Merge pull request #194 from brandonwillard/update-warnings

tests/gof/test_utils.py

 import theano
-from theano.gof.utils import give_variables_names, remove, unique
-
-
-def test_give_variables_names():
-    x = theano.tensor.matrix("x")
-    y = x + 1
-    z = theano.tensor.dot(x, y)
-    variables = (x, y, z)
-    give_variables_names(variables)
-    assert all(var.name for var in variables)
-    assert unique([var.name for var in variables])
-
-
-def test_give_variables_names_idempotence():
-    x = theano.tensor.matrix("x")
-    y = x + 1
-    z = theano.tensor.dot(x, y)
-    variables = (x, y, z)
-
-    give_variables_names(variables)
-    names = [var.name for var in variables]
-    give_variables_names(variables)
-    names2 = [var.name for var in variables]
-
-    assert names == names2
-
-
-def test_give_variables_names_small():
-    x = theano.tensor.matrix("x")
-    y = theano.tensor.dot(x, x)
-    fgraph = theano.FunctionGraph((x,), (y,))
-    give_variables_names(fgraph.variables)
-    assert all(var.name for var in fgraph.variables)
-    assert unique([var.name for var in fgraph.variables])
+from theano.gof.utils import remove
 
 
 def test_remove():

tests/tensor/test_basic.py

@@ -4114,7 +4114,7 @@ def test_make_column_matrix_broadcastable():
     assert (f(np.zeros((3, 1))) + np.ones(2) == np.ones((3, 2))).all()
 
 
-def test_flatten_outdimNone():
+def test_flatten_ndim_default():
     a = dmatrix()
     c = flatten(a)
     f = inplace_func([a], c)
@@ -4178,9 +4178,6 @@ def test_flatten_ndim2_of_3():
 
     flatten_2 = partial(flatten, ndim=2)
     utt.verify_grad(flatten_2, [a_val])
-    # test outdim parameter name
-    flatten_2 = partial(flatten, outdim=2)
-    utt.verify_grad(flatten_2, [a_val])
 
 
 def test_flatten_broadcastable():
@@ -4219,7 +4216,7 @@ def test_flatten_ndim_invalid():
 def test_is_flat():
     # tests is_flat method for constant and symbolic variables,
     # as well as reshaped constant and symbolic variables on the
-    # given outdim
+    # given `ndim`
 
     # Constant variable
     assert tt.is_flat(tt.as_tensor_variable(np.zeros(10)))
@@ -6251,10 +6248,10 @@ class TestInferShape(utt.InferShapeTester):
         # Flatten
         atens3 = tensor3()
         atens3_val = rand(4, 5, 3)
-        for outdim in (3, 2, 1):
+        for ndim in (3, 2, 1):
             self._compile_and_check(
                 [atens3],
-                [flatten(atens3, outdim)],
+                [flatten(atens3, ndim)],
                 [atens3_val],
                 Reshape,
                 excluding=["local_useless_reshape"],
@@ -6262,10 +6259,10 @@ class TestInferShape(utt.InferShapeTester):
 
         amat = matrix()
         amat_val = rand(4, 5)
-        for outdim in (2, 1):
+        for ndim in (2, 1):
             self._compile_and_check(
                 [amat],
-                [flatten(amat, outdim)],
+                [flatten(amat, ndim)],
                 [amat_val],
                 Reshape,
                 excluding=["local_useless_reshape"],
@@ -6273,10 +6270,10 @@ class TestInferShape(utt.InferShapeTester):
 
         avec = vector()
         avec_val = rand(4)
-        outdim = 1
+        ndim = 1
         self._compile_and_check(
             [avec],
-            [flatten(avec, outdim)],
+            [flatten(avec, ndim)],
             [avec_val],
             Reshape,
             excluding=["local_useless_reshape"],

tests/tensor/test_blas_c.py

-import sys
+from warnings import warn
 
 import numpy as np
 import pytest
@@ -197,7 +197,7 @@ class TestCGemv(OptimizationTestMixin):
 
     def test_force_gemv_init(self):
         if check_force_gemv_init():
-            sys.stderr.write(
+            warn(
                 "WARNING: The current BLAS requires Theano to initialize"
                 + " memory for some GEMV calls which will result in a minor"
                 + " degradation in performance for such calls."

theano/compile/debugmode.py

@@ -13,6 +13,7 @@ import sys
 from io import StringIO
 from itertools import chain
 from itertools import product as itertools_product
+from warnings import warn
 
 import numpy as np
 
@@ -934,7 +935,7 @@ def _check_strides_match(a, b, warn_err, op):
         if warn_err == 2:
             raise e
         else:
-            print("WARNING:", e, file=sys.stderr)
+            warn(str(e))
 
 
 def _lessbroken_deepcopy(a):
@@ -2459,7 +2460,7 @@ class _Maker(FunctionMaker):  # inheritance buys a few helper functions
                 l0 = fgraph0.equivalence_tracker.event_list
                 if li != l0:
                     infolog = StringIO()
-                    print("WARNING: Optimization process is unstable...", file=infolog)
+                    print("Optimization process is unstable...", file=infolog)
                     print(
                         "  (HINT: Ops that the nodes point to must compare " "equal)",
                         file=infolog,

