未验证 提交 840c521b 编写于 作者: J joanna.wozna.intel 提交者: GitHub

Fix problem with flags fp32 and int8 (#27954)

上级 5ccaaab8
......@@ -66,6 +66,7 @@ function(inference_analysis_api_int8_test_run TARGET_NAME test_binary model_dir
--infer_data=${data_path}
--warmup_batch_size=${WARMUP_BATCH_SIZE}
--batch_size=50
--enable_int8=true
--cpu_num_threads=${CPU_NUM_THREADS_ON_CI}
--iterations=2)
endfunction()
......@@ -81,6 +82,7 @@ function(inference_analysis_api_bfloat16_test_run TARGET_NAME test_binary model_
ARGS --infer_model=${model_dir}/model
--infer_data=${data_path}
--batch_size=50
--enable_bf16=true
--paddle_num_threads=${CPU_NUM_THREADS_ON_CI}
--iterations=2)
endfunction()
......@@ -92,6 +94,7 @@ function(inference_analysis_api_object_dection_int8_test_run TARGET_NAME test_bi
--infer_data=${data_path}
--warmup_batch_size=10
--batch_size=300
--enable_int8=true
--cpu_num_threads=${CPU_NUM_THREADS_ON_CI}
--iterations=1)
endfunction()
......@@ -115,6 +118,7 @@ function(inference_analysis_api_quant_test_run TARGET_NAME test_binary fp32_mode
--int8_model=${int8_model_dir}
--infer_data=${data_path}
--batch_size=50
--enable_int8=true
--cpu_num_threads=${CPU_NUM_THREADS_ON_CI}
--with_accuracy_layer=false
--iterations=2)
......
......@@ -50,8 +50,8 @@ DEFINE_bool(ernie_large, false, "Test ernie large");
DEFINE_bool(with_accuracy_layer, true,
"Calculate the accuracy while label is in the input");
DEFINE_bool(enable_fp32, true, "Enable FP32 type prediction");
DEFINE_bool(enable_bf16, true, "Enable BF16 type prediction");
DEFINE_bool(enable_int8, true, "Enable INT8 type prediction");
DEFINE_bool(enable_bf16, false, "Enable BF16 type prediction");
DEFINE_bool(enable_int8, false, "Enable INT8 type prediction");
DEFINE_int32(warmup_batch_size, 100, "batch size for quantization warmup");
// setting iterations to 0 means processing the whole dataset
DEFINE_int32(iterations, 0, "number of batches to process");
......@@ -639,8 +639,9 @@ void TestPrediction(const PaddlePredictor::Config *config,
}
}
void SummarizeAccuracy(float avg_acc_fp32, float avg_acc_int8,
int compared_idx) {
void SummarizeAccuracy(float avg_acc_ref, float avg_acc, int compared_idx) {
std::string data_type_name = "INT8";
if (FLAGS_enable_bf16) data_type_name = "BF16";
PADDLE_ENFORCE_LE(
compared_idx, 2,
platform::errors::InvalidArgument(
......@@ -659,12 +660,12 @@ void SummarizeAccuracy(float avg_acc_fp32, float avg_acc_int8,
LOG(INFO) << "--- Accuracy summary --- ";
LOG(INFO) << "Accepted " << prefix
<< "drop threshold: " << FLAGS_quantized_accuracy
<< ". (condition: (FP32_" << prefix << " - INT8_" << prefix
<< ") <= threshold)";
<< ". (condition: (FP32_" << prefix << " - " << data_type_name
<< "_" << prefix << ") <= threshold)";
LOG(INFO) << "FP32: avg " << prefix << std::fixed << std::setw(6)
<< std::setprecision(4) << avg_acc_fp32;
LOG(INFO) << "INT8: avg " << prefix << std::fixed << std::setw(6)
<< std::setprecision(4) << avg_acc_int8;
<< std::setprecision(4) << avg_acc_ref;
LOG(INFO) << data_type_name << ": avg " << prefix << std::fixed
<< std::setw(6) << std::setprecision(4) << avg_acc;
}
void SummarizePerformance(const char *title, float sample) {
......@@ -677,8 +678,9 @@ void SummarizePerformance(const char *title, float sample) {
void SummarizePerformance(const char *title_fp32, float sample_latency_fp32,
const char *title, float sample_latency) {
SummarizePerformance(title_fp32, sample_latency_fp32);
SummarizePerformance(title, sample_latency);
if (FLAGS_enable_fp32) SummarizePerformance(title_fp32, sample_latency_fp32);
if (FLAGS_enable_int8 || FLAGS_enable_bf16)
SummarizePerformance(title, sample_latency);
}
float CompareAccuracyOne(
......@@ -733,7 +735,7 @@ void CompareAccuracy(
const std::vector<std::vector<PaddleTensor>> &output_slots_quant,
const std::vector<std::vector<PaddleTensor>> &output_slots_ref,
int compared_idx) {
if ((FLAGS_enable_fp32 && FLAGS_enable_int8) &&
if ((FLAGS_enable_fp32 && (FLAGS_enable_int8 || FLAGS_enable_bf16)) &&
(output_slots_quant.size() == 0 || output_slots_ref.size()) == 0)
throw std::invalid_argument(
"CompareAccuracy: output_slots vector is empty.");
......@@ -741,7 +743,7 @@ void CompareAccuracy(
float avg_acc_quant = 0.0;
float avg_acc_ref = 0.0;
if (FLAGS_enable_int8)
if (FLAGS_enable_int8 || FLAGS_enable_bf16)
avg_acc_quant = CompareAccuracyOne(output_slots_quant, compared_idx);
if (FLAGS_enable_fp32)
......@@ -751,9 +753,9 @@ void CompareAccuracy(
if (FLAGS_enable_fp32) CHECK_GT(avg_acc_ref, 0.0);
if (FLAGS_enable_int8) CHECK_GT(avg_acc_quant, 0.0);
if (FLAGS_enable_int8 || FLAGS_enable_bf16) CHECK_GT(avg_acc_quant, 0.0);
if (FLAGS_enable_fp32 && FLAGS_enable_int8)
if (FLAGS_enable_fp32 && (FLAGS_enable_int8 || FLAGS_enable_bf16))
CHECK_LE(avg_acc_ref - avg_acc_quant, FLAGS_quantized_accuracy);
}
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册