAL_STD": 6,
+            "inference_config/POLYNOMIAL_FEATURES": "no",
+            "inference_config/REGRESSION_Y_PREPROCESS_TRANSFORMS": [None],
+            "preprocessing/append_original": True,
+            "preprocessing/categoricals": "numeric",
+            "preprocessing/global": "svd_quarter_components",
+            "preprocessing/scaling": ["quantile_uni_coarse"],
+            "softmax_temperature": 0.8,
+            "zip_model_path": ["tabpfn-v2-classifier-finetuned-zk73skhh.ckpt", "tabpfn-v2-regressor-v2_default.ckpt"],
+        },
+        {
+            "ag_args": {"name_suffix": "_r11", "priority": -6},
+            "ag_args_ensemble": {"model_random_seed": 88, "vary_seed_across_folds": True},
+            "balance_probabilities": True,
+            "inference_config/OUTLIER_REMOVAL_STD": 6,
+            "inference_config/POLYNOMIAL_FEATURES": 25,
+            "inference_config/REGRESSION_Y_PREPROCESS_TRANSFORMS": [None],
+            "preprocessing/append_original": True,
+            "preprocessing/categoricals": "onehot",
+            "preprocessing/global": "svd_quarter_components",
+            "preprocessing/scaling": ["safepower", "quantile_uni"],
+            "softmax_temperature": 0.7,
+            "zip_model_path": ["tabpfn-v2-classifier-finetuned-zk73skhh.ckpt", "tabpfn-v2-regressor-v2_default.ckpt"],
+        },
+        {
+            "ag_args": {"name_suffix": "_c1", "priority": -8},
+            "ag_args_ensemble": {"model_random_seed": 0, "vary_seed_across_folds": True},
+            "zip_model_path": ["tabpfn-v2-classifier-finetuned-zk73skhh.ckpt", "tabpfn-v2-regressor-v2_default.ckpt"],
+        },
+        {
+            "ag_args": {"name_suffix": "_r196", "priority": -10},
+            "ag_args_ensemble": {"model_random_seed": 1568, "vary_seed_across_folds": True},
+            "balance_probabilities": False,
+            "inference_config/OUTLIER_REMOVAL_STD": 12,
+            "inference_config/POLYNOMIAL_FEATURES": "no",
+            "inference_config/REGRESSION_Y_PREPROCESS_TRANSFORMS": ["kdi_alpha_1.0"],
+            "preprocessing/append_original": False,
+            "preprocessing/categoricals": "numeric",
+            "preprocessing/global": None,
+            "preprocessing/scaling": ["squashing_scaler_default"],
+            "softmax_temperature": 1.25,
+            "zip_model_path": ["tabpfn-v2-classifier-finetuned-zk73skhh.ckpt", "tabpfn-v2-regressor-v2_default.ckpt"],
+        },
+    ],
+}

autogluon/tabular/experimental/_scikit_mixin.py

@@ -44,13 +44,17 @@ class ScikitMixin:
         # Input validation
         X = check_array(X)
         if X.shape[1] != self.n_features_in_:
-            raise ValueError(f"Inconsistent number of features between fit and predict calls: ({self.n_features_in_}, {X.shape[1]})")
+            raise ValueError(
+                f"Inconsistent number of features between fit and predict calls: ({self.n_features_in_}, {X.shape[1]})"
+            )
         return X
 
     def _combine_X_y(self, X, y) -> pd.DataFrame:
         label = self.predictor_.label
         X = pd.DataFrame(X)
-        assert label not in list(X.columns), f"Cannot have column named {label}. Please rename the column to a different value."
+        assert label not in list(X.columns), (
+            f"Cannot have column named {label}. Please rename the column to a different value."
+        )
         X[label] = y
         return X
 

autogluon/tabular/experimental/_tabular_classifier.py

@@ -65,7 +65,9 @@ class TabularClassifier(BaseEstimator, ClassifierMixin, ScikitMixin):
         # Input validation
         X = check_array(X)
         if X.shape[1] != self.n_features_in_:
-            raise ValueError(f"Inconsistent number of features between fit and predict calls: ({self.n_features_in_}, {X.shape[1]})")
+            raise ValueError(
+                f"Inconsistent number of features between fit and predict calls: ({self.n_features_in_}, {X.shape[1]})"
+            )
 
