Skip to content

P
pytensor
提交 b4dc02d6 authored 作者: Brandon T. Willard's avatar Brandon T. Willard
浏览文件
操作

Replace theano.tensor alias T with tt in tests.gpuarray

上级 b4c51eb2
隐藏空白字符变更
内嵌 并排
正在显示 包含 345 行增加 和 352 行删除
tests/gpuarray/rnn_support.py
浏览文件 @ b4dc02d6
import numpy as np
import theano
import theano.tensor as T
import theano.tensor as tt
class Model(object):
......@@ -125,16 +125,16 @@ class GRU(Layer):
"""step through processed input to create output"""
def step(inp, s_prev):
i_t = T.nnet.sigmoid(
T.dot(inp, self.W_i) + T.dot(s_prev, self.R_i) + self.b_wi + self.b_ru
i_t = tt.nnet.sigmoid(
tt.dot(inp, self.W_i) + tt.dot(s_prev, self.R_i) + self.b_wi + self.b_ru
)
r_t = T.nnet.sigmoid(
T.dot(inp, self.W_r) + T.dot(s_prev, self.R_r) + self.b_wr + self.b_rr
r_t = tt.nnet.sigmoid(
tt.dot(inp, self.W_r) + tt.dot(s_prev, self.R_r) + self.b_wr + self.b_rr
)
h_hat_t = T.tanh(
T.dot(inp, self.W_h)
+ (r_t * (T.dot(s_prev, self.R_h) + self.b_rh))
h_hat_t = tt.tanh(
tt.dot(inp, self.W_h)
+ (r_t * (tt.dot(s_prev, self.R_h) + self.b_rh))
+ self.b_wh
)
......@@ -229,21 +229,21 @@ class LSTM(Layer):
"""step through processed input to create output"""
def step(x_t, h_tm1, c_tm1):
i_t = T.nnet.sigmoid(
T.dot(x_t, self.W_i) + T.dot(h_tm1, self.R_i) + self.b_wi + self.b_ri
i_t = tt.nnet.sigmoid(
tt.dot(x_t, self.W_i) + tt.dot(h_tm1, self.R_i) + self.b_wi + self.b_ri
)
f_t = T.nnet.sigmoid(
T.dot(x_t, self.W_f) + T.dot(h_tm1, self.R_f) + self.b_wf + self.b_rf
f_t = tt.nnet.sigmoid(
tt.dot(x_t, self.W_f) + tt.dot(h_tm1, self.R_f) + self.b_wf + self.b_rf
)
o_t = T.nnet.sigmoid(
T.dot(x_t, self.W_o) + T.dot(h_tm1, self.R_o) + self.b_ro + self.b_wo
o_t = tt.nnet.sigmoid(
tt.dot(x_t, self.W_o) + tt.dot(h_tm1, self.R_o) + self.b_ro + self.b_wo
)
c_hat_t = T.tanh(
T.dot(x_t, self.W_c) + T.dot(h_tm1, self.R_c) + self.b_wc + self.b_rc
c_hat_t = tt.tanh(
tt.dot(x_t, self.W_c) + tt.dot(h_tm1, self.R_c) + self.b_wc + self.b_rc
)
c_t = f_t * c_tm1 + i_t * c_hat_t
h_t = o_t * T.tanh(c_t)
h_t = o_t * tt.tanh(c_t)
return h_t, c_t
......@@ -275,7 +275,7 @@ class FC(Layer):
self.b = bias_weights((output_dim,), param_list=self.params, name=name + ".b")
def output(self):
return T.dot(self.X, self.W) + self.b
return tt.dot(self.X, self.W) + self.b
class WrapperLayer(Layer):
......
tests/gpuarray/test_basic_ops.py
浏览文件 @ b4dc02d6
......@@ -4,7 +4,7 @@ pygpu = pytest.importorskip("pygpu")
gpuarray = pygpu.gpuarray
import numpy as np
import theano
import theano.tensor as T
import theano.tensor as tt
from theano.tensor import TensorType
from theano.tensor.basic import alloc
......@@ -217,7 +217,7 @@ def makeTester(
def test_transfer_cpu_gpu():
a = T.fmatrix("a")
a = tt.fmatrix("a")
g = GpuArrayType(dtype="float32", broadcastable=(False, False))("g")
av = np.asarray(rng.rand(5, 4), dtype="float32")
......@@ -254,7 +254,7 @@ def test_transfer_strided():
# This is just to ensure that it works in theano
# libgpuarray has a much more comprehensive suit of tests to
# ensure correctness
a = T.fmatrix("a")
a = tt.fmatrix("a")
g = GpuArrayType(dtype="float32", broadcastable=(False, False))("g")
av = np.asarray(rng.rand(5, 8), dtype="float32")
......@@ -300,7 +300,7 @@ class TestGPUAlloc(TestAlloc):
dtype = "float32"
mode = mode_with_gpu
shared = staticmethod(gpuarray_shared_constructor)
allocs = [GpuAlloc(test_ctx_name), GpuAlloc(test_ctx_name), T.Alloc()]
allocs = [GpuAlloc(test_ctx_name), GpuAlloc(test_ctx_name), tt.Alloc()]
def test_alloc_empty():
......@@ -343,21 +343,21 @@ def test_shape():
assert np.all(f(v) == (3, 4, 5))
if theano.config.mode != "FAST_COMPILE":
assert len(topo) == 4
assert isinstance(topo[0].op, T.opt.Shape_i)
assert isinstance(topo[1].op, T.opt.Shape_i)
assert isinstance(topo[2].op, T.opt.Shape_i)
assert isinstance(topo[3].op, T.opt.MakeVector)
assert isinstance(topo[0].op, tt.opt.Shape_i)
assert isinstance(topo[1].op, tt.opt.Shape_i)
assert isinstance(topo[2].op, tt.opt.Shape_i)
assert isinstance(topo[3].op, tt.opt.MakeVector)
mode = mode_with_gpu.excluding("local_shape_to_shape_i")
f = theano.function([x], x.shape, mode=mode)
topo = f.maker.fgraph.toposort()
assert np.all(f(v) == (3, 4, 5))
assert len(topo) == 1
assert isinstance(topo[0].op, T.Shape)
assert isinstance(topo[0].op, tt.Shape)
def test_gpu_contiguous():
a = T.fmatrix("a")
i = T.iscalar("i")
a = tt.fmatrix("a")
i = tt.iscalar("i")
a_val = np.asarray(np.random.rand(4, 5), dtype="float32")
# The reshape is needed otherwise we make the subtensor on the CPU
# to transfer less data.
......@@ -383,8 +383,8 @@ class TestGPUReshape(TestReshape):
theano.compile.DeepCopyOp,
GpuDimShuffle,
GpuElemwise,
theano.tensor.opt.Shape_i,
theano.tensor.opt.MakeVector,
tt.opt.Shape_i,
tt.opt.MakeVector,
)
assert self.op == GpuReshape
......@@ -418,7 +418,7 @@ class TestGPUJoinAndSplit(TestJoinAndSplit):
# Also test float16 computation at the same time.
rng = np.random.RandomState(seed=utt.fetch_seed())
m = self.shared(rng.rand(4, 6).astype("float16"))
o = T.Split(2)(m, 0, [2, 2])
o = tt.Split(2)(m, 0, [2, 2])
assert o[0].dtype == "float16"
f = theano.function([], o, mode=self.mode)
assert any(
......@@ -433,22 +433,22 @@ class TestGPUJoinAndSplit(TestJoinAndSplit):
def test_gpujoin_gpualloc():
a = T.fmatrix("a")
a = tt.fmatrix("a")
a_val = np.asarray(np.random.rand(4, 5), dtype="float32")
b = T.fmatrix("b")
b = tt.fmatrix("b")
b_val = np.asarray(np.random.rand(3, 5), dtype="float32")
f = theano.function(
[a, b], T.join(0, T.zeros_like(a), T.ones_like(b)) + 4, mode=mode_without_gpu
[a, b], tt.join(0, tt.zeros_like(a), tt.ones_like(b)) + 4, mode=mode_without_gpu
)
f_gpu = theano.function(
[a, b], T.join(0, T.zeros_like(a), T.ones_like(b)), mode=mode_with_gpu
[a, b], tt.join(0, tt.zeros_like(a), tt.ones_like(b)), mode=mode_with_gpu
)
f_gpu2 = theano.function(
[a, b], T.join(0, T.zeros_like(a), T.ones_like(b)) + 4, mode=mode_with_gpu
[a, b], tt.join(0, tt.zeros_like(a), tt.ones_like(b)) + 4, mode=mode_with_gpu
)
assert sum([node.op == T.alloc for node in f.maker.fgraph.toposort()]) == 2
assert sum([node.op == T.join_ for node in f.maker.fgraph.toposort()]) == 1
assert sum([node.op == tt.alloc for node in f.maker.fgraph.toposort()]) == 2
assert sum([node.op == tt.join_ for node in f.maker.fgraph.toposort()]) == 1
assert (
sum([isinstance(node.op, GpuAlloc) for node in f_gpu.maker.fgraph.toposort()])
== 2
......@@ -471,10 +471,10 @@ def test_gpueye():
