experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Add details about failed NNAPI operation when logging NNAPI errors. It also Converts the error code into the associated NNAPI constant name.

Browse Source 
PiperOrigin-RevId: 284149818
Change-Id: I3b6be19b93b40ada764c2edcd661c86619b9b830
This commit is contained in:
Stefano Galarraga 2019-12-06 02:47:23 -08:00 committed by TensorFlower Gardener
parent 9b3930304b
commit d0e55241d5
2 changed files with 86 additions and 40 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

122
tensorflow/lite/delegates/nnapi/nnapi_delegate.cc
View File

@ -57,14 +57,47 @@ limitations under the License.
namespace tflite {
namespace {
// TODO(b/80621585): Consider printing error string, but don't for now to
// minimize binary size.
#define RETURN_TFLITE_ERROR_IF_NN_ERROR(context, code, p_errno) \
// Returns the enum name corresponding to the given error code if the given
// value corresponds to an of the error codes in the enumeration above or
// an message with the unknown code.
// LINT.IfChange(NnApiErrorDescription)
std::string NnApiErrorDescription(int error_code) {
switch (error_code) {
case ANEURALNETWORKS_NO_ERROR:
return "ANEURALNETWORKS_NO_ERROR";
case ANEURALNETWORKS_OUT_OF_MEMORY:
return "ANEURALNETWORKS_OUT_OF_MEMORY";
case ANEURALNETWORKS_INCOMPLETE:
return "ANEURALNETWORKS_INCOMPLETE";
case ANEURALNETWORKS_UNEXPECTED_NULL:
return "ANEURALNETWORKS_UNEXPECTED_NULL";
case ANEURALNETWORKS_BAD_DATA:
return "ANEURALNETWORKS_BAD_DATA";
case ANEURALNETWORKS_OP_FAILED:
return "ANEURALNETWORKS_OP_FAILED";
case ANEURALNETWORKS_BAD_STATE:
return "ANEURALNETWORKS_BAD_STATE";
case ANEURALNETWORKS_UNMAPPABLE:
return "ANEURALNETWORKS_UNMAPPABLE";
case ANEURALNETWORKS_OUTPUT_INSUFFICIENT_SIZE:
return "ANEURALNETWORKS_OUTPUT_INSUFFICIENT_SIZE";
case ANEURALNETWORKS_UNAVAILABLE_DEVICE:
return "ANEURALNETWORKS_UNAVAILABLE_DEVICE";
default:
return "Unknown NNAPI error code: " + std::to_string(error_code);
}
}
// LINT.ThenChange()
#define RETURN_TFLITE_ERROR_IF_NN_ERROR(context, code, call_desc, p_errno) \
do { \
const auto _code = (code); \
const auto _call_desc = (call_desc); \
if (_code != ANEURALNETWORKS_NO_ERROR) { \
context->ReportError(context, "NN API returned error (%d, line %d).\n", \
_code, __LINE__); \
const auto error_desc = NnApiErrorDescription(_code); \
context->ReportError(context, \
"NN API returned error %s at line %d while %s.\n", \
error_desc.c_str(), __LINE__, _call_desc); \
*p_errno = _code; \
return kTfLiteError; \
} \
@ -611,7 +644,7 @@ class NNAPIOpBuilder {
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context_,
nnapi_->ANeuralNetworksModel_addOperand(nn_model_, &operand_type),
nnapi_errno_);
"adding operand", nnapi_errno_);
dequantized_ann_index = operand_mapping_->add_new_non_tensor_operand();
// Add Dequantize operation.
