Skip to content

abandoned create_plugin() function #2146

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Aug 14, 2023
Merged

abandoned create_plugin() function #2146

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
0976470
abandoned create_plugin() function
zewenli98 Jul 27, 2023
8923872
Merge branch 'pytorch:main' into main
zewenli98 Aug 3, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
282 changes: 6 additions & 276 deletions core/conversion/converters/impl/interpolate.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -336,28 +336,7 @@ auto interpolate_registrations TORCHTRT_UNUSED =
float scale = args[3].IValue()->toDouble();
std::vector<float> padded_scales(in_shape.size(), 1);
padded_scales[padded_scales.size() - 1] = scale;
#if NV_TENSORRT_MAJOR < 7 || (NV_TENSORRT_MAJOR == 7 && NV_TENSORRT_MINOR < 1) // IF TRT VERSION <= 7.0
if (!align_corners) {
TORCHTRT_THROW_ERROR(
"Unable to convert node: "
<< util::node_info(n) << "\nupsample_linear1d only supports align_corner with TensorRT <= 7.0.");
} else {
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, true);
}
#else
TORCHTRT_CHECK(
!(align_corners && ctx->input_is_dynamic),
"Torch-TensorRT currently does not support the compilation of dynamic engines from code using using PyTorch [bi/tri]linear interpolation via scale factor and align_corners=True");
if (align_corners) {
// Align corners and scale factor behave slightly different together in TRT and PyTorch so run the
// layer in ATen to maintain consistancy between Torch-TensorRT and PyTorch
// https://pytorch.org/docs/stable/nn.functional.html#torch.nn.functional.interpolate
create_plugin(
ctx, n, in, "linear1d", in_shape, {}, {}, {scale}, std::string("linear"), align_corners, true);
} else {
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, align_corners);
}
#endif
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, align_corners);
} else {
// Case 2: user uses output size
auto out_size = util::toVec(util::toDims(args[1].unwrapToIntList()));
Expand All @@ -366,18 +345,7 @@ auto interpolate_registrations TORCHTRT_UNUSED =

auto out_shape = in_shape;
std::copy(out_size.begin(), out_size.end(), out_shape.begin() + (in_shape.size() - out_size.size()));
#if NV_TENSORRT_MAJOR < 7 || (NV_TENSORRT_MAJOR == 7 && NV_TENSORRT_MINOR < 1) // IF TRT VERSION <= 7.0
if (!align_corners) {
// align_corners not supported in TensorRT, create plugin and
// run layer through PyTorch
create_plugin(
ctx, n, in, "linear1d", in_shape, out_shape, out_size, {}, std::string("linear"), align_corners);
} else {
resize_layer_size(ctx, n, in, out_shape, {}, nvinfer1::ResizeMode::kLINEAR, true);
}
#else
resize_layer_size(ctx, n, in, out_shape, {}, nvinfer1::ResizeMode::kLINEAR, align_corners);
#endif
}

return true;
Expand All @@ -400,28 +368,7 @@ auto interpolate_registrations TORCHTRT_UNUSED =
float scale = scale_factors[0];
std::vector<float> padded_scales(in_shape.size(), 1);
padded_scales[padded_scales.size() - 1] = scale;
#if NV_TENSORRT_MAJOR < 7 || (NV_TENSORRT_MAJOR == 7 && NV_TENSORRT_MINOR < 1) // IF TRT VERSION <= 7.0
if (!align_corners) {
TORCHTRT_THROW_ERROR(
"Unable to convert node: "
<< util::node_info(n) << "\nupsample_linear1d only supports align_corner with TensorRT <= 7.0.");
} else {
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, true);
}
#else
TORCHTRT_CHECK(
!(align_corners && ctx->input_is_dynamic),
"Torch-TensorRT currently does not support the compilation of dynamc engines from code using PyTorch [bi/tri]linear interpolation via scale factor and align_corners=True");
if (align_corners) {
// Align corners and scale factor behave slightly different together in TRT and PyTorch so run the
// layer in ATen to maintain consistancy between Torch-TensorRT and PyTorch
// https://pytorch.org/docs/stable/nn.functional.html#torch.nn.functional.interpolate
create_plugin(
ctx, n, in, "linear1d", in_shape, {}, {}, {scale}, std::string("linear"), align_corners, true);
} else {
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, align_corners);
}
#endif
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, align_corners);
} else {
// Case 2: user uses output size
auto out_size = util::toVec(util::toDims(args[1].unwrapToIntList()));
Expand All @@ -430,18 +377,7 @@ auto interpolate_registrations TORCHTRT_UNUSED =