         data = pd.DataFrame(X)
         y_pred = self.predictor_.predict(data=data).to_numpy()

autogluon/tabular/experimental/_tabular_regressor.py

@@ -55,7 +55,9 @@ class TabularRegressor(BaseEstimator, RegressorMixin, ScikitMixin):
         # Input validation
         X = check_array(X)
         if X.shape[1] != self.n_features_in_:
-            raise ValueError(f"Inconsistent number of features between fit and predict calls: ({self.n_features_in_}, {X.shape[1]})")
+            raise ValueError(
+                f"Inconsistent number of features between fit and predict calls: ({self.n_features_in_}, {X.shape[1]})"
+            )
 
         data = pd.DataFrame(X)
         y_pred = self.predictor_.predict(data=data).to_numpy()

autogluon/tabular/experimental/plot_leaderboard.py

@@ -1,7 +1,7 @@
 from __future__ import annotations
 
-import pandas as pd
 import matplotlib.pyplot as plt
+import pandas as pd
 from matplotlib.figure import Figure
 
 from autogluon.tabular import TabularPredictor
@@ -9,7 +9,7 @@ from autogluon.tabular import TabularPredictor
 
 def _cumulative_min_idx(series: pd.Series) -> pd.Series:
     """
-    
+
     Parameters
     ----------
     series: pd.Series
@@ -20,7 +20,7 @@ def _cumulative_min_idx(series: pd.Series) -> pd.Series:
         The index of the cumulative min of the series values.
 
     """
-    min_val = float('inf')
+    min_val = float("inf")
     min_index = -1
     result = []
     for i, val in enumerate(series):
@@ -54,7 +54,9 @@ def compute_cumulative_leaderboard_stats(leaderboard: pd.DataFrame) -> pd.DataFr
     leaderboard["time_so_far"] = leaderboard["fit_time_marginal"].cumsum()
     leaderboard["metric_error_val_so_far"] = leaderboard["best_model_so_far"].map(leaderboard["metric_error_val"])
     if "metric_error_test" in leaderboard:
-        leaderboard["metric_error_test_so_far"] = leaderboard["best_model_so_far"].map(leaderboard["metric_error_test"])
+        leaderboard["metric_error_test_so_far"] = leaderboard["best_model_so_far"].map(
+            leaderboard["metric_error_test"]
+        )
     leaderboard = leaderboard.reset_index(drop=False).set_index("fit_order")
     return leaderboard
 
@@ -88,7 +90,7 @@ def compute_cumulative_leaderboard_stats_ensemble(
     model_fit_order = list(leaderboard_stats["model"])
     ens_names = []
     for i in range(len(model_fit_order)):
-        models_to_ens = model_fit_order[:i + 1]
+        models_to_ens = model_fit_order[: i + 1]
         ens_name = predictor.fit_weighted_ensemble(base_models=models_to_ens, name_suffix=f"_fit_{i + 1}")[0]
         ens_names.append(ens_name)
 
@@ -144,10 +146,14 @@ def plot_leaderboard_from_predictor(
     """
     leaderboard = predictor.leaderboard(test_data, score_format="error", display=False)
     if ensemble:
-        leaderboard_order_sorted = compute_cumulative_leaderboard_stats_ensemble(leaderboard=leaderboard, test_data=test_data, predictor=predictor)
+        leaderboard_order_sorted = compute_cumulative_leaderboard_stats_ensemble(
+            leaderboard=leaderboard, test_data=test_data, predictor=predictor
+        )
     else:
         leaderboard_order_sorted = compute_cumulative_leaderboard_stats(leaderboard=leaderboard)
-    return plot_leaderboard(leaderboard=leaderboard_order_sorted, preprocess=False, ensemble=ensemble, include_val=include_val)
+    return plot_leaderboard(
+        leaderboard=leaderboard_order_sorted, preprocess=False, ensemble=ensemble, include_val=include_val
+    )
 