@ -623,7 +656,7 @@ class NNAPIOpBuilder {
nnapi_->ANeuralNetworksModel_addOperation(
nn_model_, ANEURALNETWORKS_DEQUANTIZE, 1, dequantize_input, 1,
dequantize_output),
nnapi_errno_);
"adding operation", nnapi_errno_);
dequantize_mapping_->Add(ann_index, dequantized_type,
dequantized_ann_index);
}
@ -645,7 +678,7 @@ class NNAPIOpBuilder {
augmented_inputs_.data(),
static_cast<uint32_t>(augmented_outputs_.size()),
augmented_outputs_.data()),
nnapi_errno_);
"adding operation", nnapi_errno_);
augmented_inputs_.clear();
augmented_outputs_.clear();
return kTfLiteOk;
@ -660,7 +693,7 @@ class NNAPIOpBuilder {
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context_,
nnapi_->ANeuralNetworksModel_addOperand(nn_model_, &operand_type),
nnapi_errno_);
"adding operand", nnapi_errno_);
int ann_tensor_index = operand_mapping_->lite_index_to_ann(tensor_index);
if (ann_tensor_index != -1) {
augmented_inputs_.push_back(ann_tensor_index);
@ -718,7 +751,7 @@ class NNAPIOpBuilder {
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context_,
nnapi_->ANeuralNetworksModel_addOperand(nn_model_, &operand_type),
nnapi_errno_);
"adding operand", nnapi_errno_);
augmented_inputs_.push_back(ann_tensor_index);
@ -727,7 +760,7 @@ class NNAPIOpBuilder {
nnapi_->ANeuralNetworksModel_setOperandValue(
nn_model_, ann_tensor_index, new_tensor->data.raw,
new_tensor->bytes),
nnapi_errno_);
"setting new operand value", nnapi_errno_);
return kTfLiteOk;
}
@ -774,13 +807,13 @@ class NNAPIOpBuilder {
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context_,
nnapi_->ANeuralNetworksModel_addOperand(nn_model_, &operand_type),
nnapi_errno_);
"adding operand", nnapi_errno_);
const int ann_index = operand_mapping_->add_new_non_tensor_operand();
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context_,
nnapi_->ANeuralNetworksModel_setOperandValue(nn_model_, ann_index,
&value, sizeof(T)),
nnapi_errno_);
"setting new operand value", nnapi_errno_);
augmented_inputs_.push_back(ann_index);
return kTfLiteOk;
}
@ -798,14 +831,14 @@ class NNAPIOpBuilder {
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context_,
nnapi_->ANeuralNetworksModel_addOperand(nn_model_, &operand_type),
nnapi_errno_);
"adding operand", nnapi_errno_);
const int ann_index = operand_mapping_->add_new_non_tensor_operand();
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context_,
nnapi_->ANeuralNetworksModel_setOperandValue(
nn_model_, ann_index, values, sizeof(T) * num_values),
nnapi_errno_);
"settings new operand value", nnapi_errno_);
augmented_inputs_.push_back(ann_index);
return kTfLiteOk;
}
@ -840,7 +873,7 @@ class NNAPIOpBuilder {
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context_,
nnapi_->ANeuralNetworksModel_addOperand(nn_model_, &operand_type),
nnapi_errno_);
"adding operand", nnapi_errno_);
const int ann_index = operand_mapping_->add_new_non_tensor_operand();
augmented_outputs_.push_back(ann_index);
if (ann_index_out) *ann_index_out = ann_index;
@ -960,14 +993,14 @@ class NNAPIOpBuilder {
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context_,
nnapi_->ANeuralNetworksModel_addOperand(nn_model_, &operand_type),
nnapi_errno_);
"adding operand", nnapi_errno_);
if (nn_type == ANEURALNETWORKS_TENSOR_QUANT8_SYMM_PER_CHANNEL) {
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context_,
nnapi_->ANeuralNetworksModel_setOperandSymmPerChannelQuantParams(
nn_model_, ann_tensor_index, &ann_perchannel_params),
nnapi_errno_);
"setting new operand per channel quantization params", nnapi_errno_);
}
if (tensor->allocation_type == kTfLiteMmapRo) {
if (IsQuantized(tensor_type) && need_int8_conversion) {
@ -1000,7 +1033,7 @@ class NNAPIOpBuilder {
nnapi_->ANeuralNetworksModel_setOperandValue(
nn_model_, ann_tensor_index, new_tensor->data.raw,
new_tensor->bytes),
nnapi_errno_);
"setting new operand value", nnapi_errno_);
