Upload folder using huggingface_hub

.gitignore

@@ -0,0 +1,207 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[codz]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py.cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# UV
+#   Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#uv.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+#poetry.toml
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#   pdm recommends including project-wide configuration in pdm.toml, but excluding .pdm-python.
+#   https://pdm-project.org/en/latest/usage/project/#working-with-version-control
+#pdm.lock
+#pdm.toml
+.pdm-python
+.pdm-build/
+
+# pixi
+#   Similar to Pipfile.lock, it is generally recommended to include pixi.lock in version control.
+#pixi.lock
+#   Pixi creates a virtual environment in the .pixi directory, just like venv module creates one
+#   in the .venv directory. It is recommended not to include this directory in version control.
+.pixi
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.envrc
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/
+
+# Abstra
+# Abstra is an AI-powered process automation framework.
+# Ignore directories containing user credentials, local state, and settings.
+# Learn more at https://abstra.io/docs
+.abstra/
+
+# Visual Studio Code
+#  Visual Studio Code specific template is maintained in a separate VisualStudioCode.gitignore 
+#  that can be found at https://github.com/github/gitignore/blob/main/Global/VisualStudioCode.gitignore
+#  and can be added to the global gitignore or merged into this file. However, if you prefer, 
+#  you could uncomment the following to ignore the entire vscode folder
+# .vscode/
+
+# Ruff stuff:
+.ruff_cache/
+
+# PyPI configuration file
+.pypirc
+
+# Cursor
+#  Cursor is an AI-powered code editor. `.cursorignore` specifies files/directories to
+#  exclude from AI features like autocomplete and code analysis. Recommended for sensitive data
+#  refer to https://docs.cursor.com/context/ignore-files
+.cursorignore
+.cursorindexingignore
+
+# Marimo
+marimo/_static/
+marimo/_lsp/
+__marimo__/

config.json

@@ -0,0 +1,8 @@
+{
+  "architectures": ["SmSubgroupClassifier"],
+  "auto_map": {
+    "AutoConfig": "configuration_sm_subgroup_classifier.SmSubgroupClassifierConfig",
+    "AutoModel": "modeling_sm_subgroup_classifier.SmSubgroupClassifier"
+  },
+  "model_type": "sm_subgroup_classifier"
+}

configuration_sm_subgroup_classifier.py

@@ -0,0 +1,9 @@
+from transformers import PretrainedConfig
+
+
+class SmSubgroupClassifierConfig(PretrainedConfig):
+    model_type = "sm_subgroup_classifier"
+
+    def __init__(self, available_models=None, **kwargs):
+        super().__init__(**kwargs)
+        self.available_models = available_models or []

en_ID/metadata.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ca61fbfd1ba2ce417a30bde0d53adf2ffd1ea1a8c6186eff0149c770f001c10
+size 62

en_ID/model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48dc684753e19792ce04cadeae9025441fc6507396eb93ab6dc740f87f1f3dad
+size 9055

en_ID/scaler.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f230b5ecc0b0cf4a07aec1ab185ebea9060ea6eca6f7780119e9416ec15c0f1
+size 25191

en_ID/training_details.txt

@@ -0,0 +1,39 @@
+Training Details for en_ID
+========================================
+
+Language: en
+Register: ID
+Training Date: 2025-09-26 14:09:04
+
+Data Summary:
+- Total samples: 41071
+- Training samples: 32856
+- Test samples: 8215
+- Embedding dimension: 1024
+
+Classes:
+- Number of classes: 2
+- Class names: 'sports', ''
+- Class distribution: {'sports': 983, '': 40088}
+
+Cross-Validation Results:
+- CV folds: 5
+- CV scores: [0.9984783931832014, 0.9987825292953888, 0.9975650585907777, 0.9989347131334653, 0.9981737939430833]
+- CV mean: 0.9984
+- CV std: 0.0005
+- CV confidence interval: 0.9984 ± 0.0010
+
+Final Performance:
+- Test accuracy: 0.9988
+
+Model Configuration:
+- Algorithm: Logistic Regression
+- Regularization (C): 1.0
+- Feature scaling: StandardScaler
+- Random state: 42
+
+Files:
+- Classifier: model.pkl
+- Scaler: scaler.pkl
+- Metadata: metadata.pkl
+- This file: training_details.txt

en_NA-nb-OP/metadata.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33039d60b51479962f055e496f8b49f16e8c9dd1d424896502d1fe6db7cb9005
+size 84