# allowed.
if M is None:
M = N
N_symb = T.iscalar()
M_symb = T.iscalar()
k_symb = T.iscalar()
out = T.eye(N_symb, M_symb, k_symb, dtype=dtype) + np.array(1).astype(dtype)
N_symb = tt.iscalar()
M_symb = tt.iscalar()
k_symb = tt.iscalar()
out = tt.eye(N_symb, M_symb, k_symb, dtype=dtype) + np.array(1).astype(dtype)
f = theano.function([N_symb, M_symb, k_symb], out, mode=mode_with_gpu)
result = np.asarray(f(N, M, k)) - np.array(1).astype(dtype)
......@@ -511,7 +511,7 @@ def test_hostfromgpu_shape_i():
m = mode_with_gpu.including(
"local_dot_to_dot22", "local_dot22_to_dot22scalar", "specialize"
)
a = T.fmatrix("a")
a = tt.fmatrix("a")
ca = theano.gpuarray.type.GpuArrayType("float32", (False, False))()
av = np.asarray(np.random.rand(5, 4), dtype="float32")
cv = gpuarray.asarray(
......@@ -522,9 +522,9 @@ def test_hostfromgpu_shape_i():
assert any(isinstance(x.op, GpuFromHost) for x in f.maker.fgraph.toposort())
f = theano.function([a], GpuFromHost(test_ctx_name)(a).shape, mode=m)
topo = f.maker.fgraph.toposort()
assert isinstance(topo[0].op, T.opt.Shape_i)
assert isinstance(topo[1].op, T.opt.Shape_i)
assert isinstance(topo[2].op, T.opt.MakeVector)
assert isinstance(topo[0].op, tt.opt.Shape_i)
assert isinstance(topo[1].op, tt.opt.Shape_i)
assert isinstance(topo[2].op, tt.opt.MakeVector)
assert tuple(f(av)) == (5, 4)
f = theano.function([ca], host_from_gpu(ca), mode=m)
......@@ -533,7 +533,7 @@ def test_hostfromgpu_shape_i():
topo = f.maker.fgraph.toposort()
assert isinstance(topo[0].op, theano.compile.Shape_i)
assert isinstance(topo[1].op, theano.compile.Shape_i)
assert isinstance(topo[2].op, theano.tensor.opt.MakeVector)
assert isinstance(topo[2].op, tt.opt.MakeVector)
assert tuple(f(cv)) == (5, 4)
......@@ -544,10 +544,10 @@ def test_Gpujoin_inplace():
# Gpujoin function but all except one of them are empty. In this case
# Gpujoin should work inplace and the output should be the view of the
# non-empty element.
s = T.lscalar()
s = tt.lscalar()
data = np.array([3, 4, 5], dtype=theano.config.floatX)
x = gpuarray_shared_constructor(data, borrow=True)
z = T.zeros((s,))
z = tt.zeros((s,))
join = GpuJoin(view=0)
c = join(0, x, z)
......@@ -560,11 +560,11 @@ def test_Gpujoin_inplace():
def test_gpu_tril_triu():
def check_l(m, k=0):
m_symb = T.matrix(dtype=m.dtype)
k_symb = T.iscalar()
m_symb = tt.matrix(dtype=m.dtype)
k_symb = tt.iscalar()
f = theano.function(
[m_symb, k_symb], T.tril(m_symb, k_symb), mode=mode_with_gpu
[m_symb, k_symb], tt.tril(m_symb, k_symb), mode=mode_with_gpu
)
result = f(m, k)
assert np.allclose(result, np.tril(m, k))
......@@ -572,10 +572,10 @@ def test_gpu_tril_triu():
assert any([isinstance(node.op, GpuTri) for node in f.maker.fgraph.toposort()])
def check_u(m, k=0):
m_symb = T.matrix(dtype=m.dtype)
k_symb = T.iscalar()
m_symb = tt.matrix(dtype=m.dtype)
k_symb = tt.iscalar()
f = theano.function(
[m_symb, k_symb], T.triu(m_symb, k_symb), mode=mode_with_gpu
[m_symb, k_symb], tt.triu(m_symb, k_symb), mode=mode_with_gpu
)
result = f(m, k)
assert np.allclose(result, np.triu(m, k))
......@@ -624,10 +624,10 @@ def test_gputri():
# allowed.
if M is None:
M = N
N_symb = T.iscalar()
M_symb = T.iscalar()
k_symb = T.iscalar()
out = T.tri(N_symb, M_symb, k_symb, dtype=dtype) + np.array(1).astype(dtype)
N_symb = tt.iscalar()
M_symb = tt.iscalar()
k_symb = tt.iscalar()
out = tt.tri(N_symb, M_symb, k_symb, dtype=dtype) + np.array(1).astype(dtype)
f = theano.function([N_symb, M_symb, k_symb], out, mode=mode_with_gpu)
result = np.asarray(f(N, M, k)) - np.array(1).astype(dtype)
assert np.allclose(result, np.tri(N, M_, k, dtype=dtype))
......
tests/gpuarray/test_ctc.py
浏览文件 @ b4dc02d6
......@@ -2,7 +2,7 @@ import pytest
import numpy as np
import theano
import theano.tensor as T
import theano.tensor as tt
import theano.gpuarray
from theano.gpuarray.ctc import gpu_ctc, GpuConnectionistTemporalClassification
......@@ -56,7 +56,7 @@ class TestCTC:
outputs = [cpu_ctc_cost]
if compute_grad:
# Symbolic gradient of CTC cost
cpu_ctc_grad = T.grad(T.mean(cpu_ctc_cost), activations)
cpu_ctc_grad = tt.grad(tt.mean(cpu_ctc_cost), activations)
outputs += [cpu_ctc_grad]
return theano.function([], outputs, mode=mode)
......@@ -65,7 +65,7 @@ class TestCTC:
outputs = [gpu_ctc_cost]
if compute_grad:
# Symbolic gradient of CTC cost
gpu_ctc_grad = T.grad(T.mean(gpu_ctc_cost), activations)
gpu_ctc_grad = tt.grad(tt.mean(gpu_ctc_cost), activations)
outputs += [gpu_ctc_grad]
return theano.function([], outputs, mode=mode_with_gpu)
......
tests/gpuarray/test_dnn.py
浏览文件 @ b4dc02d6
......@@ -2,9 +2,10 @@ import logging
import pytest
import numpy as np
pygpu = pytest.importorskip("pygpu")
pygpu = pytest.importorskip("pygpu") # noqa
import theano
import theano.tensor as T
import theano.tensor as tt
import tests.unittest_tools as utt
from itertools import product, chain
......@@ -12,17 +13,27 @@ from collections import OrderedDict
from six import StringIO
from theano.tensor.signal.pool import pool_2d, pool_3d
from theano.tensor.signal.pool import Pool, MaxPoolGrad, AveragePoolGrad
from theano.tensor.nnet import (
bn,
softmax_op,
SoftmaxGrad,
softmax,
LogSoftmax,
Softmax,
conv2d,
)
from theano.tensor.nnet.corr3d import Corr3dMM
from theano.tensor.nnet.corr import CorrMM
from theano.tensor.nnet.abstract_conv import (
get_conv_output_shape,
get_conv_gradinputs_shape,
)
from theano.tensor.nnet import bn
from theano.tensor.signal.pool import pool_2d, pool_3d
from theano.tensor.signal.pool import Pool, MaxPoolGrad, AveragePoolGrad
from theano.configdefaults import SUPPORTED_DNN_CONV_ALGO_FWD
from theano.gpuarray import dnn
from theano.gpuarray.basic_ops import GpuAllocEmpty
from theano.gpuarray.type import gpuarray_shared_constructor, GpuArrayType
......@@ -76,7 +87,7 @@ def set_precision(floatX):
def test_dnn_conv_desc_merge():
kern_shp = T.as_tensor_variable(np.asarray([3, 1, 2, 2]).astype("int64"))
kern_shp = tt.as_tensor_variable(np.asarray([3, 1, 2, 2]).astype("int64"))
desc1 = dnn.GpuDnnConvDesc(
border_mode="valid", subsample=(2, 2), dilation=(1, 1), conv_mode="conv"
)(kern_shp)
......@@ -101,9 +112,9 @@ def test_dnn_conv_merge():
# This test that we merge correctly multiple dnn_conv.
img_shp = [2, 5, 6, 8]
kern_shp = [3, 5, 5, 6]
img = T.tensor4("img")
kern = T.tensor4("kern")
out = T.tensor4("out")
img = tt.tensor4("img")
kern = tt.tensor4("kern")
out = tt.tensor4("out")
desc = dnn.GpuDnnConvDesc(border_mode="valid")(kern.shape)
# Test forward op
......@@ -139,9 +150,9 @@ def test_dnn_conv_inplace():
img_shp = [2, 5, 6, 8]
kern_shp = [3, 5, 5, 6]
img = T.tensor4("img")
kern = T.tensor4("kern")
out = T.tensor4("out")
img = tt.tensor4("img")
kern = tt.tensor4("kern")
out = tt.tensor4("out")
desc1 = dnn.GpuDnnConvDesc(border_mode="valid", conv_mode="conv")(kern.shape)
desc2 = dnn.GpuDnnConvDesc(border_mode="valid", conv_mode="cross")(kern.shape)
......@@ -184,9 +195,9 @@ def test_dnn_conv_inplace():
def run_dnn_conv_invalid_precision(ndim):
bc = (False,) * (ndim + 2)
img = T.tensor(theano.config.floatX, broadcastable=bc)
kerns = T.tensor(theano.config.floatX, broadcastable=bc)
topgrad = T.tensor(theano.config.floatX, broadcastable=bc)
img = tt.tensor(theano.config.floatX, broadcastable=bc)
kerns = tt.tensor(theano.config.floatX, broadcastable=bc)
topgrad = tt.tensor(theano.config.floatX, broadcastable=bc)
shape = np.arange(ndim + 2)
if ndim == 2:
dnn_conv_func = dnn.dnn_conv
......@@ -242,8 +253,8 @@ def test_dnn_conv_invalid_precision():
def test_dnn_conv_mixed_dtype():
mf = T.ftensor4()
md = T.dtensor4()
mf = tt.ftensor4()
md = tt.dtensor4()
def assert_types(conv):
dt = conv.owner.inputs[0].dtype
......@@ -259,8 +270,8 @@ def test_dnn_conv_mixed_dtype():
def test_dnn_conv3d_mixed_dtype():
mf = T.ftensor5()
md = T.dtensor5()
mf = tt.ftensor5()
md = tt.dtensor5()
def assert_types(conv):
dt = conv.owner.inputs[0].dtype
......@@ -280,7 +291,7 @@ def test_pooling():
modes = get_dnn_pool_modes()
x = T.tensor4()
x = tt.tensor4()
for mode, pad in product(modes, ((0, 0), (1, 0), (0, 1), (2, 3), (3, 2))):
if pad != (0, 0) and mode == "average_exc_pad":
# Not implemented
......@@ -386,7 +397,7 @@ def test_pooling():
def run_pooling_with_tensor_vars(mode):
utt.seed_rng()
x = T.tensor4()
x = tt.tensor4()
ws = theano.shared(np.array([2, 2], dtype="int32"))
stride = theano.shared(np.array([1, 1], dtype="int32"))
pad = theano.shared(np.array([0, 0], dtype="int32"))
......@@ -446,7 +457,7 @@ def test_pooling3d():
modes = get_dnn_pool_modes()
x = T.tensor5()
x = tt.tensor5()
for mode, pad in product(
modes,
((0, 0, 0), (1, 0, 0), (0, 1, 0), (0, 0, 1), (2, 3, 2), (3, 2, 2), (2, 2, 3)),
......@@ -549,7 +560,7 @@ def test_pooling_opt():
utt.seed_rng()
# 2D pooling
x = T.matrix()
x = tt.matrix()
f = theano.function(
[x],
......@@ -564,7 +575,7 @@ def test_pooling_opt():
# gradient of 2D pooling
f = theano.function(
[x],
T.grad(
tt.grad(
pool_2d(x, ws=(2, 2), mode="average_inc_pad", ignore_border=True).sum(), x
),
mode=mode_with_gpu.including("cudnn"),
......@@ -586,7 +597,7 @@ def test_pooling_opt():
f(data)
# 3D pooling
x = T.tensor3()
x = tt.tensor3()
f = theano.function(
[x],
......@@ -601,7 +612,7 @@ def test_pooling_opt():
# gradient of 3D pooling
f = theano.function(
[x],
T.grad(
tt.grad(
pool_3d(x, ws=(2, 2, 2), mode="average_inc_pad", ignore_border=True).sum(),
x,
),
......@@ -632,7 +643,7 @@ def test_pooling_opt_arbitrary_dimensions():
for mode in modes:
out_pool = Pool(ndim=len(ws), mode=mode, ignore_border=True)(input, ws)
out_pool_grad = T.grad(T.sum(out_pool), wrt=input)
out_pool_grad = tt.grad(tt.sum(out_pool), wrt=input)
out = [out_pool, out_pool_grad]
# run on GPU
......@@ -679,14 +690,14 @@ def test_pooling_opt_arbitrary_dimensions():
def test_pooling_empty_batch():
img_shp = (0, 5, 6, 8)
img = T.ftensor4("img")
img = tt.ftensor4("img")
o = dnn.dnn_pool(img, (2, 2), (2, 2))
f = theano.function([img], o, mode=mode_with_gpu)
d = f(np.random.rand(*img_shp).astype("float32"))
assert d.shape == (0, 5, 3, 4)
g = T.grad(T.sum(o), wrt=img)
g = tt.grad(tt.sum(o), wrt=img)
f = theano.function([img], g, mode=mode_with_gpu)
d = f(np.random.rand(*img_shp).astype("float32"))