#ifdef TFLITE_NNAPI_ALLOW_MMAP_SHARING
} else if (tensor->allocation &&
static_cast<const Allocation*>(tensor->allocation)->type() ==
@ -1025,14 +1058,14 @@ class NNAPIOpBuilder {
nnapi_->ANeuralNetworksModel_setOperandValueFromMemory(
nn_model_, ann_tensor_index, ann_memory_handle, offset,
tensor->bytes),
nnapi_errno_);
"setting new operand value from memory", nnapi_errno_);
#endif
} else {
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context_,
nnapi_->ANeuralNetworksModel_setOperandValue(
nn_model_, ann_tensor_index, tensor->data.raw, tensor->bytes),
nnapi_errno_);
"setting new operand value", nnapi_errno_);
}
}
@ -2910,8 +2943,9 @@ TfLiteStatus NNAPIDelegateKernel::Init(TfLiteContext* context,
if (!nn_model_) {
ANeuralNetworksModel* model = nullptr;
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context, nnapi_->ANeuralNetworksModel_create(&model), nnapi_errno);
RETURN_TFLITE_ERROR_IF_NN_ERROR(context,
nnapi_->ANeuralNetworksModel_create(&model),
"creating NNAPI model", nnapi_errno);
nn_model_.reset(model);
TF_LITE_ENSURE_STATUS(BuildGraph(context, params->input_tensors,
@ -2927,11 +2961,12 @@ TfLiteStatus NNAPIDelegateKernel::Init(TfLiteContext* context,
nnapi_->ANeuralNetworksCompilation_createForDevices(
nn_model_.get(), nnapi_devices_.data(), nnapi_devices_.size(),
&compilation),
nnapi_errno);
"creating NNAPI model for given devices", nnapi_errno);
} else {
RETURN_TFLITE_ERROR_IF_NN_ERROR(context,
nnapi_->ANeuralNetworksCompilation_create(
nn_model_.get(), &compilation),
"creating NNAPI compilation",
nnapi_errno);
}
@ -2945,7 +2980,9 @@ TfLiteStatus NNAPIDelegateKernel::Init(TfLiteContext* context,
nnapi_->ANeuralNetworksCompilation_free(compilation);
compilation = nullptr;
}
RETURN_TFLITE_ERROR_IF_NN_ERROR(context, preference_result, nnapi_errno);
RETURN_TFLITE_ERROR_IF_NN_ERROR(context, preference_result,
"setting compilation preferences",
nnapi_errno);
}
const char* cache_dir = delegate_options.cache_dir;
@ -2978,7 +3015,8 @@ TfLiteStatus NNAPIDelegateKernel::Init(TfLiteContext* context,
nnapi_->ANeuralNetworksCompilation_free(compilation);
compilation = nullptr;
}
RETURN_TFLITE_ERROR_IF_NN_ERROR(context, set_caching_result, nnapi_errno);
RETURN_TFLITE_ERROR_IF_NN_ERROR(context, set_caching_result,
"configuring NNAPI caching", nnapi_errno);
}
const int finish_result =
nnapi_->ANeuralNetworksCompilation_finish(compilation);
@ -2986,7 +3024,8 @@ TfLiteStatus NNAPIDelegateKernel::Init(TfLiteContext* context,
nnapi_->ANeuralNetworksCompilation_free(compilation);
compilation = nullptr;
}
RETURN_TFLITE_ERROR_IF_NN_ERROR(context, finish_result, nnapi_errno);
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context, finish_result, "completing NNAPI compilation", nnapi_errno);
nn_compilation_.reset(compilation);
}
return kTfLiteOk;
@ -3007,7 +3046,7 @@ TfLiteStatus NNAPIDelegateKernel::Invoke(TfLiteContext* context,
RETURN_TFLITE_ERROR_IF_NN_ERROR(context,
nnapi_->ANeuralNetworksExecution_create(
nn_compilation_.get(), &execution),
nnapi_errno);
"creating NNAPI execution", nnapi_errno);
std::unique_ptr<ANeuralNetworksExecution, NNFreeExecution>
execution_unique_ptr(execution);
@ -3030,6 +3069,7 @@ TfLiteStatus NNAPIDelegateKernel::Invoke(TfLiteContext* context,
execution, relative_input_index, nullptr,
tensor_memory_map_->at(tensor->buffer_handle).memory, 0,
tensor->bytes),
"associating NNAPI execution input with a memory object",
nnapi_errno);
relative_input_index++;
continue;
@ -3077,6 +3117,7 @@ TfLiteStatus NNAPIDelegateKernel::Invoke(TfLiteContext* context,
nnapi_->ANeuralNetworksExecution_setInputFromMemory(
execution, relative_input_index, nullptr,
nn_input_memory_->get_handle(), input_offset, tensor_size),
"associating NNAPI execution input with a memory object",
nnapi_errno);
} else {
// copy data to pre-allocated shared memory.