en_NA-nb-OP/model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f83d56eccb7a610a0b57baa16d63af8839c571455358919a4d892bcf581a66e
+size 33671

en_NA-nb-OP/scaler.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85019b0b45399ef17ada3a02dbb3c77b0783fca6e0e18bd4e8e818e740f304c7
+size 25191

en_NA-nb-OP/training_details.txt

@@ -0,0 +1,39 @@
+Training Details for en_NA-nb-OP
+========================================
+
+Language: en
+Register: NA-nb-OP
+Training Date: 2025-09-26 14:08:46
+
+Data Summary:
+- Total samples: 1536
+- Training samples: 1228
+- Test samples: 308
+- Embedding dimension: 1024
+
+Classes:
+- Number of classes: 4
+- Class names: '', 'culture', 'dining', 'lifestyle'
+- Class distribution: {'': 747, 'culture': 327, 'dining': 172, 'lifestyle': 290}
+
+Cross-Validation Results:
+- CV folds: 5
+- CV scores: [0.9634146341463414, 0.959349593495935, 0.9634146341463414, 0.9877551020408163, 0.9755102040816327]
+- CV mean: 0.9699
+- CV std: 0.0104
+- CV confidence interval: 0.9699 ± 0.0209
+
+Final Performance:
+- Test accuracy: 0.9643
+
+Model Configuration:
+- Algorithm: Logistic Regression
+- Regularization (C): 1.0
+- Feature scaling: StandardScaler
+- Random state: 42
+
+Files:
+- Classifier: model.pkl
+- Scaler: scaler.pkl
+- Metadata: metadata.pkl
+- This file: training_details.txt

en_NA-nb/metadata.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:836ae4c831fe01dd10837795f65517e2768a1b1bd5364187226b09a39044826b
+size 64

en_NA-nb/model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:97142d5962fb70803f26b449fa21ced30015b550de8604dc850ad50c07bee259
+size 9055

en_NA-nb/scaler.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7423f9e8f84672e12ec4a45f5886d488146a6b5f20fc6a48db4977e89f663078
+size 25191

en_NA-nb/training_details.txt

@@ -0,0 +1,39 @@
+Training Details for en_NA-nb
+========================================
+
+Language: en
+Register: NA-nb
+Training Date: 2025-09-26 14:08:58
+
+Data Summary:
+- Total samples: 49026
+- Training samples: 39220
+- Test samples: 9806
+- Embedding dimension: 1024
+
+Classes:
+- Number of classes: 2
+- Class names: '', 'comments'
+- Class distribution: {'': 46239, 'comments': 2787}
+
+Cross-Validation Results:
+- CV folds: 5
+- CV scores: [0.989673635900051, 0.9886537480877103, 0.9903110657827638, 0.9899286078531362, 0.9903110657827638]
+- CV mean: 0.9898
+- CV std: 0.0006
+- CV confidence interval: 0.9898 ± 0.0012
+
+Final Performance:
+- Test accuracy: 0.9897
+
+Model Configuration:
+- Algorithm: Logistic Regression
+- Regularization (C): 1.0
+- Feature scaling: StandardScaler
+- Random state: 42
+
+Files:
+- Classifier: model.pkl
+- Scaler: scaler.pkl
+- Metadata: metadata.pkl
+- This file: training_details.txt

example.py

@@ -0,0 +1,15 @@
+import numpy as np
+from transformers import AutoModel
+
+# Load sm_subgroup_classifier
+sm_classifier = AutoModel.from_pretrained(
+    "erikhenriksson/sm-subgroup-classifier", trust_remote_code=True
+)
+
+# create a random 1024 dimensional embedding
+embedding = np.random.rand(1024).astype(np.float32)
+
+# Use - model automatically discovers what's available
+result = sm_classifier("fi", "NA-nb", embedding)
+
+print(result)

fi_ID-NA/metadata.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:836ae4c831fe01dd10837795f65517e2768a1b1bd5364187226b09a39044826b
+size 64

fi_ID-NA/model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:91641ec86de57b7e5bc916dd05cad1bc6189dcd58a67f4f3cc8516af6ac5b228
+size 9055

fi_ID-NA/scaler.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b61d42366ac1a5e42e2924c14aac56e71c18c92670fef5cef8072ae4698c0612
+size 25191