# Not sure what to assert, it should just pass, that's all.
......@@ -696,7 +707,7 @@ def test_pooling_empty_batch():
def test_dnn_tag():
# Test that if cudnn isn't avail we crash and that if it is avail, we use it.
x = T.tensor4()
x = tt.tensor4()
old = theano.config.on_opt_error
theano.config.on_opt_error = "raise"
......@@ -737,7 +748,7 @@ class TestDnnInferShapes(utt.InferShapeTester):
super().setup_method()
def test_softmax(self):
t = T.tensor4("t")
t = tt.tensor4("t")
rand_tensor = np.asarray(np.random.rand(5, 4, 3, 2), dtype=theano.config.floatX)
self._compile_and_check(
[t],
......@@ -748,7 +759,7 @@ class TestDnnInferShapes(utt.InferShapeTester):
self._compile_and_check(
[t],
[T.grad(dnn.GpuDnnSoftmax("accurate", "channel")(t).mean(), t)],
[tt.grad(dnn.GpuDnnSoftmax("accurate", "channel")(t).mean(), t)],
[rand_tensor],
dnn.GpuDnnSoftmaxGrad,
)
......@@ -815,9 +826,9 @@ class TestDnnInferShapes(utt.InferShapeTester):
dilations += [(2, 2)]
self._test_conv(
T.tensor4("img"),
T.tensor4("kerns"),
T.tensor4("out"),
tt.tensor4("img"),
tt.tensor4("kerns"),
tt.tensor4("out"),
np.random.rand(7, 2, 12, 16),
np.random.rand(8, 2, 4, 3),
border_mode,
......@@ -834,9 +845,9 @@ class TestDnnInferShapes(utt.InferShapeTester):
dilations = [(1, 1, 1), (2, 2, 2)] if dnn.version() >= 6000 else [(1, 1, 1)]
self._test_conv(
T.tensor5("img"),
T.tensor5("kerns"),
T.tensor5("out"),
tt.tensor5("img"),
tt.tensor5("kerns"),
tt.tensor5("out"),
np.random.rand(10, 2, 15, 16, 17),
np.random.rand(8, 2, 4, 3, 1),
border_mode,
......@@ -901,9 +912,9 @@ class TestDnnInferShapes(utt.InferShapeTester):
dilations = [(1, 1), (2, 2)] if dnn.version() >= 6000 else [(1, 1)]
self._test_conv_gradw(
T.tensor4("img"),
T.tensor4("topgrad"),
T.tensor4("kerns"),
tt.tensor4("img"),
tt.tensor4("topgrad"),
tt.tensor4("kerns"),
(5, 2, 6, 13),
(1, 2, 3, 7),
border_mode,
......@@ -913,9 +924,9 @@ class TestDnnInferShapes(utt.InferShapeTester):
)
def test_conv_gradi(self):
img = T.tensor4("img")
kerns = T.tensor4("kerns")
out = T.tensor4("out")
img = tt.tensor4("img")
kerns = tt.tensor4("kerns")
out = tt.tensor4("out")
kern_vals = np.asarray(np.random.rand(13, 4, 5, 6), dtype=theano.config.floatX)
out_vals = np.asarray(np.random.rand(3, 13, 9, 11), dtype=theano.config.floatX)
......@@ -948,7 +959,7 @@ class TestDnnInferShapes(utt.InferShapeTester):
)
def test_pool(self):
img = T.tensor4("img")
img = tt.tensor4("img")
img_val = np.asarray(np.random.rand(2, 3, 4, 5), dtype=theano.config.floatX)
modes = get_dnn_pool_modes()
......@@ -964,7 +975,7 @@ class TestDnnInferShapes(utt.InferShapeTester):
)
def test_pool_3d(self):
img = T.tensor5("img")
img = tt.tensor5("img")
img_val = np.asarray(np.random.rand(2, 3, 4, 5, 6), dtype=theano.config.floatX)
modes = get_dnn_pool_modes()
......@@ -980,9 +991,9 @@ class TestDnnInferShapes(utt.InferShapeTester):
)
def test_pool_grad(self):
img = T.tensor4("img")
img_grad = T.tensor4("img_grad")
out = T.tensor4("out")
img = tt.tensor4("img")
img_grad = tt.tensor4("img_grad")
out = tt.tensor4("out")
img_val = np.asarray(np.random.rand(2, 3, 4, 5), dtype=theano.config.floatX)
img_grad_val = np.asarray(
np.random.rand(2, 3, 4, 5), dtype=theano.config.floatX
......@@ -1006,9 +1017,9 @@ class TestDnnInferShapes(utt.InferShapeTester):
)
def test_pool_3d_grad(self):
img = T.tensor5("img")
img_grad = T.tensor5("img_grad")
out = T.tensor5("out")
img = tt.tensor5("img")
img_grad = tt.tensor5("img_grad")
out = tt.tensor5("out")
img_val = np.asarray(np.random.rand(2, 3, 4, 5, 6), dtype=theano.config.floatX)
img_grad_val = np.asarray(
np.random.rand(2, 3, 4, 5, 6), dtype=theano.config.floatX
......@@ -1034,8 +1045,8 @@ class TestDnnInferShapes(utt.InferShapeTester):
# this has been a problem in the past
def test_dnn_conv_border_mode():
img = T.tensor4()
kern = T.tensor4()
img = tt.tensor4()
kern = tt.tensor4()
dnn.dnn_conv(img, kern, border_mode=1)
dnn.dnn_conv(img, kern, border_mode=(2, 3))
......@@ -1047,9 +1058,9 @@ def test_dnn_conv_border_mode():
def test_dnn_conv_alpha_output_merge():
utt.seed_rng()
img = T.tensor4()
kern = T.tensor4()
out = T.tensor4()
img = tt.tensor4()
kern = tt.tensor4()
out = tt.tensor4()
b = 1
c = 4
......@@ -1313,7 +1324,7 @@ def test_conv3d_fwd():
flipped_filters = filters
# Compile a theano function for the reference implementation
conv_ref = theano.tensor.nnet.corr3d.Corr3dMM(
conv_ref = Corr3dMM(
border_mode=border_mode,
subsample=subsample,
filter_dilation=dilation,
......@@ -1365,7 +1376,7 @@ def test_conv3d_bwd():
conv_mode=conv_mode,
)
grad_i, grad_w = theano.tensor.grad(conv.sum(), [inputs, filters])
grad_i, grad_w = tt.grad(conv.sum(), [inputs, filters])
f = theano.function([], [grad_i, grad_w], mode=mode_with_gpu)
......@@ -1377,12 +1388,12 @@ def test_conv3d_bwd():
flipped_filters = filters
# Compile a theano function for the reference implementation
conv_ref = theano.tensor.nnet.corr3d.Corr3dMM(
conv_ref = Corr3dMM(
border_mode=border_mode,
subsample=subsample,
filter_dilation=dilation,
)(ref_cast(inputs), flipped_filters)
(grad_i_ref, grad_w_ref) = theano.tensor.grad(conv_ref.sum(), [inputs, filters])
(grad_i_ref, grad_w_ref) = tt.grad(conv_ref.sum(), [inputs, filters])
f_ref = theano.function([], [grad_i_ref, grad_w_ref], mode="FAST_RUN")
# Compare the results of the two implementations
......@@ -1418,22 +1429,22 @@ class TestSoftMax(test_nnet.TestSoftMax):
data = np.arange(np.product(dims), dtype=theano.config.floatX).reshape(dims)
# Verify the forward op
x_gpu = T.tensor4("x_gpu")
x_gpu = tt.tensor4("x_gpu")
f_gpu = dnn.GpuDnnSoftmax("accurate", "channel")(x_gpu)
f_gpu = theano.function([x_gpu], f_gpu, mode=self.mode)
assert f_gpu(data).shape == dims
# Verify the gradient op
dy_gpu = T.tensor4("dy_gpu")
sm_gpu = T.tensor4("sm_gpu")
dy_gpu = tt.tensor4("dy_gpu")
sm_gpu = tt.tensor4("sm_gpu")
f_grad_gpu = dnn.GpuDnnSoftmaxGrad("accurate", "channel")(dy_gpu, sm_gpu)
f_grad_gpu = theano.function([dy_gpu, sm_gpu], f_grad_gpu, mode=self.mode)
assert f_grad_gpu(data, data).shape == dims
def test_softmax_f16(self):
x = T.matrix("x", "float16")
x_gpu = T.tensor4("x_gpu", "float16")
f_z = T.nnet.softmax_op
x = tt.matrix("x", "float16")
x_gpu = tt.tensor4("x_gpu", "float16")
f_z = softmax_op
f_gpu = dnn.GpuDnnSoftmax("accurate", "channel")
def cmp(n, m, f, f_gpu):
......@@ -1455,15 +1466,15 @@ class TestSoftMax(test_nnet.TestSoftMax):
gout = np.asarray(f_gpu(gdata))[:, :, 0, 0]
utt.assert_allclose(out, gout)
x = T.matrix("x")
x_gpu = T.tensor4("x_gpu")
f_z = T.nnet.softmax_op
x = tt.matrix("x")
x_gpu = tt.tensor4("x_gpu")
f_z = softmax_op
f_gpu = dnn.GpuDnnSoftmax("accurate", "channel")
# Verify the grad operation
dims = (2, 3, 4, 5)
gdata = np.arange(np.product(dims), dtype=theano.config.floatX).reshape(dims)
T.verify_grad(f_gpu, [gdata], rng=np.random, mode=mode_with_gpu)
tt.verify_grad(f_gpu, [gdata], rng=np.random, mode=mode_with_gpu)
