saymrwulf/pytorch
1
0
Fork 0
mirror of https://github.com/saymrwulf/pytorch.git synced 2026-05-15 21:00:47 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Revert D20464855: [pytorch][PR] Add the fusion of quantized batchnorm and relu

Browse source 
Test Plan: revert-hammer

Differential Revision:
D20464855

Original commit changeset: 57090d427053

fbshipit-source-id: e7c50b5e7cd27a479539d7ee17580118377971c5
This commit is contained in:
Mike Ruberry 2020-03-19 18:27:21 -07:00 committed by Facebook GitHub Bot
parent a4afac6076
commit e7fc55ef7b
4 changed files with 36 additions and 92 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

1
aten/src/ATen/native/quantized/cpu/kernels/QuantizedOpKernels.cpp
View file

@ -1477,7 +1477,6 @@ REGISTER_DISPATCH(qcat_nhwc_stub, &qcat_nhwc_kernel<false>);
REGISTER_DISPATCH(qcat_relu_nhwc_stub, &qcat_nhwc_kernel<true>);
REGISTER_DISPATCH(qtopk_stub, &qtopk_kernel);
REGISTER_DISPATCH(qbatch_norm_stub, &q_batch_norm_kernel<false>);
REGISTER_DISPATCH(qbatch_norm_relu_stub, &q_batch_norm_kernel<true>);
REGISTER_DISPATCH(fake_quant_tensor_stub, &fake_quantize_tensor_kernel);
REGISTER_DISPATCH(fake_quant_grad_tensor_stub, &fake_quantize_grad_tensor_kernel);
REGISTER_DISPATCH(fake_quant_per_channel_stub, &fake_quant_per_channel_cpu);

75
aten/src/ATen/native/quantized/cpu/qbatch_norm.cpp
View file

@ -11,7 +11,6 @@ namespace at {
namespace native {
DEFINE_DISPATCH(qbatch_norm_stub);
DEFINE_DISPATCH(qbatch_norm_relu_stub);
namespace {
void compute_fused_params(
@ -98,31 +97,18 @@ Tensor q_batch_norm_impl(
output_scale,
alpha_data,
beta_data);
if (ReluFused) {
qbatch_norm_relu_stub(
qx.device().type(),
N,
C,
H * W,
qx.q_zero_point(),
output_zero_point,
qx_nhwc,
alpha,
beta,
qy);
} else {
qbatch_norm_stub(
qx.device().type(),
N,
C,
H * W,
qx.q_zero_point(),
output_zero_point,
qx_nhwc,
alpha,
beta,
qy);
}
qbatch_norm_stub(
qx.device().type(),
N,
C,
H * W,
qx.q_zero_point(),
output_zero_point,
qx_nhwc,
alpha,
beta,
qy);
return qy;
}
@ -187,31 +173,18 @@ Tensor q_batch_norm3d_impl(
alpha_data,
beta_data);
if (ReluFused) {
qbatch_norm_relu_stub(
qx.device().type(),
N,
C,
D * H * W,
qx.q_zero_point(),
output_zero_point,
qx_nhwc,
alpha,
beta,
qy);
} else {
qbatch_norm_stub(
qx.device().type(),
N,
C,
D * H * W,
qx.q_zero_point(),
output_zero_point,
qx_nhwc,
alpha,
beta,
qy);
}
qbatch_norm_stub(
qx.device().type(),
N,
C,
D * H * W,
qx.q_zero_point(),
output_zero_point,
qx_nhwc,
alpha,
beta,
qy);
return qy;
}

1
aten/src/ATen/native/quantized/cpu/quantized_ops.h
View file

@ -137,7 +137,6 @@ DECLARE_DISPATCH(qcat_nhwc_fn, qcat_nhwc_stub);
DECLARE_DISPATCH(qcat_nhwc_fn, qcat_relu_nhwc_stub);
DECLARE_DISPATCH(qtopk_fn, qtopk_stub);
DECLARE_DISPATCH(qbatch_norm_fn, qbatch_norm_stub);
DECLARE_DISPATCH(qbatch_norm_fn, qbatch_norm_relu_stub);
} // namespace native
} // namespace at

51
test/test_quantized.py
View file

@ -1401,10 +1401,13 @@ class TestQuantizedOps(TestCase):
min_side=1, max_side=32),
qparams=hu.qparams()),
Y_scale=st.floats(0.2, 2.6),
Y_zero_point=st.integers(0, 5))
def test_batch_norm(self, X, Y_scale, Y_zero_point):
Y_zero_point=st.integers(0, 5),
qengine=st.sampled_from(("qnnpack", "fbgemm")))
def test_batch_norm(self, X, Y_scale, Y_zero_point, qengine):
if qengine not in torch.backends.quantized.supported_engines:
return
with override_quantized_engine("fbgemm"):
with override_quantized_engine(qengine):
X, (scale_x, zero_point_x, dtype_x) = X
X = torch.from_numpy(X)
@ -1423,47 +1426,17 @@ class TestQuantizedOps(TestCase):
quantize_ref = torch.quantize_per_tensor(float_ref, Y_scale, Y_zero_point, dtype_x)
self.assertEqual(qy.int_repr().numpy(), quantize_ref.int_repr().numpy())
@given(X=hu.tensor(shapes=hu.array_shapes(min_dims=4, max_dims=5,
min_side=1, max_side=32),
qparams=hu.qparams()),
Y_scale=st.floats(0.2, 2.6),
Y_zero_point=st.integers(0, 5))
def test_batch_norm_relu(self, X, Y_scale, Y_zero_point):
with override_quantized_engine("fbgemm"):
X, (scale_x, zero_point_x, dtype_x) = X
X = torch.from_numpy(X)
c = X.shape[1]
mean = torch.rand(c).float()
var = torch.rand(c).float()
weight = torch.rand(c).float()
bias = torch.rand(c).float()
eps = 0.001
qx = torch.quantize_per_tensor(X, scale_x, zero_point_x, dtype_x)
if len(X.shape) == 4:
qy = torch.ops.quantized.batch_norm2d_relu(qx, weight, bias, mean, var, eps, Y_scale, Y_zero_point)
else:
qy = torch.ops.quantized.batch_norm3d_relu(qx, weight, bias, mean, var, eps, Y_scale, Y_zero_point)
float_ref = F.batch_norm(qx.dequantize(), weight=weight, bias=bias,
running_mean=mean, running_var=var, training=False, momentum=0, eps=eps).numpy()
float_ref_relu = float_ref.copy()
float_ref_relu[float_ref < 0] = 0
quantize_ref = torch.quantize_per_tensor(torch.from_numpy(float_ref_relu), Y_scale, Y_zero_point, dtype_x)
self.assertEqual(qy.int_repr().numpy(), quantize_ref.int_repr().numpy())
@given(X=hu.tensor(shapes=hu.array_shapes(min_dims=5, max_dims=5,
min_side=1, max_side=32),
qparams=hu.qparams()),
Y_scale=st.floats(0.2, 2.6),
Y_zero_point=st.integers(0, 5))
def test_batch_norm3d(self, X, Y_scale, Y_zero_point):
Y_zero_point=st.integers(0, 5),
qengine=st.sampled_from(("qnnpack", "fbgemm")))
def test_batch_norm3d(self, X, Y_scale, Y_zero_point, qengine):
if qengine not in torch.backends.quantized.supported_engines:
return
with override_quantized_engine("fbgemm"):
with override_quantized_engine(qengine):
X, (scale_x, zero_point_x, dtype_x) = X
X = torch.from_numpy(X)