fi_ID-NA/training_details.txt

@@ -0,0 +1,39 @@
+Training Details for fi_ID-NA
+========================================
+
+Language: fi
+Register: ID-NA
+Training Date: 2025-09-26 14:06:57
+
+Data Summary:
+- Total samples: 1729
+- Training samples: 1383
+- Test samples: 346
+- Embedding dimension: 1024
+
+Classes:
+- Number of classes: 2
+- Class names: '', 'comments'
+- Class distribution: {'': 389, 'comments': 1340}
+
+Cross-Validation Results:
+- CV folds: 5
+- CV scores: [0.9855595667870036, 0.9927797833935018, 0.9927797833935018, 0.9927536231884058, 0.9818840579710145]
+- CV mean: 0.9892
+- CV std: 0.0046
+- CV confidence interval: 0.9892 ± 0.0092
+
+Final Performance:
+- Test accuracy: 0.9913
+
+Model Configuration:
+- Algorithm: Logistic Regression
+- Regularization (C): 1.0
+- Feature scaling: StandardScaler
+- Random state: 42
+
+Files:
+- Classifier: model.pkl
+- Scaler: scaler.pkl
+- Metadata: metadata.pkl
+- This file: training_details.txt

fi_ID/metadata.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e193cf54249f175dabf353519ff468ff9b644e995a749d1abb9d45286eaa185b
+size 62

fi_ID/model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15dc245f640dd24687143cb21fdedf34968f9de671aec463f8bbc91bec7ea88f
+size 9055

fi_ID/scaler.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ed4c1193f133be974168c0f1b02f520aa06f0e056b4f5dd7c8f952cc20a8675
+size 25191

fi_ID/training_details.txt

@@ -0,0 +1,39 @@
+Training Details for fi_ID
+========================================
+
+Language: fi
+Register: ID
+Training Date: 2025-09-26 14:08:45
+
+Data Summary:
+- Total samples: 34973
+- Training samples: 27978
+- Test samples: 6995
+- Embedding dimension: 1024
+
+Classes:
+- Number of classes: 2
+- Class names: '', 'sports'
+- Class distribution: {'': 34495, 'sports': 478}
+
+Cross-Validation Results:
+- CV folds: 5
+- CV scores: [0.9980343102215868, 0.9987491065046462, 0.9992852037169406, 0.9987488829311886, 0.9978552278820375]
+- CV mean: 0.9985
+- CV std: 0.0005
+- CV confidence interval: 0.9985 ± 0.0010
+
+Final Performance:
+- Test accuracy: 0.9986
+
+Model Configuration:
+- Algorithm: Logistic Regression
+- Regularization (C): 1.0
+- Feature scaling: StandardScaler
+- Random state: 42
+
+Files:
+- Classifier: model.pkl
+- Scaler: scaler.pkl
+- Metadata: metadata.pkl
+- This file: training_details.txt

fi_NA-nb-OP-rv/metadata.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1febda2186d53eefc99e24613b84c6746d3dfb4d9d938c736199e7a18c55884
+size 91

fi_NA-nb-OP-rv/model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83cfc695439d258effeb25e30a6677c1c0e18b8d07c3a4e1c36738eac4b800ac
+size 33671

fi_NA-nb-OP-rv/scaler.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7694b06950c31c271c8d8c2a03bf252b1b59b2dae800bcfd6ff31c1b4f296919
+size 25191

fi_NA-nb-OP-rv/training_details.txt

@@ -0,0 +1,39 @@
+Training Details for fi_NA-nb-OP-rv
+========================================
+
+Language: fi
+Register: NA-nb-OP-rv
+Training Date: 2025-09-26 14:06:58
+
+Data Summary:
+- Total samples: 4929
+- Training samples: 3943
+- Test samples: 986
+- Embedding dimension: 1024
+
+Classes:
+- Number of classes: 4
+- Class names: 'books', 'dining', 'beverages', 'cosmetics'
+- Class distribution: {'books': 2958, 'dining': 294, 'beverages': 283, 'cosmetics': 1394}
+
+Cross-Validation Results:
+- CV folds: 5
+- CV scores: [0.9898605830164765, 0.9936628643852978, 0.9936628643852978, 0.9885786802030457, 0.9911167512690355]
+- CV mean: 0.9914
+- CV std: 0.0020
+- CV confidence interval: 0.9914 ± 0.0041
+
+Final Performance:
+- Test accuracy: 0.9899
+
+Model Configuration:
+- Algorithm: Logistic Regression
+- Regularization (C): 1.0
+- Feature scaling: StandardScaler
+- Random state: 42
+
+Files:
+- Classifier: model.pkl
+- Scaler: scaler.pkl
+- Metadata: metadata.pkl
+- This file: training_details.txt