# Verify that the CPU and GPU implementations return the same results
# up to a tolerance.
......@@ -1474,65 +1485,34 @@ class TestSoftMax(test_nnet.TestSoftMax):
# Verify that the SoftmaxGrad -> Gpu[Dnn]SoftmaxGrad
# optimization is applied when cudnn is required
y = T.vector("y")
f = theano.function(
[y], T.grad(T.nnet.softmax(y).mean(), y), mode=mode_with_gpu
)
y = tt.vector("y")
f = theano.function([y], tt.grad(softmax(y).mean(), y), mode=mode_with_gpu)
sorted_f = f.maker.fgraph.toposort()
val = np.random.rand(5).astype(theano.config.floatX)
out_dnn = f(val)
assert len([i for i in sorted_f if isinstance(i.op, self.gpu_grad_op)]) == 1
assert (
len(
[
i
for i in sorted_f
if isinstance(i.op, theano.tensor.nnet.SoftmaxGrad)
]
)
== 0
)
assert len([i for i in sorted_f if isinstance(i.op, SoftmaxGrad)]) == 0
# Verify that the SoftmaxGrad -> Gpu[Dnn]SoftmaxGrad
# optimization is not applied when cudnn is excluded or not
# available
mode_wo_cudnn = mode_with_gpu.excluding("cudnn")
y = T.vector("y")
f = theano.function(
[y], T.grad(T.nnet.softmax(y).mean(), y), mode=mode_wo_cudnn
)
y = tt.vector("y")
f = theano.function([y], tt.grad(softmax(y).mean(), y), mode=mode_wo_cudnn)
sorted_f = f.maker.fgraph.toposort()
out_cpu = f(val)
utt.assert_allclose(out_dnn, out_cpu)
assert len([i for i in sorted_f if isinstance(i.op, self.gpu_grad_op)]) == 0
assert (
len(
[
i
for i in sorted_f
if isinstance(i.op, theano.tensor.nnet.SoftmaxGrad)
]
)
== 1
)
assert len([i for i in sorted_f if isinstance(i.op, SoftmaxGrad)]) == 1
# Verify that the SoftmaxGrad -> GpuDnnSoftmaxGrad do not
# crash with manual graph
y = T.vector("y")
o = theano.tensor.nnet.SoftmaxGrad()(y, y * 2)
y = tt.vector("y")
o = SoftmaxGrad()(y, y * 2)
f = theano.function([y], o, mode=mode_with_gpu)
sorted_f = f.maker.fgraph.toposort()
assert len([i for i in sorted_f if isinstance(i.op, self.gpu_grad_op)]) == 1
assert (
len(
[
i
for i in sorted_f
if isinstance(i.op, theano.tensor.nnet.SoftmaxGrad)
]
)
== 0
)
assert len([i for i in sorted_f if isinstance(i.op, SoftmaxGrad)]) == 0
@pytest.mark.skipif(
dnn.version(raises=False) < 3000, reason="Log-softmax is only in cudnn v3+"
......@@ -1540,9 +1520,9 @@ class TestSoftMax(test_nnet.TestSoftMax):
def test_log_softmax(self):
# This is a test for an optimization that depends on cuDNN v3 or
# more recent. Don't test if the cuDNN version is too old.
x = T.tensor4()
x = tt.tensor4()
softmax_out = dnn.GpuDnnSoftmax("accurate", "channel")(x)
log_out = T.log(T.as_tensor_variable(softmax_out))
log_out = tt.log(tt.as_tensor_variable(softmax_out))
f = theano.function([x], log_out, mode=mode_with_gpu)
......@@ -1585,11 +1565,11 @@ class TestSoftMax(test_nnet.TestSoftMax):
# Compile a reference function, on the CPU, to be used to validate the
# results of the other function.
x = T.matrix()
f_ref = theano.function([x], T.nnet.LogSoftmax()(x))
x = tt.matrix()
f_ref = theano.function([x], LogSoftmax()(x))
# Build the first graph and ensure that the optimization is applied
log_softmax_out = T.nnet.LogSoftmax()(x)
log_softmax_out = LogSoftmax()(x)
f = theano.function([x], log_softmax_out, mode=mode_with_gpu)
dnn_softmax_nodes = [
......@@ -1603,7 +1583,7 @@ class TestSoftMax(test_nnet.TestSoftMax):
utt.assert_allclose(f(inp), f_ref(inp))
# Build the first graph and ensure that the optimization is applied
log_softmax_out = T.log(T.nnet.Softmax()(x))
log_softmax_out = tt.log(Softmax()(x))
f = theano.function([x], log_softmax_out, mode=mode_with_gpu)
dnn_softmax_nodes = [
......@@ -1618,7 +1598,7 @@ class TestSoftMax(test_nnet.TestSoftMax):
def dnn_reduction(nd, idtype, acc_dtype, odtype):
inp = T.TensorType(idtype, (False,) * nd)()
inp = tt.TensorType(idtype, (False,) * nd)()
res = inp.sum(acc_dtype=acc_dtype, dtype=odtype)
f = theano.function([inp], res, mode=mode_with_gpu)
assert any(
......@@ -1641,7 +1621,7 @@ def test_dnn_reduction_opt():
@pytest.mark.skipif(dnn.version(raises=False) < 6000, reason=dnn.dnn_available.msg)
def test_dnn_reduction_sum_squares():
M = T.matrix()
M = tt.matrix()
for axis in (None, 0, 1):
out = (M ** 2).sum(axis=axis)
f = theano.function([M], out, mode=mode_with_gpu)
......@@ -1655,7 +1635,7 @@ def test_dnn_reduction_sum_squares():
@pytest.mark.skipif(dnn.version(raises=False) < 6000, reason=dnn.dnn_available.msg)
def test_dnn_reduction_sum_abs():
M = T.matrix()
M = tt.matrix()
for axis in (None, 0, 1):
out = abs(M).sum(axis=axis)
f = theano.function([M], out, mode=mode_with_gpu)
......@@ -1669,7 +1649,7 @@ def test_dnn_reduction_sum_abs():
@pytest.mark.skipif(dnn.version(raises=False) < 6000, reason=dnn.dnn_available.msg)
def test_dnn_reduction_absmax():
M = T.matrix()
M = tt.matrix()
for axis in (None, 0, 1):
out = abs(M).max(axis=axis)
f = theano.function([M], out, mode=mode_with_gpu)
......@@ -1692,9 +1672,7 @@ def test_dnn_reduction_axis_size_one():
[(4, 1, 6, 1), (1, 3)],
]:
x = theano.tensor.TensorType(
dtype=dtype, broadcastable=[False] * len(shape)
)()
x = tt.TensorType(dtype=dtype, broadcastable=[False] * len(shape))()
sum = x.sum(axis=axis)
sum_squares = (x ** 2).sum(axis=axis)
sum_abs = abs(x).sum(axis=axis)
......@@ -1779,9 +1757,9 @@ def test_dnn_reduction_error():
slow_output = np.sum(slow_output.transpose(), axis=1)
vecT = T.vector(dtype=theano.config.floatX)
outputT = T.alloc(2.0 * vecT, 5, vecT.shape[0])
outputSummedT = T.sum(T.transpose(outputT), axis=1)
vecT = tt.vector(dtype=theano.config.floatX)
outputT = tt.alloc(2.0 * vecT, 5, vecT.shape[0])
outputSummedT = tt.sum(tt.transpose(outputT), axis=1)
f3 = theano.function(inputs=[vecT], outputs=outputSummedT)
output = f3(vec)
......@@ -1789,8 +1767,8 @@ def test_dnn_reduction_error():
def dnn_maxargmax(nd, idtype, axis):
inp = T.TensorType(idtype, (False,) * nd)()
res = T.max_and_argmax(inp, axis=axis)
inp = tt.TensorType(idtype, (False,) * nd)()
res = tt.max_and_argmax(inp, axis=axis)
f = theano.function([inp], res, mode=mode_with_gpu)
assert any(
isinstance(n.op, dnn.GpuDnnReduction) for n in f.maker.fgraph.apply_nodes
......@@ -1819,7 +1797,14 @@ def test_dnn_batchnorm_train():
utt.seed_rng()
for mode in ("per-activation", "spatial"):
for vartype in (T.tensor6, T.tensor5, T.tensor4, T.tensor3, T.matrix, T.vector):
for vartype in (
tt.tensor6,
tt.tensor5,
tt.tensor4,
tt.tensor3,
tt.matrix,
tt.vector,
):
x, scale, bias, running_mean, running_var = (
vartype(n)
for n in ("x", "scale", "bias", "running_mean", "running_var")
......@@ -1869,10 +1854,10 @@ def test_dnn_batchnorm_train():
axes = (0,) + tuple(range(2, ndim))
x_mean_ref = x.mean(axis=axes, keepdims=True)
x_var_ref = x.var(axis=axes, keepdims=True)
