Skip to content

GitLab

Unverified Commit 26c16324 authored by Evezerest's avatar Evezerest Committed by GitHub
Browse files

Merge branch 'dygraph' into dygraph

parents d9549ce6 0b37c118
Hide whitespace changes
Inline Side-by-side
Showing with 477 additions and 236 deletions
PTDN/test_inference_cpp.sh → test_tipc/test_inference_cpp.sh
View file @ 26c16324
#!/bin/bash
source tests/common_func.sh
source test_tipc/common_func.sh
FILENAME=$1
dataline=$(awk 'NR==52, NR==66{print}' $FILENAME)
......@@ -35,7 +35,7 @@ cpp_benchmark_key=$(func_parser_key "${lines[14]}")
cpp_benchmark_value=$(func_parser_value "${lines[14]}")
LOG_PATH="./tests/output"
LOG_PATH="./test_tipc/output"
mkdir -p ${LOG_PATH}
status_log="${LOG_PATH}/results_cpp.log"
......
test_tipc/test_lite.sh 0 → 100644
View file @ 26c16324
#!/bin/bash
source ./common_func.sh
export LD_LIBRARY_PATH=${PWD}:$LD_LIBRARY_PATH
FILENAME=$1
dataline=$(awk 'NR==101, NR==110{print}' $FILENAME)
echo $dataline
# parser params
IFS=$'\n'
lines=(${dataline})
# parser lite inference
lite_inference_cmd=$(func_parser_value "${lines[1]}")
lite_model_dir_list=$(func_parser_value "${lines[2]}")
lite_cpu_threads_list=$(func_parser_value "${lines[3]}")
lite_batch_size_list=$(func_parser_value "${lines[4]}")
lite_power_mode_list=$(func_parser_value "${lines[5]}")
lite_infer_img_dir_list=$(func_parser_value "${lines[6]}")
lite_config_dir=$(func_parser_value "${lines[7]}")
lite_rec_dict_dir=$(func_parser_value "${lines[8]}")
lite_benchmark_value=$(func_parser_value "${lines[9]}")
LOG_PATH="./output"
mkdir -p ${LOG_PATH}
status_log="${LOG_PATH}/results.log"
function func_lite(){
IFS='|'
_script=$1
_lite_model=$2
_log_path=$3
_img_dir=$4
_config=$5
if [[ $lite_model =~ "slim" ]]; then
precision="INT8"
else
precision="FP32"
fi
is_single_img=$(echo $_img_dir | grep -E ".jpg|.jpeg|.png|.JPEG|.JPG")
if [[ "$is_single_img" != "" ]]; then
single_img="True"
else
single_img="False"
fi
# lite inference
for num_threads in ${lite_cpu_threads_list[*]}; do
for power_mode in ${lite_power_mode_list[*]}; do
for batchsize in ${lite_batch_size_list[*]}; do
model_name=$(echo $lite_model | awk -F "/" '{print $NF}')
_save_log_path="${_log_path}/lite_${model_name}_precision_${precision}_batchsize_${batchsize}_threads_${num_threads}_powermode_${power_mode}_singleimg_${single_img}.log"
command="${_script} ${lite_model} ${precision} ${num_threads} ${batchsize} ${power_mode} ${_img_dir} ${_config} ${lite_benchmark_value} > ${_save_log_path} 2>&1"
eval ${command}
status_check $? "${command}" "${status_log}"
done
done
done
}
echo "################### run test ###################"
IFS="|"
for lite_model in ${lite_model_dir_list[*]}; do
#run lite inference
for img_dir in ${lite_infer_img_dir_list[*]}; do
func_lite "${lite_inference_cmd}" "${lite_model}" "${LOG_PATH}" "${img_dir}" "${lite_config_dir}"
done
done
test_tipc/test_paddle2onnx.sh 0 → 100644
View file @ 26c16324
#!/bin/bash
source test_tipc/common_func.sh
FILENAME=$1
dataline=$(cat ${FILENAME})
lines=(${dataline})
# common params
model_name=$(func_parser_value "${lines[1]}")
python=$(func_parser_value "${lines[2]}")
# parser params
dataline=$(awk 'NR==111, NR==123{print}' $FILENAME)
IFS=$'\n'
lines=(${dataline})
# parser paddle2onnx
padlle2onnx_cmd=$(func_parser_value "${lines[1]}")
infer_model_dir_key=$(func_parser_key "${lines[2]}")
infer_model_dir_value=$(func_parser_value "${lines[2]}")
model_filename_key=$(func_parser_key "${lines[3]}")
model_filename_value=$(func_parser_value "${lines[3]}")
params_filename_key=$(func_parser_key "${lines[4]}")
params_filename_value=$(func_parser_value "${lines[4]}")
save_file_key=$(func_parser_key "${lines[5]}")
save_file_value=$(func_parser_value "${lines[5]}")
opset_version_key=$(func_parser_key "${lines[6]}")
opset_version_value=$(func_parser_value "${lines[6]}")
enable_onnx_checker_key=$(func_parser_key "${lines[7]}")
enable_onnx_checker_value=$(func_parser_value "${lines[7]}")
# parser onnx inference
inference_py=$(func_parser_value "${lines[8]}")
use_gpu_key=$(func_parser_key "${lines[9]}")
use_gpu_value=$(func_parser_value "${lines[9]}")
det_model_key=$(func_parser_key "${lines[10]}")