fi_NA-nb-OP/metadata.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8181ed2154a1e738d8731ba8b6559a50b8d98ff39db75d319023ec18339569f3
+size 77

fi_NA-nb-OP/model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f00c094692e5bb83931d54deef16ef5f7b807f0537d15a1ed6345acab21e88f
+size 25471

fi_NA-nb-OP/scaler.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5dac617f742be37b2fd9976ce2704013ecaa88a3eb69a0c6f1894edd5b2b6434
+size 25191

fi_NA-nb-OP/training_details.txt

@@ -0,0 +1,39 @@
+Training Details for fi_NA-nb-OP
+========================================
+
+Language: fi
+Register: NA-nb-OP
+Training Date: 2025-09-26 14:06:57
+
+Data Summary:
+- Total samples: 1969
+- Training samples: 1575
+- Test samples: 394
+- Embedding dimension: 1024
+
+Classes:
+- Number of classes: 3
+- Class names: '', 'culture', 'consumption'
+- Class distribution: {'': 874, 'culture': 695, 'consumption': 400}
+
+Cross-Validation Results:
+- CV folds: 5
+- CV scores: [0.9841269841269841, 0.9936507936507937, 0.9841269841269841, 0.9777777777777777, 0.9650793650793651]
+- CV mean: 0.9810
+- CV std: 0.0094
+- CV confidence interval: 0.9810 ± 0.0188
+
+Final Performance:
+- Test accuracy: 0.9848
+
+Model Configuration:
+- Algorithm: Logistic Regression
+- Regularization (C): 1.0
+- Feature scaling: StandardScaler
+- Random state: 42
+
+Files:
+- Classifier: model.pkl
+- Scaler: scaler.pkl
+- Metadata: metadata.pkl
+- This file: training_details.txt

fi_NA-nb/metadata.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:836ae4c831fe01dd10837795f65517e2768a1b1bd5364187226b09a39044826b
+size 64

fi_NA-nb/model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73957b464aecf86217f464d4db098528d7f7447904f721044e0a93987372e896
+size 9055

fi_NA-nb/scaler.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ee14f4b973f3e012477bb168eb62930d29c713bd94ef1f4eccef46f3b32fcca
+size 25191

fi_NA-nb/training_details.txt

@@ -0,0 +1,39 @@
+Training Details for fi_NA-nb
+========================================
+
+Language: fi
+Register: NA-nb
+Training Date: 2025-09-26 14:08:40
+
+Data Summary:
+- Total samples: 218088
+- Training samples: 174470
+- Test samples: 43618
+- Embedding dimension: 1024
+
+Classes:
+- Number of classes: 2
+- Class names: '', 'comments'
+- Class distribution: {'': 193558, 'comments': 24530}
+
+Cross-Validation Results:
+- CV folds: 5
+- CV scores: [0.9926061787126726, 0.9932939760417264, 0.9922622800481458, 0.9933512924858142, 0.992778128044936]
+- CV mean: 0.9929
+- CV std: 0.0004
+- CV confidence interval: 0.9929 ± 0.0008
+
+Final Performance:
+- Test accuracy: 0.9927
+
+Model Configuration:
+- Algorithm: Logistic Regression
+- Regularization (C): 1.0
+- Feature scaling: StandardScaler
+- Random state: 42
+
+Files:
+- Classifier: model.pkl
+- Scaler: scaler.pkl
+- Metadata: metadata.pkl
+- This file: training_details.txt