x_invstd_ref = T.inv(T.sqrt(x_var_ref + eps))
scale_ref = T.addbroadcast(scale, *axes)
bias_ref = T.addbroadcast(bias, *axes)
m = T.cast(T.prod(x.shape) / T.prod(scale.shape), theano.config.floatX)
x_invstd_ref = tt.inv(tt.sqrt(x_var_ref + eps))
scale_ref = tt.addbroadcast(scale, *axes)
bias_ref = tt.addbroadcast(bias, *axes)
m = tt.cast(tt.prod(x.shape) / tt.prod(scale.shape), theano.config.floatX)
out_ref = (x - x_mean_ref) * (scale_ref * x_invstd_ref) + bias_ref
out_running_mean_ref = (
running_mean * (1 - running_average_factor)
......@@ -1884,12 +1869,12 @@ def test_dnn_batchnorm_train():
)
# backward pass
dy = vartype("dy")
grads_gpu = T.grad(None, wrt=[x, scale, bias], known_grads={out_gpu: dy})
grads_abstract = T.grad(
grads_gpu = tt.grad(None, wrt=[x, scale, bias], known_grads={out_gpu: dy})
grads_abstract = tt.grad(
None, wrt=[x, scale, bias], known_grads={out_abstract: dy}
)
# reference backward pass
grads_ref = T.grad(None, wrt=[x, scale, bias], known_grads={out_ref: dy})
grads_ref = tt.grad(None, wrt=[x, scale, bias], known_grads={out_ref: dy})
# compile
f_gpu = theano.function(
[x, scale, bias, running_mean, running_var, dy],
......@@ -2011,10 +1996,10 @@ def test_dnn_batchnorm_train_without_running_averages():
utt.seed_rng()
x, scale, bias, dy = (
T.tensor4("x"),
T.tensor4("scale"),
T.tensor4("bias"),
T.tensor4("dy"),
tt.tensor4("x"),
tt.tensor4("scale"),
tt.tensor4("bias"),
tt.tensor4("dy"),
)
data_shape = (5, 10, 30, 25)
param_shape = (1, 10, 30, 25)
......@@ -2027,8 +2012,8 @@ def test_dnn_batchnorm_train_without_running_averages():
x, scale, bias, "per-activation"
)
# backward pass
grads_gpu = T.grad(None, wrt=[x, scale, bias], known_grads={out_gpu: dy})
grads_abstract = T.grad(None, wrt=[x, scale, bias], known_grads={out_abstract: dy})
grads_gpu = tt.grad(None, wrt=[x, scale, bias], known_grads={out_gpu: dy})
grads_abstract = tt.grad(None, wrt=[x, scale, bias], known_grads={out_abstract: dy})
# compile
f_gpu = theano.function(
[x, scale, bias, dy],
......@@ -2081,10 +2066,10 @@ def test_without_dnn_batchnorm_train_without_running_averages():
utt.seed_rng()
x, scale, bias, dy = (
T.tensor4("x"),
T.tensor4("scale"),
T.tensor4("bias"),
T.tensor4("dy"),
tt.tensor4("x"),
tt.tensor4("scale"),
tt.tensor4("bias"),
tt.tensor4("dy"),
)
data_shape = (5, 10, 30, 25)
param_shape = (1, 10, 30, 25)
......@@ -2094,7 +2079,7 @@ def test_without_dnn_batchnorm_train_without_running_averages():
x, scale, bias, "per-activation"
)
# backward pass
grads_abstract = T.grad(None, wrt=[x, scale, bias], known_grads={out_abstract: dy})
grads_abstract = tt.grad(None, wrt=[x, scale, bias], known_grads={out_abstract: dy})
# compile
f_abstract = theano.function(
[x, scale, bias, dy],
......@@ -2143,7 +2128,7 @@ def test_dnn_batchnorm_train_inplace():
# test inplace_running_mean and inplace_running_var
utt.seed_rng()
x, scale, bias = T.tensor4("x"), T.tensor4("scale"), T.tensor4("bias")
x, scale, bias = tt.tensor4("x"), tt.tensor4("scale"), tt.tensor4("bias")
data_shape = (5, 10, 30, 25)
param_shape = (1, 10, 30, 25)
running_mean = gpuarray_shared_constructor(
......@@ -2199,7 +2184,14 @@ def test_batchnorm_inference():
utt.seed_rng()
for mode in ("per-activation", "spatial"):
for vartype in (T.tensor6, T.tensor5, T.tensor4, T.tensor3, T.matrix, T.vector):
for vartype in (
tt.tensor6,
tt.tensor5,
tt.tensor4,
tt.tensor3,
tt.matrix,
tt.vector,
):
x, scale, bias, mean, var = (
vartype(n) for n in ("x", "scale", "bias", "mean", "var")
)
......@@ -2220,19 +2212,19 @@ def test_batchnorm_inference():
elif mode == "spatial":
axes = (0,) + tuple(range(2, ndim))
scale_ref, bias_ref, mean_ref, var_ref = (
T.addbroadcast(t, *axes) for t in (scale, bias, mean, var)
tt.addbroadcast(t, *axes) for t in (scale, bias, mean, var)
)
out_ref = (x - mean_ref) * (scale_ref / T.sqrt(var_ref + eps)) + bias_ref
out_ref = (x - mean_ref) * (scale_ref / tt.sqrt(var_ref + eps)) + bias_ref
# backward pass
dy = vartype("dy")
grads_gpu = T.grad(
grads_gpu = tt.grad(
None, wrt=[x, scale, bias, mean, var], known_grads={out_gpu: dy}
)
grads_abstract = T.grad(
grads_abstract = tt.grad(
None, wrt=[x, scale, bias, mean, var], known_grads={out_abstract: dy}
)
# reference backward pass
grads_ref = T.grad(
grads_ref = tt.grad(
None, wrt=[x, scale, bias, mean, var], known_grads={out_ref: dy}
)
# compile
......@@ -2318,7 +2310,7 @@ def test_batchnorm_inference_inplace():
utt.seed_rng()
x, scale, bias, mean, var = (
T.tensor4(n) for n in ("x", "scale", "bias", "mean", "var")
tt.tensor4(n) for n in ("x", "scale", "bias", "mean", "var")
)
data_shape = (5, 10, 30, 25)
param_shape = (1, 10, 30, 25)
......@@ -2345,7 +2337,7 @@ def test_batchnorm_inference_inplace():
def test_dnn_batchnorm_valid_and_invalid_axes():
for vartype in (T.tensor5, T.tensor4, T.tensor3, T.matrix):
for vartype in (tt.tensor5, tt.tensor4, tt.tensor3, tt.matrix):
x, scale, bias, mean, var, dy = (
vartype(n) for n in ("x", "scale", "bias", "mean", "var", "dy")
)
......@@ -2363,10 +2355,10 @@ def test_dnn_batchnorm_valid_and_invalid_axes():
out_test = bn.batch_normalization_test(x, scale, bias, mean, var, axes)
# backward pass
dy = vartype("dy")
grads_train = T.grad(
grads_train = tt.grad(
None, wrt=[x, scale, bias], known_grads={out_train: dy}
)
grads_test = T.grad(
grads_test = tt.grad(
None, wrt=[x, scale, bias, mean, var], known_grads={out_test: dy}
)
# compile
......@@ -2439,9 +2431,9 @@ def test_dnn_rnn_gru():
timesteps = 5
# test code
X = T.tensor3("X")
Y = T.tensor3("Y")
h0 = T.tensor3("h0")
X = tt.tensor3("X")
Y = tt.tensor3("Y")
h0 = tt.tensor3("h0")
rnnb = dnn.RNNBlock(theano.config.floatX, hidden_dim, depth, "gru")
psize = rnnb.get_param_size([batch_size, input_dim])
......@@ -2465,10 +2457,10 @@ def test_dnn_rnn_gru():
def funcs(out, params, hy=None):
cost = 0
if out:
cost += T.mean((Y - out) ** 2)
cost += tt.mean((Y - out) ** 2)
if hy:
cost += T.mean(hy ** 2)
grad = T.grad(cost, [X, h0] + params)
cost += tt.mean(hy ** 2)
grad = tt.grad(cost, [X, h0] + params)
grad_fn = theano.function(
[X, Y, h0], grad, mode=mode_with_gpu, on_unused_input="ignore"
)
......@@ -2477,7 +2469,7 @@ def test_dnn_rnn_gru():
ref_y = last_layer.output()
# This will grab the hy from the scan implementation
ref_hy = T.stack(
ref_hy = tt.stack(
[model.layers[0].Y[-1], model.layers[1].Y[-1], model.layers[2].Y[-1]]
)
......@@ -2548,9 +2540,9 @@ def test_dnn_rnn_gru_bidi():
timesteps = 5
# test code
X = T.tensor3("X")
Y = T.tensor3("Y")
h0 = T.tensor3("h0")
X = tt.tensor3("X")
Y = tt.tensor3("Y")
h0 = tt.tensor3("h0")
rnnb = dnn.RNNBlock(
