scikit-learn
diff --git a/‎doc/whats_new/v0.23.rst
Lines changed: 2 additions & 1 deletion b/‎doc/whats_new/v0.23.rst
Lines changed: 2 additions & 1 deletion
diff --git a/‎sklearn/svm/_base.py
Lines changed: 7 additions & 12 deletions b/‎sklearn/svm/_base.py
Lines changed: 7 additions & 12 deletions
diff --git a/‎sklearn/svm/tests/test_svm.py
Lines changed: 6 additions & 8 deletions b/‎sklearn/svm/tests/test_svm.py
Lines changed: 6 additions & 8 deletions
@@ -23,7 +23,8 @@ parameters, may produce different models from the previous version. This often
 occurs due to changes in the modelling logic (bug fixes or enhancements), or in
 random sampling procedures.
 
-- models come here
+- :class:`svm.SVC` and :class:`svm.SVR` when `kernel` is callable and input is
+  not castable as float array. |Fix|
 
 Details are listed in the changelog below.
 
 
@@ -177,12 +177,12 @@ def fit(self, X, y, sample_weight=None):
                              "boolean masks (use `indices=True` in CV)."
                              % (sample_weight.shape, X.shape))
 
-        kernel = self.kernel
-        if callable(kernel):
-            kernel = 'precomputed'
+        kernel = 'precomputed' if callable(self.kernel) else self.kernel
 
         if kernel == 'precomputed':
-            self._gamma = 0.  # unused but needs to be a float
+            # unused but needs to be a float for cython code that ignores
+            # it anyway
+            self._gamma = 0.
         elif isinstance(self.gamma, str):
             if self.gamma == 'scale':
                 # var = E[X^2] - E[X]^2 if sparse
@@ -207,10 +207,7 @@ def fit(self, X, y, sample_weight=None):
         fit(X, y, sample_weight, solver_type, kernel, random_seed=seed)
         # see comment on the other call to np.iinfo in this file
 
-        if hasattr(X, 'shape'):
-            self.shape_fit_ = X.shape
-        else:
-            self.shape_fit_ = (_num_samples(X),)
+        self.shape_fit_ = X.shape if hasattr(X, "shape") else (n_samples, )
 
         # In binary case, we need to flip the sign of coef, intercept and
         # decision function. Use self._intercept_ and self._dual_coef_
@@ -467,18 +464,16 @@ def _validate_for_predict(self, X):
             raise ValueError(
                 "cannot use sparse input in %r trained on dense data"
                 % type(self).__name__)
-        if not callable(self.kernel):
-            n_features = X.shape[1]
 
         if self.kernel == "precomputed":
             if X.shape[1] != self.shape_fit_[0]:
                 raise ValueError("X.shape[1] = %d should be equal to %d, "
                                  "the number of samples at training time" %
                                  (X.shape[1], self.shape_fit_[0]))
-        elif not callable(self.kernel) and n_features != self.shape_fit_[1]:
+        elif not callable(self.kernel) and X.shape[1] != self.shape_fit_[1]:
             raise ValueError("X.shape[1] = %d should be equal to %d, "
                              "the number of features at training time" %
-                             (n_features, self.shape_fit_[1]))
+                             (X.shape[1], self.shape_fit_[1]))
         return X
 
     @property
 
@@ -1281,14 +1281,12 @@ def string_kernel(X1, X2):
     assert svc1.score(data, y) == svc3.score(K, y)
     assert svc1.score(data, y) == svc2.score(X, y)
     if hasattr(svc1, 'decision_function'):  # classifier
-        assert_array_almost_equal(svc1.decision_function(data),
-                                  svc2.decision_function(X))
-        assert_array_almost_equal(svc1.decision_function(data),
-                                  svc3.decision_function(K))
+        assert_allclose(svc1.decision_function(data),
+                        svc2.decision_function(X))
+        assert_allclose(svc1.decision_function(data),
+                        svc3.decision_function(K))
         assert_array_equal(svc1.predict(data), svc2.predict(X))
         assert_array_equal(svc1.predict(data), svc3.predict(K))
     else:  # regressor
-        assert_array_almost_equal(svc1.predict(data),
-                                  svc2.predict(X))
-        assert_array_almost_equal(svc1.predict(data),
-                                  svc3.predict(K))
+        assert_allclose(svc1.predict(data), svc2.predict(X))
+        assert_allclose(svc1.predict(data), svc3.predict(K))