modeling_sm_subgroup_classifier.py

@@ -0,0 +1,138 @@
+import os
+import pickle
+
+import joblib
+import torch
+from transformers import PreTrainedModel
+
+from .configuration_sm_subgroup_classifier import SmSubgroupClassifierConfig
+
+
+class SmSubgroupClassifier(PreTrainedModel):
+    config_class = SmSubgroupClassifierConfig
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.config = config
+        self._loaded_classifiers = {}
+        self.model_dir = None
+        self._available_models = None
+
+    @property
+    def available_models(self):
+        """Auto-discover available models"""
+        if self._available_models is None:
+            self._available_models = self._discover_available_models()
+        return self._available_models
+
+    def _discover_available_models(self):
+        """Scan model directory for available models"""
+        if not self.model_dir or not os.path.exists(self.model_dir):
+            return []
+
+        models = []
+        for item in os.listdir(self.model_dir):
+            item_path = os.path.join(self.model_dir, item)
+            if os.path.isdir(item_path):
+                # Verify it's a valid model directory
+                required_files = ["model.pkl", "scaler.pkl", "metadata.pkl"]
+                if all(
+                    os.path.exists(os.path.join(item_path, f)) for f in required_files
+                ):
+                    models.append(item)
+
+        return sorted(models)
+
+    def _load_classifier(self, model_key):
+        """Load a specific classifier by model key (e.g., 'en_OP-ob')"""
+        if model_key in self._loaded_classifiers:
+            return self._loaded_classifiers[model_key]
+
+        if model_key not in self.available_models:
+            raise ValueError(
+                f"Model '{model_key}' not available. Available: {self.available_models}"
+            )
+
+        # Path to classifier
+        classifier_path = os.path.join(self.model_dir, model_key)
+        if not os.path.exists(classifier_path):
+            raise FileNotFoundError(f"Classifier not found at {classifier_path}")
+
+        # Load components
+        classifier = joblib.load(os.path.join(classifier_path, "model.pkl"))
+        scaler = joblib.load(os.path.join(classifier_path, "scaler.pkl"))
+        with open(os.path.join(classifier_path, "metadata.pkl"), "rb") as f:
+            metadata = pickle.load(f)
+
+        classifier_info = {
+            "classifier": classifier,
+            "scaler": scaler,
+            "class_names": metadata["class_names"],
+        }
+
+        self._loaded_classifiers[model_key] = classifier_info
+        return classifier_info
+
+    def forward(self, language, model_name, embeddings):
+        """
+        Args:
+            language: Language code (en, fi, sv)
+            model_name: Model name (OP-ob, NA, etc.)
+            embeddings: Pre-computed embeddings
+        """
+        # Create model key
+        model_key = f"{language}_{model_name}"
+
+        # Convert embeddings to numpy if needed
+        if torch.is_tensor(embeddings):
+            embeddings = embeddings.detach().cpu().numpy()
+        if embeddings.ndim == 1:
+            embeddings = embeddings.reshape(1, -1)
+
+        # Load classifier
+        classifier_info = self._load_classifier(model_key)
+
+        # Scale and predict
+        embeddings_scaled = classifier_info["scaler"].transform(embeddings)
+        predictions = classifier_info["classifier"].predict(embeddings_scaled)
+        probabilities = classifier_info["classifier"].predict_proba(embeddings_scaled)
+
+        # Format results - just use class names and probabilities
+        results = []
+        for pred, probs in zip(predictions, probabilities):
+            predicted_class_name = classifier_info["class_names"][pred]
+            # Get all class probabilities
+            all_probs = {
+                classifier_info["class_names"][i]: float(prob)
+                for i, prob in enumerate(probs)
+            }
+            results.append(
+                {
+                    "label": predicted_class_name,
+                    "probabilities": all_probs,
+                }
+            )
+
+        return results[0] if len(results) == 1 else results
+
+    @classmethod
+    def from_pretrained(cls, pretrained_model_name_or_path, **kwargs):
+        # Load config
+        config = SmSubgroupClassifierConfig.from_pretrained(
+            pretrained_model_name_or_path, **kwargs
+        )
+
+        # Create model instance (skip the pytorch weight loading)
+        model = cls(config)
+
+        # For HF Hub, we need to resolve to the actual cached directory
+        try:
+            from huggingface_hub import snapshot_download
+
+            # Download/get the cached directory path
+            model.model_dir = snapshot_download(pretrained_model_name_or_path)
+        except ImportError:
+            # Fallback if huggingface_hub not available
+            model.model_dir = pretrained_model_name_or_path
+
+        return model

sv_HI-NA-nb-re/metadata.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a0f52bdc3d0252e9176382affb050c74fa24ed744785573a67ae28cdb9d785f
+size 62

sv_HI-NA-nb-re/model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45b36afa0b2d6b74474f7ec1ae3779322e637f64af1bfe522083fe356a22a194
+size 9055

sv_HI-NA-nb-re/scaler.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7732e98d25e09a32de626c8996637f3e844deb95f1495a77e72cd6cc957da2f4
+size 25191