theano.config.floatX, hidden_dim, depth, "gru", direction_mode="bidirectional"
......@@ -2563,10 +2555,10 @@ def test_dnn_rnn_gru_bidi():
def funcs(out, params, hy=None):
cost = 0
if out:
cost += T.mean((Y - out) ** 2)
cost += tt.mean((Y - out) ** 2)
if hy:
cost += T.mean(hy ** 2)
grad = T.grad(cost, [X, h0] + params)
cost += tt.mean(hy ** 2)
grad = tt.grad(cost, [X, h0] + params)
grad_fn = theano.function(
[X, Y, h0], grad, mode=mode_with_gpu, on_unused_input="ignore"
)
......@@ -2609,10 +2601,10 @@ def test_dnn_rnn_lstm():
timesteps = 5
# test code
X = T.tensor3("X")
Y = T.tensor3("Y")
h0 = T.tensor3("h0")
c0 = T.tensor3("c0")
X = tt.tensor3("X")
Y = tt.tensor3("Y")
h0 = tt.tensor3("h0")
c0 = tt.tensor3("c0")
rnnb = dnn.RNNBlock(theano.config.floatX, hidden_dim, depth, "lstm")
psize = rnnb.get_param_size([batch_size, input_dim])
......@@ -2635,8 +2627,8 @@ def test_dnn_rnn_lstm():
def funcs(out, params):
fn = theano.function([X, h0, c0], out, mode=mode_with_gpu)
cost = T.mean((Y - out) ** 2)
grad = T.grad(cost, [X, h0, c0] + params)
cost = tt.mean((Y - out) ** 2)
grad = tt.grad(cost, [X, h0, c0] + params)
grad_fn = theano.function([X, Y, h0, c0], grad, mode=mode_with_gpu)
return fn, grad_fn
......@@ -2695,10 +2687,10 @@ def test_dnn_rnn_lstm_grad_c():
timesteps = 5
# test code
X = T.tensor3("X")
CY = T.tensor3("CY")
h0 = T.tensor3("h0")
c0 = T.tensor3("c0")
X = tt.tensor3("X")
CY = tt.tensor3("CY")
h0 = tt.tensor3("h0")
c0 = tt.tensor3("c0")
rnnb = dnn.RNNBlock(theano.config.floatX, hidden_dim, depth, "lstm")
psize = rnnb.get_param_size([batch_size, input_dim])
......@@ -2720,13 +2712,13 @@ def test_dnn_rnn_lstm_grad_c():
p[:] = layer_params[j].get_value(borrow=True, return_internal_type=True)
def funcs(out, params):
cost = T.mean((CY - out) ** 2)
grad = T.grad(cost, [X, h0, c0] + params)
cost = tt.mean((CY - out) ** 2)
grad = tt.grad(cost, [X, h0, c0] + params)
grad_fn = theano.function([X, CY, h0, c0], grad, mode=mode_with_gpu)
return grad_fn
_, _, cy = rnnb.apply(params_cudnn, X, h0, c0)
ref_cy = T.stack(
ref_cy = tt.stack(
[model.layers[0].C[-1], model.layers[1].C[-1], model.layers[2].C[-1]]
)
......@@ -2797,14 +2789,14 @@ def test_dnn_spatialtf():
def spatialtf_cpu(inp, theta, scale_height, scale_width, border_mode="nearest"):
num_batch, num_channels, height, width = inp.shape
theta = T.reshape(theta, (-1, 2, 3))
theta = tt.reshape(theta, (-1, 2, 3))
# grid of (x_t, y_t, 1), eq (1) in ref [1]
out_height = T.cast(T.ceil(height * scale_height), "int64")
out_width = T.cast(T.ceil(width * scale_width), "int64")
out_height = tt.cast(tt.ceil(height * scale_height), "int64")
out_width = tt.cast(tt.ceil(width * scale_width), "int64")
grid = _meshgrid(out_height, out_width)
# transform a x (x_t, y_t, 1)^t -> (x_s, y_s)
t_g = T.dot(theta, grid)
t_g = tt.dot(theta, grid)
x_s = t_g[:, 0]
y_s = t_g[:, 1]
x_s_flat = x_s.flatten()
......@@ -2816,7 +2808,7 @@ def test_dnn_spatialtf():
input_dim, x_s_flat, y_s_flat, out_height, out_width, border_mode
)
output = T.reshape(
output = tt.reshape(
input_transformed, (num_batch, out_height, out_width, num_channels)
)
output = output.dimshuffle(0, 3, 1, 2) # dimshuffle to conv format
......@@ -2825,8 +2817,8 @@ def test_dnn_spatialtf():
def _interpolate(im, x, y, out_height, out_width, border_mode):
# *_f are floats
num_batch, height, width, channels = im.shape
height_f = T.cast(height, theano.config.floatX)
width_f = T.cast(width, theano.config.floatX)
height_f = tt.cast(height, theano.config.floatX)
width_f = tt.cast(width, theano.config.floatX)
# scale coordinates from [-1, 1] to [0, dimension - 1], where dimension
# can be the width or height
......@@ -2835,42 +2827,42 @@ def test_dnn_spatialtf():
# obtain indices of the 2x2 pixel neighborhood surrounding the coordinates;
# we need those in floatX for interpolation and in int64 for indexing.
x0_f = T.floor(x)
y0_f = T.floor(y)
x0_f = tt.floor(x)
y0_f = tt.floor(y)
x1_f = x0_f + 1
y1_f = y0_f + 1
# for indexing, we need to take care of the border mode for outside pixels.
if border_mode == "nearest":
x0 = T.clip(x0_f, 0, width_f - 1)
x1 = T.clip(x1_f, 0, width_f - 1)
y0 = T.clip(y0_f, 0, height_f - 1)
y1 = T.clip(y1_f, 0, height_f - 1)
x0 = tt.clip(x0_f, 0, width_f - 1)
x1 = tt.clip(x1_f, 0, width_f - 1)
y0 = tt.clip(y0_f, 0, height_f - 1)
y1 = tt.clip(y1_f, 0, height_f - 1)
elif border_mode == "mirror":
w = 2 * (width_f - 1)
x0 = T.minimum(x0_f % w, -x0_f % w)
x1 = T.minimum(x1_f % w, -x1_f % w)
x0 = tt.minimum(x0_f % w, -x0_f % w)
x1 = tt.minimum(x1_f % w, -x1_f % w)
h = 2 * (height_f - 1)
y0 = T.minimum(y0_f % h, -y0_f % h)
y1 = T.minimum(y1_f % h, -y1_f % h)
y0 = tt.minimum(y0_f % h, -y0_f % h)
y1 = tt.minimum(y1_f % h, -y1_f % h)
elif border_mode == "wrap":
x0 = T.mod(x0_f, width_f)
x1 = T.mod(x1_f, width_f)
y0 = T.mod(y0_f, height_f)
y1 = T.mod(y1_f, height_f)
x0 = tt.mod(x0_f, width_f)
x1 = tt.mod(x1_f, width_f)
y0 = tt.mod(y0_f, height_f)
y1 = tt.mod(y1_f, height_f)
else:
raise ValueError(
"border_mode must be one of " "'nearest', 'mirror', 'wrap'"
)
x0, x1, y0, y1 = (T.cast(v, "int64") for v in (x0, x1, y0, y1))
x0, x1, y0, y1 = (tt.cast(v, "int64") for v in (x0, x1, y0, y1))
# The input is [num_batch, height, width, channels]. We do the lookup in
# the flattened input, i.e [num_batch*height*width, channels]. We need
# to offset all indices to match the flat version
dim2 = width
dim1 = width * height
base = T.repeat(
T.arange(num_batch, dtype="int64") * dim1, out_height * out_width
base = tt.repeat(
tt.arange(num_batch, dtype="int64") * dim1, out_height * out_width
)
base_y0 = base + y0 * dim2
base_y1 = base + y1 * dim2
......@@ -2891,16 +2883,16 @@ def test_dnn_spatialtf():
wb = ((x1_f - x) * (y - y0_f)).dimshuffle(0, "x")
wc = ((x - x0_f) * (y1_f - y)).dimshuffle(0, "x")
wd = ((x - x0_f) * (y - y0_f)).dimshuffle(0, "x")
output = T.sum([wa * Ia, wb * Ib, wc * Ic, wd * Id], axis=0)
output = tt.sum([wa * Ia, wb * Ib, wc * Ic, wd * Id], axis=0)
return output
def _linspace(start, stop, num):
# Theano linspace. Behaves similar to np.linspace
start = T.cast(start, theano.config.floatX)
stop = T.cast(stop, theano.config.floatX)
num = T.cast(num, theano.config.floatX)
start = tt.cast(start, theano.config.floatX)
stop = tt.cast(stop, theano.config.floatX)
num = tt.cast(num, theano.config.floatX)
step = (stop - start) / (num - 1)
return T.arange(num, dtype=theano.config.floatX) * step + start
return tt.arange(num, dtype=theano.config.floatX) * step + start
def _meshgrid(height, width):
# This function is the grid generator from eq. (1) in reference [1].
......@@ -2913,13 +2905,17 @@ def test_dnn_spatialtf():
