Upload scikit-learn/SCIKIT-LEARN_USER_GUIDE.txt with huggingface_hub

scikit-learn/SCIKIT-LEARN_USER_GUIDE.txt

@@ -0,0 +1,106 @@
+================================================================================
+  SCIKIT-LEARN - USER GUIDE (Android Python STB) - Generated by RIMI
+================================================================================
+
+What is scikit-learn?
+---------------------
+scikit-learn is a Python module for machine learning built on top of SciPy
+(numpy, scipy) and distributed under BSD-3-Clause. It provides simple and
+efficient tools for predictive data analysis: classification, regression,
+clustering, dimensionality reduction, model selection, preprocessing.
+
+This wheel is cross-compiled for Android (cp312, android_24_x86_64 /
+android_24_arm64_v8a, 16KB page, NoRELRO, single libc++_shared.so).
+
+Installation
+------------
+Via PipManager in app or via Install_WHEELS_to_DEVICE.ps1 (option 1 ALL or 2 single):
+  pip install scikit_learn-1.7.1-cp312-cp312-android_24_x86_64.whl
+
+Requires at runtime (all in WHEELS, pure+native):
+  numpy>=1.22.0 (6.4M android_24_x86_64, 16KB)
+  scipy>=1.8.0 (31M android_24_x86_64, 16KB, no C: RUNPATH)
+  joblib>=1.2.0 (0.29M pure)
+  threadpoolctl>=3.1.0 (0.02M pure)
+
+No extra build deps needed at runtime (cython, meson-python only for building).
+
+Quick Start (Android)
+---------------------
+import sklearn
+print(sklearn.__version__)  # 1.7.1
+
+from sklearn.datasets import load_iris
+X, y = load_iris(return_X_y=True)  # X shape (150,4)
+
+from sklearn.preprocessing import StandardScaler
+scaler = StandardScaler()
+Xt = scaler.fit_transform(X)
+
+from sklearn.decomposition import PCA
+pca = PCA(n_components=2)
+Xt2 = pca.fit_transform(X)  # shape (150,2)
+
+from sklearn.cluster import KMeans
+km = KMeans(n_clusters=3, n_init=10, random_state=42)
+labels = km.fit_predict(X)
+
+from sklearn.linear_model import LogisticRegression
+clf = LogisticRegression(max_iter=200)
+clf.fit(X[:100], y[:100])
+pred = clf.predict(X[:5])
+
+from sklearn.ensemble import RandomForestClassifier
+clf = RandomForestClassifier(n_estimators=10, random_state=42)
+clf.fit(X[:100], y[:100])
+
+from sklearn.metrics import accuracy_score
+acc = accuracy_score(y_true, y_pred)
+
+from sklearn.model_selection import train_test_split, cross_val_score
+Xtr, Xte, ytr, yte = train_test_split(X, y, test_size=0.25, random_state=42)
+scores = cross_val_score(clf, X, y, cv=3)
+
+API Gotchas (Android)
+---------------------
+- All 69 Cython extensions are 16KB page aligned (LOAD 0x4000, no GNU_RELRO,
+  NEEDED libc++_shared.so only, no C: RUNPATH). Verified via llvm-readelf.
+- sklearn imports numpy/scipy/joblib/threadpoolctl — ensure those wheels
+  are installed first (Install_WHEELS_to_DEVICE.ps1 does it in sorted order).
+- No tkinter/matplotlib required for core; plotting via matplotlib is optional
+  (extra, not needed for ML).
+- For XGBoost sklearn interface (XGBRegressor/XGBClassifier), sklearn must be
+  installed (now in WHEELS 1.7.1, 7-8M).
+
+Tested Features (Test_ScikitLearn.py)
+-------------------------------------
+- import sklearn, C extension _tree
+- load_iris, load_digits
+- StandardScaler, MinMaxScaler
+- PCA (10->2)
+- KMeans (3 clusters)
+- LogisticRegression, LinearRegression
+- RandomForestClassifier (10 trees)
+- accuracy_score, mean_squared_error
+- train_test_split, cross_val_score (3-fold)
+
+All 15 tests PASS, 0 SKIP, 0 FAIL on Android x86_64 emulator (Python 3.12.14).
+
+Platform Notes (Android)
+------------------------
+- Arch: x86_64 (emulator) and arm64-v8a (phone) — both 16KB, NoRELRO
+- NDK r26, clang 17.0.2, LLD, -z norelro -z max-page-size=16384
+- Links only to /data/app/.../lib/x86_64/libc++_shared.so (custom 1.3M, not NDK 4K)
+  + librimi.so + libpython3.12.so.1.0 — no libc++.so/libc++abi.so/libunwind.
+- No Windows C:/ paths in any .so (strings check 0).
+
+Troubleshooting
+---------------
+- ImportError: No module named 'sklearn' → install wheel via PipManager
+- ImportError: libomp.so → already handled via threadpoolctl, no extra
+- If Test_ScikitLearn shows SKIP for missing joblib/threadpoolctl, install
+  those pure wheels first (they are in WHEELS).
+
+================================================================================
+  Generated by RIMI
+================================================================================