theano/compile/mode.py

@@ -123,7 +123,7 @@ class AddDestroyHandler(gof.GlobalOptimizer):
                 break
         if not supervisor_added:
             warnings.warn(
-                "WARNING: Supervisor is not added. Please build a FunctionGraph"
+                "Supervisor is not added. Please build a FunctionGraph"
                 "via theano.compile.function.types.std_graph()"
                 "or add the Supervisor class manually.",
                 stacklevel=3,

theano/gof/cmodule.py

@@ -2014,7 +2014,7 @@ class GCC_compiler(Compiler):
             and "icpc" not in theano.config.cxx
         ):
             _logger.warning(
-                "OPTIMIZATION WARNING: your Theano flag `cxx` seems not to be"
+                "Your Theano flag `cxx` seems not to be"
                 " the g++ compiler. So we disable the compiler optimization"
                 " specific to g++ that tell to compile for a specific CPU."
                 " At worst, this could cause slow down.\n"
@@ -2083,7 +2083,7 @@ class GCC_compiler(Compiler):
                 else:
                     reported_lines = native_lines
                 _logger.warning(
-                    "OPTIMIZATION WARNING: Theano was not able to find the"
+                    "Theano was not able to find the"
                     " g++ parameters that tune the compilation to your "
                     " specific CPU. This can slow down the execution of Theano"
                     " functions. Please submit the following lines to"
@@ -2095,7 +2095,7 @@ class GCC_compiler(Compiler):
                 _logger.info(f"g++ default lines: {default_lines}")
                 if len(default_lines) < 1:
                     _logger.warning(
-                        "OPTIMIZATION WARNING: Theano was not able to find the"
+                        "Theano was not able to find the"
                         " default g++ parameters. This is needed to tune"
                         " the compilation to your specific"
                         " CPU. This can slow down the execution of Theano"

theano/gof/link.py

@@ -3,6 +3,7 @@ import traceback
 from copy import copy, deepcopy
 from io import StringIO
 from sys import getsizeof
+from warnings import warn
 
 import numpy as np
 
@@ -334,7 +335,7 @@ def raise_with_op(fgraph, node, thunk=None, exc_info=None, storage_map=None):
             str(exc_value) + detailed_err_msg + "\n" + "\n".join(hints)
         )
     except TypeError:
-        print(f"WARNING: {exc_type} error does not allow us to add extra error message")
+        warn(f"{exc_type} error does not allow us to add extra error message")
         # Some exception need extra parameter in inputs. So forget the
         # extra long error message in that case.
     raise exc_value.with_traceback(exc_trace)

theano/gof/toolbox.py

@@ -610,17 +610,15 @@ class ReplaceValidate(History, Validator):
             if rm in fgraph.apply_nodes or rm in fgraph.variables:
                 fgraph.revert(chk)
                 if warn:
-                    out = sys.stderr
-                    print(
-                        "WARNING: An optimization wanted to replace a Variable"
+                    warn(
+                        "An optimization wanted to replace a Variable"
                         " in the graph, but the replacement for it doesn't"
                         " remove it. We disabled the optimization."
                         " Your function runs correctly, but it would be"
                         " appreciated if you submit this problem to the"
-                        " mailing list theano-users so that we can fix it.",
-                        file=out,
+                        " mailing list theano-users so that we can fix it."
+                        f"{reason}: {replacements}",
                     )
-                    print(reason, replacements, file=out)
                 raise ReplacementDidNotRemoveError()
 
     def __getstate__(self):

theano/gof/utils.py

@@ -376,36 +376,6 @@ def memoize(f):
     return rval
 
 
-def deprecated(filename, msg=""):
-    """
-    Decorator which will print a warning message on the first call.
-
-    Use it like this::
-
-      @deprecated('myfile', 'do something different...')
-      def fn_name(...)
-          ...
-
-    And it will print::
-
-      WARNING myfile.fn_name deprecated. do something different...
-
-    """
-
-    def _deprecated(f):
-        printme = [True]
-
-        def g(*args, **kwargs):
-            if printme[0]:
-                print(f"WARNING: {filename}.{f.__name__} deprecated. {msg}")
-                printme[0] = False
-            return f(*args, **kwargs)
-
-        return g
-
-    return _deprecated
-
-
 def uniq(seq):
     """
     Do not use set, this must always return the same value at the same index.
@@ -617,10 +587,6 @@ def flatten(a):
         return [a]
 
 
-def unique(x):
-    return len(set(x)) == len(x)
-
-
 def hist(coll):
     counts = {}
     for elem in coll:
@@ -628,31 +594,6 @@ def hist(coll):
     return counts
 