# Note: If the image size is known at layer construction time, we could
# compute the meshgrid offline in numpy instead of doing it dynamically
# in Theano. However, it hardly affected performance when we tried.
x_t = T.dot(T.ones((height, 1)), _linspace(-1.0, 1.0, width).dimshuffle("x", 0))
y_t = T.dot(_linspace(-1.0, 1.0, height).dimshuffle(0, "x"), T.ones((1, width)))
x_t = tt.dot(
tt.ones((height, 1)), _linspace(-1.0, 1.0, width).dimshuffle("x", 0)
)
y_t = tt.dot(
_linspace(-1.0, 1.0, height).dimshuffle(0, "x"), tt.ones((1, width))
)
x_t_flat = x_t.reshape((1, -1))
y_t_flat = y_t.reshape((1, -1))
ones = T.ones_like(x_t_flat)
grid = T.concatenate([x_t_flat, y_t_flat, ones], axis=0)
ones = tt.ones_like(x_t_flat)
grid = tt.concatenate([x_t_flat, y_t_flat, ones], axis=0)
return grid
img_dims = (5, 3, 16, 16)
......@@ -2933,8 +2929,8 @@ def test_dnn_spatialtf():
theta = np.asarray(img_dims[0] * [transform], dtype=theano.config.floatX)
# Create symbolic variables for inputs and transformations
t_img = T.tensor4("img")
t_theta = T.tensor3("theta")
t_img = tt.tensor4("img")
t_theta = tt.tensor3("theta")
st_dnn = dnn.dnn_spatialtf(
t_img, t_theta, scale_height=scale_height, scale_width=scale_width
......@@ -2963,8 +2959,8 @@ def test_dnn_spatialtf():
def test_dnn_spatialtf_invalid_shapes():
inputs = T.tensor4("inputs")
theta = T.tensor3("theta")
inputs = tt.tensor4("inputs")
theta = tt.tensor3("theta")
st_dnn = dnn.dnn_spatialtf(inputs, theta)
st_dnn_func = theano.function([inputs, theta], st_dnn, mode=mode_with_gpu)
......@@ -2994,13 +2990,13 @@ def test_dnn_spatialtf_invalid_shapes():
def test_dnn_spatialtf_grad():
utt.seed_rng()
inputs = T.tensor4("inputs")
theta = T.tensor3("theta")
inputs = tt.tensor4("inputs")
theta = tt.tensor3("theta")
out = dnn.dnn_spatialtf(inputs, theta, scale_height=0.25, scale_width=0.75)
out_mean = T.mean(out)
mean_gi = T.grad(out_mean, [inputs])
mean_gt = T.grad(out_mean, [theta])
out_mean = tt.mean(out)
mean_gi = tt.grad(out_mean, [inputs])
mean_gt = tt.grad(out_mean, [theta])
f_gi = theano.function([inputs, theta], mean_gi, mode=mode_with_gpu)
assert any(
......@@ -3053,7 +3049,7 @@ def test_dnn_spatialtf_grad():
class TestDnnConv2DRuntimeAlgorithms(object):
ndim = 2
cpu_conv_class = theano.tensor.nnet.corr.CorrMM
cpu_conv_class = CorrMM
runtime_shapes = [
(3, [(2, 3, 10, 9), (5, 3, 7, 7)]),
(1, [(1, 1, 100, 200), (1, 1, 50, 200)]),
......@@ -3080,8 +3076,8 @@ class TestDnnConv2DRuntimeAlgorithms(object):
_broadcastable = [False] * (2 + self.ndim)
def run_fwd_runtime_algorithm(algo):
inputs = theano.tensor.TensorType(dtype, _broadcastable)()
filters = theano.tensor.TensorType(dtype, _broadcastable)()
inputs = tt.TensorType(dtype, _broadcastable)()
filters = tt.TensorType(dtype, _broadcastable)()
# Scale down the input values to prevent very large absolute errors
# due to float rounding
lower_inputs = inputs / 10
......@@ -3127,8 +3123,8 @@ class TestDnnConv2DRuntimeAlgorithms(object):
def run_gradinput_runtime_algorithm(algo):
theano.config.dnn.conv.algo_bwd_data = algo
inputs = theano.tensor.TensorType(dtype, _broadcastable)()
filters = theano.tensor.TensorType(dtype, _broadcastable)()
inputs = tt.TensorType(dtype, _broadcastable)()
filters = tt.TensorType(dtype, _broadcastable)()
conv = dnn.dnn_conv(
img=inputs,
kerns=filters,
......@@ -3137,7 +3133,7 @@ class TestDnnConv2DRuntimeAlgorithms(object):
subsample=unit_shape,
dilation=unit_shape,
)
(grad_i,) = theano.tensor.grad(conv.sum(), [inputs])
(grad_i,) = tt.grad(conv.sum(), [inputs])
f = theano.function([inputs, filters], grad_i, mode=mode_with_gpu)
assert 1 == len(
[
......@@ -3161,7 +3157,7 @@ class TestDnnConv2DRuntimeAlgorithms(object):
conv_ref = self.cpu_conv_class(subsample=unit_shape)(
ref_cast(inputs), flipped_filters
)
(grad_i_ref,) = theano.tensor.grad(conv_ref.sum(), [inputs])
(grad_i_ref,) = tt.grad(conv_ref.sum(), [inputs])
f_ref = theano.function([inputs, filters], grad_i_ref, mode="FAST_RUN")
runtime_shapes = self.runtime_shapes
if algo in ("time_once", "guess_once"):
......@@ -3185,8 +3181,8 @@ class TestDnnConv2DRuntimeAlgorithms(object):
def run_gradweight_runtime_algorithm(algo):
theano.config.dnn.conv.algo_bwd_filter = algo
inputs = theano.tensor.TensorType(dtype, _broadcastable)()
filters = theano.tensor.TensorType(dtype, _broadcastable)()
inputs = tt.TensorType(dtype, _broadcastable)()
filters = tt.TensorType(dtype, _broadcastable)()
conv = dnn.dnn_conv(
img=inputs,
kerns=filters,
......@@ -3195,7 +3191,7 @@ class TestDnnConv2DRuntimeAlgorithms(object):
subsample=unit_shape,
dilation=unit_shape,
)
(grad_w,) = theano.tensor.grad(conv.sum(), [filters])
(grad_w,) = tt.grad(conv.sum(), [filters])
f = theano.function([inputs, filters], grad_w, mode=mode_with_gpu)
assert 1 == len(
[
......@@ -3219,7 +3215,7 @@ class TestDnnConv2DRuntimeAlgorithms(object):
conv_ref = self.cpu_conv_class(subsample=unit_shape)(
ref_cast(inputs), flipped_filters
)
(grad_w_ref,) = theano.tensor.grad(conv_ref.sum(), [filters])
(grad_w_ref,) = tt.grad(conv_ref.sum(), [filters])
f_ref = theano.function([inputs, filters], grad_w_ref, mode="FAST_RUN")
runtime_shapes = self.runtime_shapes
if algo in ("time_once", "guess_once"):
......@@ -3239,7 +3235,7 @@ class TestDnnConv2DRuntimeAlgorithms(object):
class TestDnnConv3DRuntimeAlgorithms(TestDnnConv2DRuntimeAlgorithms):
ndim = 3
cpu_conv_class = theano.tensor.nnet.corr3d.Corr3dMM
cpu_conv_class = Corr3dMM
runtime_shapes = [
(3, [(2, 3, 5, 10, 9), (5, 3, 4, 7, 7)]),
(1, [(1, 1, 5, 100, 200), (1, 1, 4, 50, 200)]),
......@@ -3293,9 +3289,9 @@ def test_conv_guess_once_with_dtypes():
def test_opt_f16_prec32():
inputs = T.TensorType("float16", (False,) * 4)()
filters = T.TensorType("float16", (False,) * 4)()
conv = T.nnet.conv2d(inputs, filters)
inputs = tt.TensorType("float16", (False,) * 4)()
filters = tt.TensorType("float16", (False,) * 4)()
conv = conv2d(inputs, filters)
gfilt = theano.grad(conv.sum(), filters)
......
tests/gpuarray/test_extra_ops.py
浏览文件 @ b4dc02d6
......@@ -3,12 +3,11 @@ import pytest
import numpy as np
import theano
import theano.tensor as tt
from functools import partial
from itertools import product
from theano import tensor as T
from theano.tensor.extra_ops import CumOp
from theano.gpuarray.extra_ops import GpuCumOp
from theano.gpuarray.type import get_context
......@@ -33,13 +32,13 @@ class TestGpuCumOp(TestCumOp):
# The CPU implementation is not so accurate, which throws out DebugMode.
# Since propagating .tag.values_eq_approx to the output of every
# GpuFromHost seems overkill, we just relax the rtol for these tests
self.old_rtol = theano.tensor.float32_rtol
theano.tensor.basic.float32_rtol *= 2
self.old_rtol = tt.float32_rtol
tt.float32_rtol *= 2
def teardown_method(self):
super().teardown_method()
# Restore rtol
theano.tensor.basic.float32_rtol = self.old_rtol
tt.float32_rtol = self.old_rtol
@pytest.mark.skipif(
theano.config.floatX != "float32",
......@@ -48,7 +47,7 @@ class TestGpuCumOp(TestCumOp):
@pytest.mark.parametrized("mode", ["mul", "add"])
def test_infer_shape(self, mode):
op_class = partial(self.op_class, mode=mode)
x = T.tensor3("x")
x = tt.tensor3("x")
a = np.random.random((3, 5, 2)).astype(theano.config.floatX)
for axis in range(-len(a.shape), len(a.shape)):
......@@ -58,7 +57,7 @@ class TestGpuCumOp(TestCumOp):
def test_Strides1D(self, mode):
op_class = partial(self.op_class, mode=mode)
np_func = dict(add=np.cumsum, mul=np.cumprod)[mode]
x = T.fvector("x")
x = tt.fvector("x")
for axis in [0, None, -1]:
a = np.random.random((42,)).astype("float32")
......@@ -89,7 +88,7 @@ class TestGpuCumOp(TestCumOp):
def test_Strides2D(self, mode):
np_func = dict(add=np.cumsum, mul=np.cumprod)[mode]
op_class = partial(self.op_class, mode=mode)
x = T.fmatrix("x")
x = tt.fmatrix("x")
for axis in [0, 1, None, -1, -2]:
a = np.random.random((42, 30)).astype("float32")
......@@ -120,7 +119,7 @@ class TestGpuCumOp(TestCumOp):
def test_Strides3D(self, mode):
np_func = dict(add=np.cumsum, mul=np.cumprod)[mode]
op_class = partial(self.op_class, mode=mode)
x = T.ftensor3("x")
x = tt.ftensor3("x")
for axis in [0, 1, 2, None, -1, -2, -3]:
a = np.random.random((42, 30, 25)).astype("float32")
......@@ -153,7 +152,7 @@ class TestGpuCumOp(TestCumOp):
op_class = partial(self.op_class, mode=mode)
block_max_size = self.max_threads_dim0 * 2
x = T.fvector("x")
x = tt.fvector("x")
f = theano.function([x], op_class(axis=0)(x), mode=self.mode)
assert [n for n in f.maker.fgraph.toposort() if isinstance(n.op, GpuCumOp)]
......@@ -176,7 +175,7 @@ class TestGpuCumOp(TestCumOp):
op_class = partial(self.op_class, mode=mode)
block_max_size = self.max_threads_dim0 * 2
x = T.fmatrix("x")
x = tt.fmatrix("x")
for shape_axis, axis in zip([0, 1, 0, 1, 0], [0, 1, None, -1, -2]):
f = theano.function([x], op_class(axis=axis)(x), mode=self.mode)
assert [n for n in f.maker.fgraph.toposort() if isinstance(n.op, GpuCumOp)]
......@@ -217,7 +216,7 @@ class TestGpuCumOp(TestCumOp):
op_class = partial(self.op_class, mode=mode)
block_max_size = self.max_threads_dim0 * 2
x = T.ftensor3("x")
x = tt.ftensor3("x")
for shape_axis, axis in zip([0, 1, 2, 0, 2, 1, 0], [0, 1, 2, None, -1, -2, -3]):
f = theano.function([x], op_class(axis=axis)(x), mode=self.mode)
assert [n for n in f.maker.fgraph.toposort() if isinstance(n.op, GpuCumOp)]
......@@ -267,6 +266,6 @@ class TestGpuCumOp(TestCumOp):
def test_GpuCumOp4D(self, mode):
op_class = partial(self.op_class, mode=mode)