sv_HI-NA-nb-re/training_details.txt

@@ -0,0 +1,39 @@
+Training Details for sv_HI-NA-nb-re
+========================================
+
+Language: sv
+Register: HI-NA-nb-re
+Training Date: 2025-09-26 14:05:16
+
+Data Summary:
+- Total samples: 6804
+- Training samples: 5443
+- Test samples: 1361
+- Embedding dimension: 1024
+
+Classes:
+- Number of classes: 2
+- Class names: 'crafts', ''
+- Class distribution: {'crafts': 135, '': 6669}
+
+Cross-Validation Results:
+- CV folds: 5
+- CV scores: [0.9963269054178145, 0.9963269054178145, 0.9972451790633609, 0.9944852941176471, 0.9954044117647058]
+- CV mean: 0.9960
+- CV std: 0.0009
+- CV confidence interval: 0.9960 ± 0.0019
+
+Final Performance:
+- Test accuracy: 0.9956
+
+Model Configuration:
+- Algorithm: Logistic Regression
+- Regularization (C): 1.0
+- Feature scaling: StandardScaler
+- Random state: 42
+
+Files:
+- Classifier: model.pkl
+- Scaler: scaler.pkl
+- Metadata: metadata.pkl
+- This file: training_details.txt

sv_ID-NA-nb/metadata.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:836ae4c831fe01dd10837795f65517e2768a1b1bd5364187226b09a39044826b
+size 64

sv_ID-NA-nb/model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3e6a7f1c0e31e315d84e43bc75299c48fe02215b7def50de7fc350316f807cb
+size 9055

sv_ID-NA-nb/scaler.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0321f17c25e743d5a5318c5011f9b6bdbe864aa683f9ff00d7bc2a6efc28cbb
+size 25191

sv_ID-NA-nb/training_details.txt

@@ -0,0 +1,39 @@
+Training Details for sv_ID-NA-nb
+========================================
+
+Language: sv
+Register: ID-NA-nb
+Training Date: 2025-09-26 14:05:20
+
+Data Summary:
+- Total samples: 1189
+- Training samples: 951
+- Test samples: 238
+- Embedding dimension: 1024
+
+Classes:
+- Number of classes: 2
+- Class names: '', 'comments'
+- Class distribution: {'': 852, 'comments': 337}
+
+Cross-Validation Results:
+- CV folds: 5
+- CV scores: [0.9214659685863874, 0.9526315789473684, 0.9315789473684211, 0.968421052631579, 0.9263157894736842]
+- CV mean: 0.9401
+- CV std: 0.0177
+- CV confidence interval: 0.9401 ± 0.0354
+
+Final Performance:
+- Test accuracy: 0.9706
+
+Model Configuration:
+- Algorithm: Logistic Regression
+- Regularization (C): 1.0
+- Feature scaling: StandardScaler
+- Random state: 42
+
+Files:
+- Classifier: model.pkl
+- Scaler: scaler.pkl
+- Metadata: metadata.pkl
+- This file: training_details.txt

sv_ID/metadata.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:21413604878beb792fbc2f549db7a71b67cfaaca8200989e74e5df69a0e6c7b1
+size 69

sv_ID/model.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f22126201e503512b9bbf51ed4c310d6e59ae89dd12e37835154c15a382d2612
+size 25471

sv_ID/scaler.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f5634d9ac1f9b6a3d8e303ca1a929f6ade9ac33f1acdf697ae765a60646fd66
+size 25191

sv_ID/training_details.txt

@@ -0,0 +1,39 @@
+Training Details for sv_ID
+========================================
+
+Language: sv
+Register: ID
+Training Date: 2025-09-26 14:06:57
+
+Data Summary:
+- Total samples: 15060
+- Training samples: 12048
+- Test samples: 3012
+- Embedding dimension: 1024
+
+Classes:
+- Number of classes: 3
+- Class names: '', 'sports', 'help'
+- Class distribution: {'': 14108, 'sports': 542, 'help': 410}
+
+Cross-Validation Results:
+- CV folds: 5
+- CV scores: [0.9962655601659751, 0.9954356846473029, 0.9983402489626556, 1.0, 0.9991697799916978]
+- CV mean: 0.9978
+- CV std: 0.0017
+- CV confidence interval: 0.9978 ± 0.0035
+
+Final Performance:
+- Test accuracy: 0.9977
+
+Model Configuration:
+- Algorithm: Logistic Regression
+- Regularization (C): 1.0
+- Feature scaling: StandardScaler
+- Random state: 42
+
+Files:
+- Classifier: model.pkl
+- Scaler: scaler.pkl
+- Metadata: metadata.pkl
+- This file: training_details.txt