 
-@deprecated("theano.gof.utils", msg="Use a_theano_variable.auto_name instead")
-def give_variables_names(variables):
-    """
-    Gives unique names to an iterable of variables. Modifies input.
-
-    This function is idempotent.
-
-    """
-    names = [var.name for var in variables]
-    h = hist(names)
-
-    def bad_var(var):
-        return not var.name or h[var.name] > 1
-
-    for i, var in enumerate(filter(bad_var, variables)):
-        var.name = (var.name or "") + f"_{int(i)}"
-
-    if not unique([str(v) for v in variables]):
-        raise ValueError(
-            "Not all variables have unique names. Maybe you've "
-            "named some of the variables identically"
-        )
-    return variables
-
-
 def remove(predicate, coll):
     """
     Return those items of collection for which predicate(item) is true.

theano/gpuarray/__init__.py

@@ -24,11 +24,21 @@ try:
 except ImportError:
     pygpu = None
 
-from . import ctc, dnn, extra_ops, fft, multinomial, opt, reduction, rng_mrg, sort
-from .basic_ops import as_gpuarray_variable
+from theano.gpuarray import (
+    ctc,
+    dnn,
+    extra_ops,
+    fft,
+    multinomial,
+    opt,
+    reduction,
+    rng_mrg,
+    sort,
+)
+from theano.gpuarray.basic_ops import as_gpuarray_variable
 
 # This is for documentation not to depend on the availability of pygpu
-from .type import (
+from theano.gpuarray.type import (
     ContextNotDefined,
     GpuArrayConstant,
     GpuArraySharedVariable,
@@ -160,8 +170,8 @@ def init_dev(dev, name=None, preallocate=None):
                     f" {int(context.free_gmem / MB)} MB are available."
                 )
             elif gmem > context.free_gmem - 50 * MB:
-                print(
-                    "WARNING: Preallocating too much memory can prevent cudnn and cublas from working properly"
+                warnings.warn(
+                    "Preallocating too much memory can prevent cudnn and cublas from working properly"
                 )
 
             # This will allocate and immediately free an object of size gmem

theano/gradient.py

@@ -2157,6 +2157,7 @@ def consider_constant(x):
             "consider_constant() is deprecated, use zero_grad() or "
             "disconnected_grad() instead."
         ),
+        category=DeprecationWarning,
         stacklevel=3,
     )
 

theano/sandbox/blocksparse.py

-import warnings
-
-from theano.tensor.nnet.blocksparse import (
-    SparseBlockGemv,
-    SparseBlockOuter,
-    sparse_block_dot,
-    sparse_block_gemv,
-    sparse_block_gemv_inplace,
-    sparse_block_outer,
-    sparse_block_outer_inplace,
-)
-
-
-__all__ = [
-    SparseBlockGemv,
-    SparseBlockOuter,
-    sparse_block_dot,
-    sparse_block_gemv,
-    sparse_block_gemv_inplace,
-    sparse_block_outer,
-    sparse_block_outer_inplace,
-]
-
-warnings.warn(
-    "DEPRECATION: theano.sandbox.blocksparse does not exist anymore,"
-    "it has been moved to theano.tensor.nnet.blocksparse.",
-    category=DeprecationWarning,
-)

theano/sandbox/conv.py

-import sys
-
-
-print(
-    "DEPRECATION: theano.sandbox.conv no longer provides conv. "
-    "They have been moved to theano.tensor.nnet.conv",
-    file=sys.stderr,
-)

theano/sandbox/softsign.py

-from theano.tensor.nnet.nnet import softsign  # noqa
-import sys
-
-
-print(
-    "DEPRECATION WARNING: softsign was moved from theano.sandbox.softsign to "
-    "theano.tensor.nnet.nnet ",
-    file=sys.stderr,
-)

theano/scan/basic.py

@@ -377,9 +377,8 @@ def scan(
     for i in range(n_outs):
         if outs_info[i] is not None:
             if isinstance(outs_info[i], dict):
-                # DEPRECATED :
                 if outs_info[i].get("return_steps", None) is not None:
-                    raise ValueError(
+                    raise DeprecationWarning(
                         "Using `return_steps` has been deprecated. "
                         "Simply select the entries you need using a "
                         "subtensor. Scan will optimize memory "
@@ -396,12 +395,9 @@ def scan(
             ):
                 # ^ no initial state but taps provided
                 raise ValueError(
-                    (
                     "If you are using slices of an output "
                     "you need to provide a initial state "
-                        "for it"
-                    ),
-                    outs_info[i],
+                    f"for it: {outs_info[i]}"
                 )
             elif (
                 outs_info[i].get("initial", None) is not None

theano/scan/opt.py

@@ -101,14 +101,6 @@ list_opt_slice = [
 ]
 
 
-def warning(*msg):
-    _logger.warning("WARNING theano.scan: " + " ".join(msg))
-
-
-def info(*msg):
-    _logger.info("INFO theano.scan: " + " ".join(msg))
-
-
 @gof.local_optimizer([Scan])
 def remove_constants_and_unused_inputs_scan(fgraph, node):
     """

theano/scan/utils.py

@@ -173,12 +173,7 @@ def hash_listsDictsTuples(x):
     return hash_value
 
 
-DEPRECATED_ARG = object()
-
-
-def clone(
-    output, replace=None, strict=True, share_inputs=True, copy_inputs=DEPRECATED_ARG
-):
+def clone(output, replace=None, strict=True, share_inputs=True):
     """
     Function that allows replacing subgraphs of a computational graph.
 