auto out_shape = in_shape;
std::copy(out_size.begin(), out_size.end(), out_shape.begin() + (in_shape.size() - out_size.size()));
#if NV_TENSORRT_MAJOR < 7 || (NV_TENSORRT_MAJOR == 7 && NV_TENSORRT_MINOR < 1) // IF TRT VERSION <= 7.0
if (!align_corners) {
// align_corners not supported in TensorRT, create plugin and
// run layer through PyTorch
create_plugin(
ctx, n, in, "linear1d", in_shape, out_shape, out_size, {}, std::string("linear"), align_corners);
} else {
resize_layer_size(ctx, n, in, out_shape, {}, nvinfer1::ResizeMode::kLINEAR, true);
}
#else
resize_layer_size(ctx, n, in, out_shape, {}, nvinfer1::ResizeMode::kLINEAR, align_corners);
#endif
}

return true;
Expand All @@ -464,38 +400,7 @@ auto interpolate_registrations TORCHTRT_UNUSED =
std::vector<float> padded_scales(in_shape.size(), 1);
padded_scales[padded_scales.size() - 2] = scale_h;
padded_scales[padded_scales.size() - 1] = scale_w;
#if NV_TENSORRT_MAJOR < 7 || (NV_TENSORRT_MAJOR == 7 && NV_TENSORRT_MINOR < 1) // IF TRT VERSION <= 7.0
if (!align_corners) {
TORCHTRT_THROW_ERROR(
"Unable to convert node: "
<< util::node_info(n) << "\nupsample_linear2d only supports align_corner with TensorRT <= 7.0.");
} else {
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, true);
}
#else
TORCHTRT_CHECK(
!(align_corners && ctx->input_is_dynamic),
"Torch-TensorRT currently does not support the compilation of dynamc engines from code using PyTorch [bi/tri]linear interpolation via scale factor and align_corners=True");
if (align_corners) {
// Align corners and scale factor behave slightly different together in TRT and PyTorch so run the
// layer in ATen to maintain consistancy between Torch-TensorRT and PyTorch
// https://pytorch.org/docs/stable/nn.functional.html#torch.nn.functional.interpolate
create_plugin(
ctx,
n,
in,
"bilinear2d",
in_shape,
{},
{},
{scale_h, scale_w},
std::string("bilinear"),
align_corners,
true);
} else {
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, align_corners);
}
#endif
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, align_corners);
} else {
// Case 2: user uses output size
auto out_size = util::toVec(util::toDims(args[1].unwrapToIntList()));
Expand All @@ -505,28 +410,7 @@ auto interpolate_registrations TORCHTRT_UNUSED =

auto out_shape = in_shape;
std::copy(out_size.begin(), out_size.end(), out_shape.begin() + (in_shape.size() - out_size.size()));

#if NV_TENSORRT_MAJOR < 7 || (NV_TENSORRT_MAJOR == 7 && NV_TENSORRT_MINOR < 1) // IF TRT VERSION <= 7.0
if (!align_corners) {
// align_corners not supported in TensorRT, create plugin and
// run layer through PyTorch
create_plugin(
ctx,
n,
in,
"bilinear2d",
in_shape,
out_shape,
out_size,
{},
std::string("bilinear"),
align_corners);
} else {
resize_layer_size(ctx, n, in, out_shape, {}, nvinfer1::ResizeMode::kLINEAR, true);
}
#else
resize_layer_size(ctx, n, in, out_shape, {}, nvinfer1::ResizeMode::kLINEAR, align_corners);
#endif
}