# Should not use the GPU version.
x = T.ftensor4("x")
x = tt.ftensor4("x")
f = theano.function([x], op_class(axis=1)(x), mode=self.mode)
assert [n for n in f.maker.fgraph.toposort() if isinstance(n.op, CumOp)]
tests/gpuarray/test_fft.py
浏览文件 @ b4dc02d6
......@@ -3,7 +3,7 @@ import numpy as np
import pytest
import theano
import theano.tensor as T
import theano.tensor as tt
import theano.gpuarray.fft
from theano.gpuarray.fft import pygpu_available, skcuda_available, pycuda_available
......@@ -27,7 +27,7 @@ class TestFFT:
def test_1Dfft(self):
inputs_val = np.random.random((1, N)).astype("float32")
x = T.matrix("x", dtype="float32")
x = tt.matrix("x", dtype="float32")
rfft = theano.gpuarray.fft.curfft(x)
f_rfft = theano.function([x], rfft, mode=mode_with_gpu)
res_rfft = f_rfft(inputs_val)
......
tests/gpuarray/test_nnet.py
浏览文件 @ b4dc02d6
import numpy as np
import theano
import theano.tensor as T
import theano.tensor as tt
import tests.unittest_tools as utt
from theano.tensor.nnet import crossentropy_softmax_1hot_with_bias_dx
from theano.gpuarray.nnet import (
GpuCrossentropySoftmaxArgmax1HotWithBias,
GpuCrossentropySoftmax1HotWithBiasDx,
GpuSoftmaxWithBias,
GpuSoftmax,
)
from tests.gpuarray.config import mode_with_gpu, mode_without_gpu
mode_wo_cudnn = mode_with_gpu.excluding("cudnn")
......@@ -29,16 +29,16 @@ def test_GpuCrossentropySoftmaxArgmax1HotWithBias():
n_in = 4098
n_out = 4099
y = T.lvector("y")
y = tt.lvector("y")
b = T.fvector("b")
b = tt.fvector("b")
# we precompute the dot with big shape before to allow the test of
# GpuCrossentropySoftmax1HotWithBiasDx to don't fail with the error
# (the launch timed out and was terminated) on GPU card not
# powerful enough. We need the big shape to check for corner
# case.
dot_result = T.fmatrix("dot_result")
dot_result = tt.fmatrix("dot_result")
# Seed numpy.random with config.unittests.rseed
utt.seed_rng()
......@@ -50,10 +50,10 @@ def test_GpuCrossentropySoftmaxArgmax1HotWithBias():
dot_value = np.asarray(np.dot(xx, W_values), dtype="float32")
del W_values
p_y_given_x = T.nnet.softmax(dot_result + b)
y_pred = T.argmax(p_y_given_x, axis=-1)
loss = -T.mean(T.log(p_y_given_x)[T.arange(y.shape[0]), y])
dW = T.grad(loss, dot_result)
p_y_given_x = tt.nnet.softmax(dot_result + b)
y_pred = tt.argmax(p_y_given_x, axis=-1)
loss = -tt.mean(tt.log(p_y_given_x)[tt.arange(y.shape[0]), y])
dW = tt.grad(loss, dot_result)
classify = theano.function(
inputs=[y, b, dot_result], outputs=[loss, y_pred, dW], mode=mode_without_gpu
)
......@@ -63,7 +63,7 @@ def test_GpuCrossentropySoftmaxArgmax1HotWithBias():
assert any(
[
isinstance(node.op, T.nnet.CrossentropySoftmaxArgmax1HotWithBias)
isinstance(node.op, tt.nnet.CrossentropySoftmaxArgmax1HotWithBias)
for node in classify.maker.fgraph.toposort()
]
)
......@@ -100,11 +100,9 @@ def test_GpuCrossentropySoftmax1HotWithBiasDx():
dnll_value = np.asarray(np.random.rand(batch_size), dtype="float32")
y_idx_value = np.random.randint(low=0, high=5, size=batch_size)
softmax_output = T.fmatrix()
softmax_output = tt.fmatrix()
softmax_output /= softmax_output.sum(axis=1).reshape(softmax_output.shape[1], 1)
op = theano.tensor.nnet.crossentropy_softmax_1hot_with_bias_dx(
dnll_value, softmax_output, y_idx_value
)
op = crossentropy_softmax_1hot_with_bias_dx(dnll_value, softmax_output, y_idx_value)
cpu_f = theano.function([softmax_output], op, mode=mode_without_gpu)
gpu_f = theano.function([softmax_output], op, mode=mode_with_gpu)
......@@ -113,7 +111,7 @@ def test_GpuCrossentropySoftmax1HotWithBiasDx():
assert any(
[
isinstance(node.op, T.nnet.CrossentropySoftmax1HotWithBiasDx)
isinstance(node.op, tt.nnet.CrossentropySoftmax1HotWithBiasDx)
for node in cpu_f.maker.fgraph.toposort()
]
)
......@@ -156,14 +154,14 @@ def softmax_with_bias_unittest_template(dtypeInput, dtypeBias):
# TODO: check that we loop when there are too many threads. (THIS IS
# NOT IMPLEMENTED)
x = T.matrix("x", dtype=dtypeInput)
b = T.vector("b", dtype=dtypeBias)
x = tt.matrix("x", dtype=dtypeInput)
b = tt.vector("b", dtype=dtypeBias)
z = T.nnet.softmax_with_bias(x, b)
z = tt.nnet.softmax_with_bias(x, b)
f = theano.function([x, b], z, mode=mode_without_gpu)
f_gpu = theano.function([x, b], z, mode=mode_with_gpu)
assert f.maker.fgraph.toposort()[-1].op == T.nnet.softmax_with_bias
assert f.maker.fgraph.toposort()[-1].op == tt.nnet.softmax_with_bias
assert isinstance(f_gpu.maker.fgraph.toposort()[-2].op, GpuSoftmaxWithBias)
def cmp(n, m):
......@@ -209,12 +207,12 @@ def softmax_unittest_template(dtypeInput):
# We check that we loop when their is too much block
# We use slower code when there isn't enough shared memory
x = T.matrix("x", dtype=dtypeInput)
x = tt.matrix("x", dtype=dtypeInput)
z = T.nnet.softmax(x)
z = tt.nnet.softmax(x)
f = theano.function([x], z, mode=mode_without_gpu)
f_gpu = theano.function([x], z, mode=mode_wo_cudnn)
assert f.maker.fgraph.toposort()[-1].op == T.nnet.softmax_op
assert f.maker.fgraph.toposort()[-1].op == tt.nnet.softmax_op
assert isinstance(f_gpu.maker.fgraph.toposort()[-2].op, GpuSoftmax)
def cmp(n, m):
......@@ -256,7 +254,7 @@ class TestSoftMax:
f = theano.function([x], f_z_out, mode=mode_without_gpu)
f_gpu = theano.function([x_gpu], f_gpu_z_out, mode=self.mode)
self._check_types(f, f_gpu, T.nnet.Softmax, self.gpu_op)
self._check_types(f, f_gpu, tt.nnet.Softmax, self.gpu_op)
# we need to test n>32*1024 to check that we make the block loop.
cmp(1, 5, f, f_gpu)
......@@ -303,16 +301,16 @@ class TestSoftMax:
)
def test_softmax(self):
x = T.fmatrix("x")
z = T.nnet.softmax_op
x = tt.fmatrix("x")
z = tt.nnet.softmax_op
f, f_gpu = self._test_softmax(x, x, z, z, self._cmp)
self._cmp(2 << 15, 5, f, f_gpu)
def test_softmax_shape_0(self):
x = T.fmatrix("x")
z = T.nnet.softmax_op
x = tt.fmatrix("x")
z = tt.nnet.softmax_op
f, f_gpu = self._test_softmax(x, x, z, z, self._cmp)
# Theano can handle that case, but cudnn can't
......
tests/gpuarray/test_reduction.py
浏览文件 @ b4dc02d6
......@@ -5,7 +5,7 @@ import pytest
import numpy as np
import theano
import theano.tensor as T
import theano.tensor as tt
from theano.gpuarray import GpuArrayType
from theano.gpuarray.reduction import GpuMaxAndArgmax
......@@ -96,7 +96,7 @@ class BaseTest:
def get_host_tensor(self):
broadcastable = (False,) * self.tensor_size
return T.tensor(self.dtype, broadcastable)
return tt.tensor(self.dtype, broadcastable)
def get_gpu_tensor(self):
broadcastable = (False,) * self.tensor_size
......@@ -116,7 +116,7 @@ class BaseTest:
M = self.get_host_tensor()
f = theano.function(
[M],
[T.max(M, axis=axis), T.argmax(M, axis=axis)],
[tt.max(M, axis=axis), tt.argmax(M, axis=axis)],
name="shape:" + str(test_tensor.shape) + "/axis:" + str(axis) + "/HOST",
mode=mode_without_gpu,
)
......@@ -131,7 +131,7 @@ class BaseTest:
M = self.get_gpu_tensor()
f = theano.function(
[M],
[T.max(M, axis=axis), T.argmax(M, axis=axis)],
[tt.max(M, axis=axis), tt.argmax(M, axis=axis)],
name="shape:" + str(test_gpu_tensor.shape) + "/axis:" + str(axis) + "/GPU",
mode=mode_with_gpu,
)
......
Markdown 格式
0% 或
您添加了 0 人 到此讨论。请谨慎行事。
请先完成此评论的编辑！
请 注册 或者 后发表评论