 
 def plot_leaderboard(
@@ -198,36 +204,84 @@ def plot_leaderboard(
 
     # TODO: View on inference time, can take from ensemble model, 3rd dimension, color?
     fig, axes = plt.subplots(1, 2, sharey=True)
-    fig.suptitle('AutoGluon Metric Error Over Time')
+    fig.suptitle("AutoGluon Metric Error Over Time")
 
     ax = axes[0]
 
     if include_test:
-        ax.plot(leaderboard_order_sorted.index, leaderboard_order_sorted["metric_error_test_so_far"].values, '-', color="b", label="test")
+        ax.plot(
+            leaderboard_order_sorted.index,
+            leaderboard_order_sorted["metric_error_test_so_far"].values,
+            "-",
+            color="b",
+            label="test",
+        )
     if include_val:
-        ax.plot(leaderboard_order_sorted.index, leaderboard_order_sorted["metric_error_val_so_far"].values, '-', color="orange", label="val")
+        ax.plot(
+            leaderboard_order_sorted.index,
+            leaderboard_order_sorted["metric_error_val_so_far"].values,
+            "-",
+            color="orange",
+            label="val",
+        )
     if ensemble:
         if include_test:
-            ax.plot(leaderboard_order_sorted.index, leaderboard_order_sorted["metric_error_test_so_far_ens"].values, '--', color="b", label="test (ens)")
+            ax.plot(
+                leaderboard_order_sorted.index,
+                leaderboard_order_sorted["metric_error_test_so_far_ens"].values,
+                "--",
+                color="b",
+                label="test (ens)",
+            )
         if include_val:
-            ax.plot(leaderboard_order_sorted.index, leaderboard_order_sorted["metric_error_val_so_far_ens"].values, '--', color="orange", label="val (ens)")
+            ax.plot(
+                leaderboard_order_sorted.index,
+                leaderboard_order_sorted["metric_error_val_so_far_ens"].values,
+                "--",
+                color="orange",
+                label="val (ens)",
+            )
     ax.set_xlim(left=1, right=leaderboard_order_sorted.index.max())
-    ax.set_xlabel('# Models Fit')
-    ax.set_ylabel(f'Metric Error ({eval_metric})')
+    ax.set_xlabel("# Models Fit")
+    ax.set_ylabel(f"Metric Error ({eval_metric})")
     ax.grid()
 
     ax = axes[1]
 
     if include_test:
-        ax.plot(leaderboard_order_sorted["time_so_far"].values, leaderboard_order_sorted["metric_error_test_so_far"].values, '-', color="b", label="test")
+        ax.plot(
+            leaderboard_order_sorted["time_so_far"].values,
+            leaderboard_order_sorted["metric_error_test_so_far"].values,
+            "-",
+            color="b",
+            label="test",
+        )
     if include_val:
-        ax.plot(leaderboard_order_sorted["time_so_far"].values, leaderboard_order_sorted["metric_error_val_so_far"].values, '-', color="orange", label="val")
+        ax.plot(
+            leaderboard_order_sorted["time_so_far"].values,
+            leaderboard_order_sorted["metric_error_val_so_far"].values,
+            "-",
+            color="orange",
+            label="val",
+        )
     if ensemble:
         if include_test:
-            ax.plot(leaderboard_order_sorted["time_so_far"].values, leaderboard_order_sorted["metric_error_test_so_far_ens"].values, '--', color="b", label="test (ens)")
+            ax.plot(
+                leaderboard_order_sorted["time_so_far"].values,
+                leaderboard_order_sorted["metric_error_test_so_far_ens"].values,
+                "--",
+                color="b",
+                label="test (ens)",
+            )
         if include_val:
-            ax.plot(leaderboard_order_sorted["time_so_far"].values, leaderboard_order_sorted["metric_error_val_so_far_ens"].values, '--', color="orange", label="val (ens)")
-    ax.set_xlabel('Time Elapsed (s)')
+            ax.plot(
+                leaderboard_order_sorted["time_so_far"].values,
+                leaderboard_order_sorted["metric_error_val_so_far_ens"].values,
+                "--",
+                color="orange",
+                label="val (ens)",
+            )
+    ax.set_xlabel("Time Elapsed (s)")
     ax.grid()
     ax.legend()