@ -3087,6 +3128,7 @@ TfLiteStatus NNAPIDelegateKernel::Invoke(TfLiteContext* context,
nnapi_->ANeuralNetworksExecution_setInputFromMemory(
execution, relative_input_index, nullptr,
nn_input_memory_->get_handle(), input_offset, tensor->bytes),
"associating NNAPI execution input with a memory object",
nnapi_errno);
tensor_size = tensor->bytes;
}
@ -3114,7 +3156,7 @@ TfLiteStatus NNAPIDelegateKernel::Invoke(TfLiteContext* context,
execution, relative_output_index, nullptr,
tensor_memory_map_->at(tensor->buffer_handle).memory, 0,
tensor->bytes),
nnapi_errno);
"associating NNAPI execution output to a memory object", nnapi_errno);
} else {
RETURN_TFLITE_ERROR_IF_NN_ERROR(
@ -3122,7 +3164,7 @@ TfLiteStatus NNAPIDelegateKernel::Invoke(TfLiteContext* context,
nnapi_->ANeuralNetworksExecution_setOutputFromMemory(
execution, relative_output_index, nullptr,
nn_output_memory_->get_handle(), output_offset, tensor->bytes),
nnapi_errno);
"associating NNAPI execution output to a memory object", nnapi_errno);
output_offset += tensor->bytes;
output_offset += getNumPaddingBytes(tensor->bytes);
}
@ -3142,7 +3184,7 @@ TfLiteStatus NNAPIDelegateKernel::Invoke(TfLiteContext* context,
nnapi_->ANeuralNetworksExecution_setOutput(
execution, relative_output_index, nullptr, tensor->data.raw,
tensor->bytes),
nnapi_errno);
"associating NNAPI execution output to a buffer", nnapi_errno);
relative_output_index++;
}
// Invoke ANN in blocking fashion.
@ -3151,15 +3193,17 @@ TfLiteStatus NNAPIDelegateKernel::Invoke(TfLiteContext* context,
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context,
nnapi_->ANeuralNetworksExecution_startCompute(execution, &event),
nnapi_errno);
"starting async computation", nnapi_errno);
const int wait_result = nnapi_->ANeuralNetworksEvent_wait(event);
nnapi_->ANeuralNetworksEvent_free(event);
RETURN_TFLITE_ERROR_IF_NN_ERROR(context, wait_result, nnapi_errno);
RETURN_TFLITE_ERROR_IF_NN_ERROR(context, wait_result,
"waiting for async computation completion",
nnapi_errno);
} else {
// Use synchronous execution for NNAPI 1.2+.
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context, nnapi_->ANeuralNetworksExecution_compute(execution),
nnapi_errno);
"running computation", nnapi_errno);
}
// copy results from shared memory to the destination.
@ -3567,21 +3611,19 @@ TfLiteStatus NNAPIDelegateKernel::BuildGraph(
nnapi_->ANeuralNetworksModel_identifyInputsAndOutputs(
nn_model_.get(), inputs.size(), inputs.data(), outputs.size(),
outputs.data()),
nnapi_errno);
"identifying model inputs and outputs", nnapi_errno);
// Set relaxed computation mode for fp32 if possible.
if (nnapi_->android_sdk_version >= kMinSdkVersionForNNAPI11) {
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context,
nnapi_->ANeuralNetworksModel_relaxComputationFloat32toFloat16(
nn_model_.get(), context->allow_fp32_relax_to_fp16),
nnapi_errno);
"set relaxed computation mode for fp32 if possible", nnapi_errno);
}
// Finalize the model
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context, nnapi_->ANeuralNetworksModel_finish(nn_model_.get()),
nnapi_errno);
"finalizing the model", nnapi_errno);
// Create shared memory pool for inputs and outputs.
nn_input_memory_.reset(
@ -3740,7 +3782,7 @@ TfLiteStatus StatefulNnApiDelegate::DoPrepare(TfLiteContext* context,
uint32_t device_count = 0;
RETURN_TFLITE_ERROR_IF_NN_ERROR(
context, nnapi->ANeuralNetworks_getDeviceCount(&device_count),
nnapi_errno);
"getting number of NNAPI devices", nnapi_errno);
if (device_count <= 1) {
return kTfLiteOk;
}

4
tensorflow/lite/nnapi/NeuralNetworksTypes.h
View File

@ -18,6 +18,8 @@ limitations under the License.
#include <stdint.h>
#include <stdio.h>
#include <string>
typedef struct AHardwareBuffer AHardwareBuffer;
// NN api types based on NNAPI header file
@ -159,6 +161,7 @@ enum {
/**
* Result codes.
*/
// LINT.IfChange
enum {
ANEURALNETWORKS_NO_ERROR = 0,
ANEURALNETWORKS_OUT_OF_MEMORY = 1,
@ -171,6 +174,7 @@ enum {
ANEURALNETWORKS_OUTPUT_INSUFFICIENT_SIZE = 8,
ANEURALNETWORKS_UNAVAILABLE_DEVICE = 9,
};
// LINT.ThenChange(//tensorflow/lite/delegates/nnapi/nnapi_delegate.cc:NnApiErrorDescription)
/**
* Implicit padding algorithms.