return true;
Expand All @@ -551,38 +435,7 @@ auto interpolate_registrations TORCHTRT_UNUSED =
std::vector<float> padded_scales(in_shape.size(), 1);
padded_scales[padded_scales.size() - 2] = scale_h;
padded_scales[padded_scales.size() - 1] = scale_w;
#if NV_TENSORRT_MAJOR < 7 || (NV_TENSORRT_MAJOR == 7 && NV_TENSORRT_MINOR < 1) // IF TRT VERSION <= 7.0
if (!align_corners) {
TORCHTRT_THROW_ERROR(
"Unable to convert node: "
<< util::node_info(n) << "\nupsample_linear2d only supports align_corner with TensorRT <= 7.0.");
} else {
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, true);
}
#else
TORCHTRT_CHECK(
!(align_corners && ctx->input_is_dynamic),
"Torch-TensorRT currently does not support the compilation of dynamc engines from code using PyTorch [bi/tri]linear interpolation via scale factor and align_corners=True");
if (align_corners) {
// Align corners and scale factor behave slightly different together in TRT and PyTorch so run the
// layer in ATen to maintain consistancy between Torch-TensorRT and PyTorch
// https://pytorch.org/docs/stable/nn.functional.html#torch.nn.functional.interpolate
create_plugin(
ctx,
n,
in,
"bilinear2d",
in_shape,
{},
{},
scale_factors.vec(),
std::string("bilinear"),
align_corners,
true);
} else {
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, align_corners);
}
#endif
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, align_corners);
} else {
// Case 2: user uses output size
auto out_size = util::toVec(util::toDims(args[1].unwrapToIntList()));
Expand All @@ -592,28 +445,7 @@ auto interpolate_registrations TORCHTRT_UNUSED =

auto out_shape = in_shape;
std::copy(out_size.begin(), out_size.end(), out_shape.begin() + (in_shape.size() - out_size.size()));

#if NV_TENSORRT_MAJOR < 7 || (NV_TENSORRT_MAJOR == 7 && NV_TENSORRT_MINOR < 1) // IF TRT VERSION <= 7.0
if (!align_corners) {
// align_corners not supported in TensorRT, create plugin and
// run layer through PyTorch
create_plugin(
ctx,
n,
in,
"bilinear2d",
in_shape,
out_shape,
out_size,
{},
std::string("bilinear"),
align_corners);
} else {
resize_layer_size(ctx, n, in, out_shape, {}, nvinfer1::ResizeMode::kLINEAR, true);
}
#else
resize_layer_size(ctx, n, in, out_shape, {}, nvinfer1::ResizeMode::kLINEAR, align_corners);
#endif
}

return true;
Expand All @@ -638,38 +470,7 @@ auto interpolate_registrations TORCHTRT_UNUSED =
padded_scales[padded_scales.size() - 3] = scale_d;
padded_scales[padded_scales.size() - 2] = scale_h;
padded_scales[padded_scales.size() - 1] = scale_w;
#if NV_TENSORRT_MAJOR < 7 || (NV_TENSORRT_MAJOR == 7 && NV_TENSORRT_MINOR < 1) // IF TRT VERSION <= 7.0
if (!align_corners) {
TORCHTRT_THROW_ERROR(
"Unable to convert node: "
<< util::node_info(n) << "\nupsample_linear3d only supports align_corner with TensorRT <= 7.0.");
} else {
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, true);
}
#else
TORCHTRT_CHECK(
!(align_corners && ctx->input_is_dynamic),
"Torch-TensorRT currently does not support the compilation of dynamc engines from code using PyTorch [bi/tri]linear interpolation via scale factor and align_corners=True");
if (align_corners) {
// Align corners and scale factor behave slightly different together in TRT and PyTorch so run the
// layer in ATen to maintain consistancy between Torch-TensorRT and PyTorch
// https://pytorch.org/docs/stable/nn.functional.html#torch.nn.functional.interpolate
create_plugin(
ctx,
n,
in,
"trilinear3d",
in_shape,
{},
{},
{scale_d, scale_h, scale_w},
std::string("trilinear"),
align_corners,
true);
} else {
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, align_corners);
}
#endif
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, align_corners);
} else {
// Case 2: user uses output size
auto out_size = util::toVec(util::toDims(args[1].unwrapToIntList()));
Expand All @@ -679,27 +480,7 @@ auto interpolate_registrations TORCHTRT_UNUSED =

