Skip to content
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

[CoreML] Adapt to MLMultiArray.dataPointer deprecation #17726

Merged
merged 15 commits into from
Nov 18, 2023
Merged

[CoreML] Adapt to MLMultiArray.dataPointer deprecation #17726

Changes from 9 commits
Commits
Show all changes
15 commits
Select commit Hold shift + click to select a range
724f9a1
fixed performance tuning link
NickLucche Sep 4, 2023
9e68d5e
fixed python api link
NickLucche Sep 4, 2023
4e498dc
support only contiguous model outputs
NickLucche Sep 27, 2023
3ddcad7
Merge branch 'microsoft:main' into main
NickLucche Sep 27, 2023
ce77170
replace MLMultiArray.dataPointer with getBytesWithHandler
NickLucche Sep 27, 2023
b2e8fdb
pr review changes
NickLucche Sep 30, 2023
37e39e5
checking getBytesWithHandler returned size against expected one
NickLucche Sep 30, 2023
6dda63f
120 chars line length limit
NickLucche Oct 2, 2023
7b1128e
fixed comparison types errors
NickLucche Oct 4, 2023
c664a00
checking getBytesWithHandler availability with @available
NickLucche Nov 7, 2023
bfe9062
refactored tensor buffer copy into own function
NickLucche Nov 10, 2023
f29e5ae
fixed variable qualifiers and naming
NickLucche Nov 13, 2023
a78ef97
better var qualifiers
NickLucche Nov 15, 2023
f88132d
changed tensor_info to pointer to avoid a shape vector copy
NickLucche Nov 16, 2023
dc82dfb
formatting
NickLucche Nov 17, 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
24 changes: 22 additions & 2 deletions onnxruntime/core/providers/coreml/model/model.mm
View file
Open in desktop
NickLucche marked this conversation as resolved.
Show resolved Hide resolved
Original file line number Diff line number Diff line change
Expand Up @@ -169,6 +169,14 @@ Status CreateInputFeatureProvider(const std::unordered_map<std::string, OnnxTens
conversion_buffers_out = std::move(conversion_buffers);
return Status::OK();
}

bool IsArrayContiguous(MLMultiArray* array) {
NickLucche marked this conversation as resolved.
Show resolved Hide resolved
int64_t batch_stride = [array.strides[0] longLongValue];
const auto* shape = array.shape;
int64_t batch_elems = 1;
for (unsigned long i = 1; i < shape.count; i++) batch_elems *= [shape[i] longLongValue];
return batch_stride == batch_elems;
}
} // namespace

NS_ASSUME_NONNULL_BEGIN
Expand Down Expand Up @@ -324,7 +332,14 @@ - (Status)predict:(const std::unordered_map<std::string, OnnxTensorData>&)inputs
") do not match");
}

const void* model_output_buffer = data.dataPointer;
ORT_RETURN_IF_NOT(IsArrayContiguous(data),
"Non-contiguous output MLMultiArray is not currently supported");
__block const void* model_output_buffer = nil;
__block unsigned long coreml_buffer_size = 0;
[data getBytesWithHandler:^(const void* bytes, NSInteger size) {
model_output_buffer = bytes;
coreml_buffer_size = size;
NickLucche marked this conversation as resolved.
Show resolved Hide resolved
}];

if (model_output_buffer == nullptr) {
return ORT_MAKE_STATUS(ONNXRUNTIME, FAIL, "model_output_buffer has no data for ", output_name);
Expand All @@ -334,11 +349,15 @@ - (Status)predict:(const std::unordered_map<std::string, OnnxTensorData>&)inputs
switch (onnx_data_type) {
case ONNX_NAMESPACE::TensorProto_DataType_FLOAT: {
const auto output_data_byte_size = num_elements * sizeof(float);
ORT_RETURN_IF_NOT(coreml_buffer_size == output_data_byte_size,
"CoreML output buffer size and expected output size differ");
memcpy(output_buffer, model_output_buffer, output_data_byte_size);
break;
}
case ONNX_NAMESPACE::TensorProto_DataType_INT32: {
const auto output_data_byte_size = num_elements * sizeof(int32_t);
ORT_RETURN_IF_NOT(coreml_buffer_size == output_data_byte_size,
"CoreML output buffer size and expected output size differ");
memcpy(output_buffer, model_output_buffer, output_data_byte_size);
break;
}
Expand All @@ -348,7 +367,8 @@ - (Status)predict:(const std::unordered_map<std::string, OnnxTensorData>&)inputs
case ONNX_NAMESPACE::TensorProto_DataType_INT64: {
ORT_RETURN_IF_NOT(data.dataType == MLMultiArrayDataTypeInt32,
"CoreML output data type is not MLMultiArrayDataTypeInt32");

ORT_RETURN_IF_NOT(coreml_buffer_size == num_elements * sizeof(int32_t),
"CoreML output buffer size and expected output size differ");
const auto model_output_span = gsl::span{static_cast<const int32_t*>(model_output_buffer), num_elements};
const auto output_span = gsl::span{static_cast<int64_t*>(output_buffer), num_elements};
std::transform(model_output_span.begin(), model_output_span.end(), output_span.begin(),
Expand Down