@@ -196,17 +191,8 @@ def clone(
         graph. If False, clone them. Note that cloned shared variables still
         use the same underlying storage, so they will always have the same
         value.
-    copy_inputs
-        Deprecated, use share_inputs.
 
     """
-    if copy_inputs is not DEPRECATED_ARG:
-        warnings.warn(
-            "In `clone()` function, the argument `copy_inputs` has been deprecated and renamed into `share_inputs`"
-        )
-        assert share_inputs  # since we used `copy_inputs` we should have default value for `share_inputs`
-        share_inputs = copy_inputs
-
     if isinstance(replace, dict):
         items = list(replace.items())
     elif isinstance(replace, (list, tuple)):

theano/sparse/__init__.py

-import sys
+from warnings import warn
 
 
 try:
     import scipy
 
-    scipy_ver = [int(n) for n in scipy.__version__.split(".")[:2]]
-    enable_sparse = bool(scipy_ver >= [0, 7])
-    if not enable_sparse:
-        sys.stderr.write(
-            f"WARNING: scipy version = {scipy.__version__}."
-            " We request version >=0.7.0 for the sparse code as it has"
-            " bugs fixed in the sparse matrix code.\n"
-        )
+    enable_sparse = True
 except ImportError:
     enable_sparse = False
-    sys.stderr.write(
-        "WARNING: scipy can't be imported." " We disable the sparse matrix code."
-    )
+    warn("SciPy can't be imported.  Sparse matrix support is disabled.")
 
 from theano.sparse.type import *
 

theano/sparse/basic.py

@@ -3,13 +3,11 @@ Classes for handling sparse matrices.
 
 To read about different sparse formats, see
 http://www-users.cs.umn.edu/~saad/software/SPARSKIT/paper.ps
-"""
-
 
-# TODO
-# Automatic methods for determining best sparse format?
+TODO: Automatic methods for determining best sparse format?
 
-import sys
+"""
+from warnings import warn
 
 import numpy as np
 import scipy.sparse
@@ -907,9 +905,8 @@ class DenseFromSparse(gof.op.Op):
         (x,) = inputs
         (out,) = outputs
         if _is_dense(x):
-            print(
-                ("WARNING: You just called DenseFromSparse on a dense matrix."),
-                file=sys.stderr,
+            warn(
+                "You just called DenseFromSparse on a dense matrix.",
             )
             out[0] = x
         else:
@@ -2398,10 +2395,8 @@ class MulSD(gof.op.Op):
                         z_data[j_idx] *= y[i, j]
                 out[0] = z
             else:
-                print(
-                    ("WARNING: crappy implementation of MulSD"),
-                    x.format,
-                    file=sys.stderr,
+                warn(
+                    "This implementation of MulSD is deficient: {x.format}",
                 )
                 out[0] = type(x)(x.toarray() * y)
 

theano/tensor/basic.py

@@ -5244,8 +5244,8 @@ class Flatten(Op):
     """
     Flatten a tensor.
 
-    Flattens a tensor to `outdim` dimensions by preserving the leading
-    outdim - 1 shape components.
+    Flattens a tensor to `ndim` dimensions by preserving the leading
+    ndim - 1 shape components.
 
     .. note:: The interface Flatten(Op) is deprecated, you should use flatten.
     """
@@ -5253,24 +5253,24 @@ class Flatten(Op):
     view_map = {0: [0]}
 
     check_input = False
-    __props__ = ("outdim",)
+    __props__ = ("ndim",)
 
-    def __init__(self, outdim=1):
+    def __init__(self, ndim=1):
         warnings.warn(
             "Flatten class is deprecated, " "please use flatten method instead.",
             DeprecationWarning,
             stacklevel=4,
         )
-        self.outdim = int(outdim)
+        self.ndim = int(ndim)
 
     def __str__(self):
-        return f"{self.__class__.__name__}{{{self.outdim}}}"
+        return f"{self.__class__.__name__}{{{self.ndim}}}"
 
     def make_node(self, x):
         t_x = as_tensor_variable(x)
-        if self.outdim < 1 or (x.ndim and self.outdim > x.ndim):
+        if self.ndim < 1 or (x.ndim and self.ndim > x.ndim):
             raise ValueError(
-                f"invalid output ndimensions ({self.outdim}) for tensor of "
+                f"invalid output ndimensions ({self.ndim}) for tensor of "
                 f"rank {t_x.ndim}"
             )
 