auto out_shape = in_shape;
std::copy(out_size.begin(), out_size.end(), out_shape.begin() + (in_shape.size() - out_size.size()));
#if NV_TENSORRT_MAJOR < 7 || (NV_TENSORRT_MAJOR == 7 && NV_TENSORRT_MINOR < 1) // IF TRT VERSION <= 7.0
if (!align_corners) {
// align_corners not supported in TensorRT, create plugin and
// run layer through PyTorch
create_plugin(
ctx,
n,
in,
"trilinear3d",
in_shape,
out_shape,
out_size,
{},
std::string("trilinear"),
align_corners);
} else {
resize_layer_size(ctx, n, in, out_shape, {}, nvinfer1::ResizeMode::kLINEAR, true);
}
#else
resize_layer_size(ctx, n, in, out_shape, {}, nvinfer1::ResizeMode::kLINEAR, align_corners);
#endif
}

return true;
Expand All @@ -726,38 +507,7 @@ auto interpolate_registrations TORCHTRT_UNUSED =
padded_scales[padded_scales.size() - 3] = scale_d;
padded_scales[padded_scales.size() - 2] = scale_h;
padded_scales[padded_scales.size() - 1] = scale_w;
#if NV_TENSORRT_MAJOR < 7 || (NV_TENSORRT_MAJOR == 7 && NV_TENSORRT_MINOR < 1) // IF TRT VERSION <= 7.0
if (!align_corners) {
TORCHTRT_THROW_ERROR(
"Unable to convert node: "
<< util::node_info(n) << "\nupsample_linear3d only supports align_corner with TensorRT <= 7.0.");
} else {
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, true);
}
#else
TORCHTRT_CHECK(
!(align_corners && ctx->input_is_dynamic),
"Torch-TensorRT currently does not support the compilation of dynamc engines from code using PyTorch [bi/tri]linear interpolation via scale factor and align_corners=True");
if (align_corners) {
// Align corners and scale factor behave slightly different together in TRT and PyTorch so run the
// layer in ATen to maintain consistancy between Torch-TensorRT and PyTorch
// https://pytorch.org/docs/stable/nn.functional.html#torch.nn.functional.interpolate
create_plugin(
ctx,
n,
in,
"trilinear3d",
in_shape,
{},
{},
scale_factors.vec(),
std::string("trilinear"),
align_corners,
true);
} else {
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, align_corners);
}
#endif
resize_layer_size(ctx, n, in, {}, padded_scales, nvinfer1::ResizeMode::kLINEAR, align_corners);
} else {
// Case 2: user uses output size
auto out_size = util::toVec(util::toDims(args[1].unwrapToIntList()));
Expand All @@ -767,27 +517,7 @@ auto interpolate_registrations TORCHTRT_UNUSED =

auto out_shape = in_shape;
std::copy(out_size.begin(), out_size.end(), out_shape.begin() + (in_shape.size() - out_size.size()));
#if NV_TENSORRT_MAJOR < 7 || (NV_TENSORRT_MAJOR == 7 && NV_TENSORRT_MINOR < 1) // IF TRT VERSION <= 7.0
if (!align_corners) {
// align_corners not supported in TensorRT, create plugin and
// run layer through PyTorch
create_plugin(
ctx,
n,
in,
"trilinear3d",
in_shape,
out_shape,
out_size,
{},
std::string("trilinear"),
align_corners);
} else {
resize_layer_size(ctx, n, in, out_shape, {}, nvinfer1::ResizeMode::kLINEAR, true);
}
#else
resize_layer_size(ctx, n, in, out_shape, {}, nvinfer1::ResizeMode::kLINEAR, align_corners);
#endif
}

return true;
Expand Down