Skip to content

[feat]Add converter for aten::where #1421

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 1 commit into from
Oct 26, 2022
Merged

[feat]Add converter for aten::where #1421

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
17 changes: 17 additions & 0 deletions core/conversion/converters/impl/select.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -721,6 +721,23 @@ auto select_registrations TORCHTRT_UNUSED =

layer->setName(util::node_info(n).c_str());

auto out_tensor = ctx->AssociateValueAndTensor(n->outputs()[0], layer->getOutput(0));
LOG_DEBUG("Output shape: " << out_tensor->getDimensions());
return true;
}})
.pattern(
{"aten::where.self(Tensor condition, Tensor self, Tensor other) -> (Tensor)",
[](ConversionCtx* ctx, const torch::jit::Node* n, args& args) -> bool {
auto condition = args[0].ITensorOrFreeze(ctx);
auto x = args[1].ITensorOrFreeze(ctx);
auto y = args[2].ITensorOrFreeze(ctx);

auto layer = ctx->net->addSelect(*condition, *x, *y);

TORCHTRT_CHECK(layer, "Unable to create select layer for aten::where.self");

layer->setName(util::node_info(n).c_str());

auto out_tensor = ctx->AssociateValueAndTensor(n->outputs()[0], layer->getOutput(0));
LOG_DEBUG("Output shape: " << out_tensor->getDimensions());
return true;
Expand Down
30 changes: 30 additions & 0 deletions tests/core/conversion/converters/test_select.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1138,3 +1138,33 @@ TEST(Converters, ScatterSrcConvertsCorrectly) {
ASSERT_TRUE(torch_tensorrt::tests::util::almostEqual(jit_results[i], trt, 2e-6));
}
}

TEST(Converters, WhereConvertsCorrectly) {
const auto graph = R"IR(
graph(%condition : Tensor,
%x : Tensor,
%y : Tensor):
%out : Tensor = aten::where(%condition, %x, %y)
return (%out))IR";

auto g = std::make_shared<torch::jit::Graph>();

torch::jit::parseIR(graph, g.get());

auto condition = at::randint(0, 2, {5, 5}, {at::kCUDA}).to(torch::kBool);
auto x = at::randn({5, 5}, {at::kCUDA});
auto y = at::randn({5, 5}, {at::kCUDA});

auto jit_condition = at::clone(condition);
auto jit_x = at::clone(x);
auto jit_y = at::clone(y);
auto params = torch_tensorrt::core::ir::get_static_params(g->inputs(), {});
auto jit_results = torch_tensorrt::tests::util::RunGraph(g, params, {jit_condition, jit_x, jit_y});

auto trt_condition = at::clone(condition);
auto trt_x = at::clone(x);
auto trt_y = at::clone(y);
auto trt_results = torch_tensorrt::tests::util::RunGraphEngine(g, params, {trt_condition, trt_x, trt_y});

ASSERT_TRUE(torch_tensorrt::tests::util::almostEqual(jit_results[0], trt_results[0], 2e-6));
}