@@ -5279,8 +5279,8 @@ class Flatten(Op):
         # For the dimension resulting from the collapse of other dimensions,
         # it should be broadcastable iff all the collapsed dimensions were
         # broadcastable.
-        bcast_kept_dims = x.broadcastable[: self.outdim - 1]
-        bcast_new_dim = builtins.all(x.broadcastable[self.outdim - 1 :])
+        bcast_kept_dims = x.broadcastable[: self.ndim - 1]
+        bcast_new_dim = builtins.all(x.broadcastable[self.ndim - 1 :])
         broadcastable = bcast_kept_dims + (bcast_new_dim,)
 
         return gof.Apply(self, [t_x], [tensor(x.type.dtype, broadcastable)])
@@ -5288,22 +5288,22 @@ class Flatten(Op):
     def perform(self, node, inp, out_):
         (x,) = inp
         (out,) = out_
-        outdim = self.outdim
-        if outdim == 1:
+        ndim = self.ndim
+        if ndim == 1:
             try:
                 out[0] = x.reshape(x.size)
             except AttributeError:
                 out[0] = x.reshape((np.prod(x.shape),))
-        elif outdim == len(x.shape):
+        elif ndim == len(x.shape):
             out[0] = x
         else:
-            newshape = x.shape[: outdim - 1] + (np.prod(x.shape[outdim - 1 :]),)
+            newshape = x.shape[: ndim - 1] + (np.prod(x.shape[ndim - 1 :]),)
             out[0] = x.reshape(newshape)
 
     def infer_shape(self, fgraph, node, in_shapes):
         (in_shp,) = in_shapes
-        part1 = in_shp[: self.outdim - 1]
-        part2 = in_shp[self.outdim - 1 :]
+        part1 = in_shp[: self.ndim - 1]
+        part2 = in_shp[self.ndim - 1 :]
 
         if len(part2) > 1:
             part2 = (prod(part2, dtype="int64"),)
@@ -5311,11 +5311,11 @@ class Flatten(Op):
             # We do not want to force an upcast of part2 if its length is 1
             pass
         else:
-            if len(in_shp) == 0 and self.outdim == 1:
+            if len(in_shp) == 0 and self.ndim == 1:
                 part2 = (1,)
             else:
                 raise ValueError(
-                    f"invalid output ndimensions ({self.outdim}) for tensor "
+                    f"invalid output ndimensions ({self.ndim}) for tensor "
                     f"of rank {len(in_shp)}"
                 )
 
@@ -5338,11 +5338,11 @@ class Flatten(Op):
     def c_code(self, node, name, inputs, outputs, sub):
         (x,) = inputs
         (out,) = outputs
-        outdim = self.outdim
+        ndim = self.ndim
         fail = sub["fail"]
         return (
             """
-        if (%(outdim)s == PyArray_NDIM(%(x)s))
+        if (%(ndim)s == PyArray_NDIM(%(x)s))
         {
             Py_XDECREF(%(out)s);
             Py_XINCREF(%(x)s);
@@ -5352,7 +5352,7 @@ class Flatten(Op):
         {
             Py_XDECREF(%(out)s);
 
-            if (%(outdim)s == 1)
+            if (%(ndim)s == 1)
             {
                 npy_intp size = PyArray_SIZE(%(x)s);
                 PyArray_Dims newshape;
@@ -5365,20 +5365,20 @@ class Flatten(Op):
             else
             {
                 npy_intp *oldshape = PyArray_DIMS(%(x)s);
-                npy_intp newshape_dims[%(outdim)s];
+                npy_intp newshape_dims[%(ndim)s];
 
                 int i;
-                for (i = 0; i < %(outdim)s - 1; ++i)
+                for (i = 0; i < %(ndim)s - 1; ++i)
                     newshape_dims[i] = oldshape[i];
 
                 newshape_dims[i] = 1;
 
-                for (int j = %(outdim)s - 1; j < PyArray_NDIM(%(x)s); ++j)
+                for (int j = %(ndim)s - 1; j < PyArray_NDIM(%(x)s); ++j)
                     newshape_dims[i] *= oldshape[j];
 
                 PyArray_Dims newshape;
                 newshape.ptr = newshape_dims;
-                newshape.len = %(outdim)s;
+                newshape.len = %(ndim)s;
                 %(out)s = (PyArrayObject*)PyArray_Newshape(%(x)s,
                                                            &newshape,
                                                            NPY_CORDER);
@@ -5428,36 +5428,29 @@ def is_flat(var, ndim=None, outdim=None):
     return var.ndim == ndim
 
 
-def flatten(x, ndim=None, outdim=None):
-    """
-    Reshapes the variable x by keeping
-    the first outdim-1 dimension size(s) of x the same,
-    and making the last dimension size of x equal to
-    the multiplication of its remaining dimension size(s).
+def flatten(x, ndim=1):
+    """Return a copy of the array collapsed into one dimension.
+
+    Reshapes the variable `x` by keeping the first outdim-1 dimension size(s)
+    of `x` the same, and making the last dimension size of `x` equal to the
+    multiplication of its remaining dimension size(s).
 
