Skip to content

fix embedding_4bit resize #3118

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

Closed
wants to merge 1 commit into from
Closed

fix embedding_4bit resize #3118

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
2 changes: 1 addition & 1 deletion kernels/quantized/cpu/op_embedding4b.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -195,7 +195,7 @@ void resize_out_tensor(
for (size_t i = 0; i < indices.dim(); i++) {
expected_output_size[i] = indices.size(i);
}
const size_t embedding_dim = weight.size(1);
const size_t embedding_dim = weight.size(1) * 2;
expected_output_size[out.dim() - 1] = embedding_dim;

exec_aten::ArrayRef<exec_aten::SizesType> output_size{
Expand Down
15 changes: 15 additions & 0 deletions kernels/quantized/test/op_embedding4b_test.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,6 +19,7 @@
using namespace ::testing;
using exec_aten::ArrayRef;
using exec_aten::optional;
using exec_aten::RuntimeContext;
using exec_aten::ScalarType;
using exec_aten::Tensor;
using torch::executor::native::quantized_embedding_4bit_out;
Expand Down Expand Up @@ -60,6 +61,20 @@ TEST(OpQuantizedEmbedding4bTest, TestGroupWiseQuantizedEmbedding) {

EXPECT_TENSOR_EQ(out, expected);

out = tf.zeros({3, 4});
auto context = RuntimeContext();
torch::executor::native::quantized_embedding_4bit_out(
context,
qweight,
weight_scales,
weight_zero_points,
quant_min,
quant_max,
indices,
out);

EXPECT_TENSOR_EQ(out, expected);

// Groupwise quantization. groupsize = 2
weight_scales = tf.make({3, 2}, {0.5, 1.0, 1.5, 2.0, 2.5, 3.0});
weight_zero_points = tf.make({3, 2}, {1, -5, 0, 2, -3, -1});
Expand Down