     Parameters
     ----------
     x : theano.tensor.var.TensorVariable
-            the variable that should be reshaped.
-
+        The variable to be reshaped.
     ndim : int
-            the number of dimensions of the returned variable
-            Default 1.
-        outdim : int
-            DEPRECATED synonym for ndim
+        The number of dimensions of the returned variable
+        The default value is ``1``.
+
     Returns
     -------
     theano.tensor.var.TensorVariable
         the flattend variable with dimensionality of outdim
     """
-    if outdim is None and ndim is None:
+    if ndim is None:
         ndim = 1
-    elif outdim is not None and ndim is not None:
-        raise ValueError("You should only specify ndim")
-    elif outdim is not None:
-        warnings.warn("flatten outdim parameter is deprecated, use ndim instead.")
 
-        ndim = outdim
     # Any input variable can be flattened to have ndim of 1,
     # even if it's a scalar. Otherwise, ndim must be positive
     # and smaller than x.ndim.
@@ -5473,32 +5466,11 @@ def flatten(x, ndim=None, outdim=None):
     bcast_new_dim = builtins.all(x.broadcastable[ndim - 1 :])
     broadcastable = bcast_kept_dims + (bcast_new_dim,)
     x_reshaped = theano.tensor.addbroadcast(
-        x_reshaped, *filter(lambda i: broadcastable[i], range(ndim))
+        x_reshaped, *[i for i in range(ndim) if broadcastable[i]]
     )
     return x_reshaped
 
 
-# class TileGrad(Op):
-#     """
-#     Calculates the gradient of the Tile Op.
-#     """
-#     # this is so weird, I can't think of how to make this a general thing.
-#     def make_node(self, x, reps, g_out):
-#         return gof.Apply(self, [x, reps, g_out], [x.type()])
-#
-#     def perform(self, node, inp, out):
-#         x, reps, g_out = inp
-#         gx, = out
-#         xsh = x.shape
-#         if len(reps) == 2 and reps[1] == 1 and len(x.shape) == 1:
-#             gx[0] = numpy.sum(g_out, axis=0)
-#         else:
-#             raise NotImplementedError('x.shape, reps combination not '
-#                                       'supported', (x.shape, reps))
-#
-# tilegrad = TileGrad()
-
-
 class Tile(Op):
     """
     Construct an array by repeating the input x according to reps pattern.

theano/tensor/nnet/conv.py

@@ -554,12 +554,12 @@ class ConvOp(OpenMPOp):
                     new -= 1
 
                 warnstr = (
-                    "OPTIMISATION WARNING: in ConvOp.__init__() "
-                    "unroll_batch(%i) must be 0 or a divisor of"
-                    " bsize(%i). We revert it to %i. This"
+                    "In ConvOp.__init__(): "
+                    f"unroll_batch({self.unroll_batch}) must be 0 or a divisor of"
+                    f" bsize({self.bsize}). We revert it to {new}. This"
                     " won't change the result, but may make it slower."
                 )
-                _logger.warning(warnstr, self.unroll_batch, self.bsize, new)
+                _logger.warning(warnstr)
 
                 self.unroll_batch = new
 
@@ -580,12 +580,12 @@ class ConvOp(OpenMPOp):
                     new -= 1
 
                 warnstr = (
-                    "OPTIMISATION WARNING: in ConvOp.__init__()"
-                    " unroll_kern(%i) should be 0 or a divisor of"
-                    " nkern(%i). We revert it to %i. This"
+                    "In ConvOp.__init__(): "
+                    f"unroll_kern({self.unroll_kern}) must be 0 or a divisor of"
+                    f" nkern({self.nkern}). We revert it to {new}. This"
                     " won't change the result, but may make it slower."
                 )
-                _logger.warning(warnstr, self.unroll_kern, self.nkern, new)
+                _logger.warning(warnstr)
                 self.unroll_kern = new
 
         self.outshp = get_conv_output_shape(

theano/tensor/nnet/nnet.py

@@ -440,9 +440,10 @@ class Softmax(Op):
             raise ValueError(f"x must be 1-d or 2-d tensor of floats. Got {x.type}")
         if x.ndim == 1:
             warnings.warn(
-                "DEPRECATION: If x is a vector, Softmax will not automatically pad x "
+                "If x is a vector, Softmax will not automatically pad x "
                 "anymore in next releases. If you need it, please do it manually. The "
                 "vector case is gonna be supported soon and the output will be a vector.",
+                category=PendingDeprecationWarning,
                 stacklevel=4,
             )
             x = tt.shape_padleft(x, n_ones=1)
@@ -638,9 +639,10 @@ class LogSoftmax(Op):
             raise ValueError(f"x must be 1-d or 2-d tensor of floats. Got {x.type}")
         if x.ndim == 1:
             warnings.warn(
-                "DEPRECATION: If x is a vector, LogSoftmax will not automatically pad x "
+                "If x is a vector, LogSoftmax will not automatically pad x "
                 "anymore in next releases. If you need it, please do it manually. The "
                 "vector case is gonna be supported soon and the output will be a vector.",
+                category=PendingDeprecationWarning,
                 stacklevel=4,
             )
             x = tt.shape_padleft(x, n_ones=1)
@@ -1661,8 +1663,7 @@ def local_argmax_pushdown(fgraph, node):
     ):
         if theano.config.warn__argmax_pushdown_bug:
             logging.getLogger("theano.tensor.nnet.nnet").warn(
-                "WARNING: there "
-                "was a bug in Theano fixed on May 27th, 2010 in this case."
+                "There was a bug in Theano fixed on May 27th, 2010 in this case."
                 " I.E. when we take the max of a softplus, softmax, exp, "
                 "log, tanh, sigmoid, softmax_with_bias op, we were doing "
                 "the max of the parent of the input. To remove this "

theano/tensor/opt.py

@@ -5821,7 +5821,7 @@ def local_sum_prod_div_dimshuffle(fgraph, node):
 
                 if compatible_dims:
                     _logger.warning(
-                        "WARNING: Your current code is fine, but"
+                        "Your current code is fine, but"
                         " Theano versions between "
                         "rev. 3bd9b789f5e8 (2010-06-16) and"
                         " cfc6322e5ad4 (2010-08-03) would "
@@ -5876,7 +5876,7 @@ def local_sum_prod_div_dimshuffle(fgraph, node):
                             node.op, Sum
                         ):
                             _logger.warning(
-                                "WARNING: Your current code is fine,"
+                                "Your current code is fine,"
                                 " but Theano versions between "
                                 "rev. 3bd9b789f5e8 (2010-06-16) and"
                                 " cfc6322e5ad4 (2010-08-03) would "
@@ -5986,7 +5986,7 @@ def local_op_of_op(fgraph, node):
                     and len(newaxis) == len(newaxis_old)
                 ):
                     _logger.warning(
-                        "WARNING (YOUR CURRENT CODE IS FINE): Theano "
+                        "(YOUR CURRENT CODE IS FINE): Theano "
                         "versions between version 9923a40c7b7a and August "
                         "2nd, 2010 generated bugged code in this case. "
                         "This happens when there are two consecutive sums "

theano/tensor/raw_random.py

 """Define random number Type (`RandomStateType`) and Op (`RandomFunction`)."""
 
-
-import sys
 from copy import copy
 from functools import reduce
+from warnings import warn
 
 import numpy as np
 
@@ -200,10 +199,8 @@ class RandomFunction(gof.Op):
         assert shape.type.ndim == 1
         assert (shape.type.dtype == "int64") or (shape.type.dtype == "int32")
         if not isinstance(r.type, RandomStateType):
-            print(
-                "WARNING: RandomState instances should be in RandomStateType",
-                file=sys.stderr,
-            )
+            warn("RandomState instances should be in RandomStateType")
+
         # the following doesn't work because we want to ignore the
         # broadcastable flags in shape.type
         # assert shape.type == tensor.lvector
@@ -532,9 +529,9 @@ def binomial(random_state, size=None, n=1, p=0.5, ndim=None, dtype="int64", prob
     """
     if prob is not None:
         p = prob
-        print(
-            "DEPRECATION WARNING: the parameter prob to the binomal fct have been renamed to p to have the same name as np.",
-            file=sys.stderr,
+        warn(
+            "The parameter prob to the binomal fct have been renamed to p to have the same name as np.",
+            category=DeprecationWarning,
         )
     n = tensor.as_tensor_variable(n)
     p = tensor.as_tensor_variable(p)
@@ -964,9 +961,9 @@ class RandomStreamsBase:
         """
         if prob is not None:
             p = prob
-            print(
-                "DEPRECATION WARNING: the parameter prob to the binomal fct have been renamed to p to have the same name as numpy.",
-                file=sys.stderr,
+            warn(
+                "The parameter prob to the binomal fct have been renamed to p to have the same name as numpy.",
+                category=DeprecationWarning,
             )
         return self.gen(binomial, size, n, p, ndim=ndim, dtype=dtype)
 

theano/tensor/signal/pool.py

@@ -93,9 +93,10 @@ def pool_2d(
             )
         else:
             warnings.warn(
-                "DEPRECATION: the 'ds' parameter is not going to exist"
+                "The 'ds' parameter is not going to exist"
                 " anymore as it is going to be replaced by the parameter"
                 " 'ws'.",
+                category=DeprecationWarning,
                 stacklevel=2,
             )
             ws = ds
@@ -110,9 +111,10 @@ def pool_2d(
             )
         else:
             warnings.warn(
-                "DEPRECATION: the 'st' parameter is not going to exist"
+                "The 'st' parameter is not going to exist"
                 " anymore as it is going to be replaced by the parameter"
                 " 'stride'.",
+                category=DeprecationWarning,
                 stacklevel=2,
             )
             stride = st
@@ -125,9 +127,10 @@ def pool_2d(
             )
         else:
             warnings.warn(
-                "DEPRECATION: the 'padding' parameter is not going to exist"
+                "The 'padding' parameter is not going to exist"
                 " anymore as it is going to be replaced by the parameter"
                 " 'pad'.",
+                category=DeprecationWarning,
                 stacklevel=2,
             )
             pad = padding
@@ -145,6 +148,7 @@ def pool_2d(
             " GPU combination supported is when"
             " `ws == stride and pad == (0, 0) and mode == 'max'`."
             " Otherwise, the convolution will be executed on CPU.",
+            category=DeprecationWarning,
             stacklevel=2,
         )
         ignore_border = False
@@ -210,9 +214,10 @@ def pool_3d(
             )
         else:
             warnings.warn(
-                "DEPRECATION: the 'ds' parameter is not going to exist"
+                "The 'ds' parameter is not going to exist"
                 " anymore as it is going to be replaced by the parameter"
                 " 'ws'.",
+                category=DeprecationWarning,
                 stacklevel=2,
             )
             ws = ds
@@ -227,9 +232,10 @@ def pool_3d(
             )
         else:
             warnings.warn(
-                "DEPRECATION: the 'st' parameter is not going to exist"
+                "The 'st' parameter is not going to exist"
                 " anymore as it is going to be replaced by the parameter"
                 " 'stride'.",
+                category=DeprecationWarning,
                 stacklevel=2,
             )
             stride = st
@@ -242,9 +248,10 @@ def pool_3d(
             )
         else:
             warnings.warn(
-                "DEPRECATION: the 'padding' parameter is not going to exist"
+                "The 'padding' parameter is not going to exist"
                 " anymore as it is going to be replaced by the parameter"
                 " 'pad'.",
+                category=DeprecationWarning,
                 stacklevel=2,
             )
             pad = padding
@@ -262,6 +269,7 @@ def pool_3d(
             " GPU combination supported is when"
             " `ws == stride and pad == (0, 0, 0) and mode == 'max'`."
             " Otherwise, the convolution will be executed on CPU.",
+            category=DeprecationWarning,
             stacklevel=2,
         )
         ignore_border = False
@@ -387,9 +395,10 @@ class Pool(OpenMPOp):
                 )
             else:
                 warnings.warn(
-                    "DEPRECATION: the 'ds' parameter is not going to exist"
+                    "The 'ds' parameter is not going to exist"
                     " anymore as it is going to be replaced by the parameter"
                     " 'ws'.",
+                    category=DeprecationWarning,
                     stacklevel=2,
                 )
                 ws = ds
@@ -404,9 +413,10 @@ class Pool(OpenMPOp):
                 )
             else:
                 warnings.warn(
-                    "DEPRECATION: the 'st' parameter is not going to exist"
+                    "The 'st' parameter is not going to exist"
                     " anymore as it is going to be replaced by the parameter"
                     " 'stride'.",
+                    category=DeprecationWarning,
                     stacklevel=2,
                 )
                 stride = st
@@ -420,9 +430,10 @@ class Pool(OpenMPOp):
                 )
             else:
                 warnings.warn(
-                    "DEPRECATION: the 'padding' parameter is not going to"
+                    "The 'padding' parameter is not going to"
                     " exist anymore as it is going to be replaced by the"
                     " parameter 'pad'.",
+                    category=DeprecationWarning,
                     stacklevel=2,
                 )
                 pad = padding
@@ -1032,9 +1043,10 @@ class PoolGrad(OpenMPOp):
                 )
             else:
                 warnings.warn(
-                    "DEPRECATION: the 'ds' parameter in PoolGrad is not going"
+                    "The 'ds' parameter in PoolGrad is not going"
                     " to exist anymore as it is going to be replaced by the"
                     " parameter 'ws'.",
+                    category=DeprecationWarning,
                     stacklevel=2,
                 )
                 ws = ds
@@ -1049,9 +1061,10 @@ class PoolGrad(OpenMPOp):
                 )
             else:
                 warnings.warn(
-                    "DEPRECATION: the 'st' parameter in PoolGrad is not going"
+                    "The 'st' parameter in PoolGrad is not going"
                     " to exist anymore as it is going to be replaced by the"
                     " parameter 'stride'.",
+                    category=DeprecationWarning,
                     stacklevel=2,
                 )
                 stride = st
@@ -1064,9 +1077,10 @@ class PoolGrad(OpenMPOp):
                 )
             else:
                 warnings.warn(
-                    "DEPRECATION: the 'padding' parameter in PoolGrad is not"
+                    "The 'padding' parameter in PoolGrad is not"
                     " going to exist anymore as it is going to be replaced"
                     " by the parameter 'pad'.",
+                    category=DeprecationWarning,
                     stacklevel=2,
                 )
                 pad = padding

theano/tensor/type.py

@@ -55,9 +55,10 @@ class TensorType(Type):
         self.sparse_grad = sparse_grad
         if sparse_grad:
             warnings.warn(
-                "DEPRECATION WARNING: You use an old interface to"
+                "You use an old interface to"
                 " AdvancedSubtensor1 sparse_grad. Now use"
-                " theano.sparse_grad(a_tensor[an_int_vector])."
+                " theano.sparse_grad(a_tensor[an_int_vector]).",
+                category=DeprecationWarning,
             )
 
     def clone(self, dtype=None, broadcastable=None):