image_dir_key=$(func_parser_key "${lines[11]}")
image_dir_value=$(func_parser_value "${lines[11]}")
LOG_PATH="./test_tipc/output"
mkdir -p ./test_tipc/output
status_log="${LOG_PATH}/results_paddle2onnx.log"
function func_paddle2onnx(){
IFS='|'
_script=$1
# paddle2onnx
_save_log_path="${LOG_PATH}/paddle2onnx_infer_cpu.log"
set_dirname=$(func_set_params "${infer_model_dir_key}" "${infer_model_dir_value}")
set_model_filename=$(func_set_params "${model_filename_key}" "${model_filename_value}")
set_params_filename=$(func_set_params "${params_filename_key}" "${params_filename_value}")
set_save_model=$(func_set_params "${save_file_key}" "${save_file_value}")
set_opset_version=$(func_set_params "${opset_version_key}" "${opset_version_value}")
set_enable_onnx_checker=$(func_set_params "${enable_onnx_checker_key}" "${enable_onnx_checker_value}")
trans_model_cmd="${padlle2onnx_cmd} ${set_dirname} ${set_model_filename} ${set_params_filename} ${set_save_model} ${set_opset_version} ${set_enable_onnx_checker}"
eval $trans_model_cmd
last_status=${PIPESTATUS[0]}
status_check $last_status "${trans_model_cmd}" "${status_log}"
# python inference
set_gpu=$(func_set_params "${use_gpu_key}" "${use_gpu_value}")
set_model_dir=$(func_set_params "${det_model_key}" "${save_file_value}")
set_img_dir=$(func_set_params "${image_dir_key}" "${image_dir_value}")
infer_model_cmd="${python} ${inference_py} ${set_gpu} ${set_img_dir} ${set_model_dir} --use_onnx=True > ${_save_log_path} 2>&1 "
eval $infer_model_cmd
status_check $last_status "${infer_model_cmd}" "${status_log}"
}
echo "################### run test ###################"
export Count=0
IFS="|"
func_paddle2onnx
\ No newline at end of file
PTDN/test_serving.sh → test_tipc/test_serving.sh
View file @ 26c16324
#!/bin/bash
source tests/common_func.sh
source test_tipc/common_func.sh
FILENAME=$1
dataline=$(awk 'NR==67, NR==83{print}' $FILENAME)
......@@ -36,8 +36,8 @@ web_precision_key=$(func_parser_key "${lines[15]}")
web_precision_list=$(func_parser_value "${lines[15]}")
pipeline_py=$(func_parser_value "${lines[16]}")
LOG_PATH="../../tests/output"
mkdir -p ./tests/output
LOG_PATH="../../test_tipc/output"
mkdir -p ./test_tipc/output
status_log="${LOG_PATH}/results_serving.log"
function func_serving(){
......
PTDN/test_train_inference_python.sh → test_tipc/test_train_inference_python.sh
View file @ 26c16324
#!/bin/bash
source tests/common_func.sh
source test_tipc/common_func.sh
FILENAME=$1
# MODE be one of ['lite_train_infer' 'whole_infer' 'whole_train_infer', 'infer', 'klquant_infer']
# MODE be one of ['lite_train_lite_infer' 'lite_train_whole_infer' 'whole_train_whole_infer', 'whole_infer', 'klquant_whole_infer']
MODE=$2
dataline=$(awk 'NR==1, NR==51{print}' $FILENAME)
......@@ -59,6 +59,7 @@ export_key1=$(func_parser_key "${lines[33]}")
export_value1=$(func_parser_value "${lines[33]}")
export_key2=$(func_parser_key "${lines[34]}")
export_value2=$(func_parser_value "${lines[34]}")
inference_dir=$(func_parser_value "${lines[35]}")
# parser inference model
infer_model_dir_list=$(func_parser_value "${lines[36]}")
......@@ -88,7 +89,7 @@ infer_key1=$(func_parser_key "${lines[50]}")
infer_value1=$(func_parser_value "${lines[50]}")
# parser klquant_infer
if [ ${MODE} = "klquant_infer" ]; then
if [ ${MODE} = "klquant_whole_infer" ]; then
dataline=$(awk 'NR==82, NR==98{print}' $FILENAME)
lines=(${dataline})
# parser inference model
......@@ -119,7 +120,7 @@ if [ ${MODE} = "klquant_infer" ]; then
infer_value1=$(func_parser_value "${lines[15]}")
fi
LOG_PATH="./tests/output"
LOG_PATH="./test_tipc/output"
mkdir -p ${LOG_PATH}
status_log="${LOG_PATH}/results_python.log"
......@@ -202,7 +203,7 @@ function func_inference(){
done
}
if [ ${MODE} = "infer" ] || [ ${MODE} = "klquant_infer" ]; then
if [ ${MODE} = "whole_infer" ] || [ ${MODE} = "klquant_whole_infer" ]; then
GPUID=$3
if [ ${#GPUID} -le 0 ];then
env=" "
......@@ -245,6 +246,7 @@ else
for gpu in ${gpu_list[*]}; do
use_gpu=${USE_GPU_KEY[Count]}
Count=$(($Count + 1))
ips=""
if [ ${gpu} = "-1" ];then
env=""
elif [ ${#gpu} -le 1 ];then
......@@ -264,6 +266,11 @@ else
env=" "
fi
for autocast in ${autocast_list[*]}; do
if [ ${autocast} = "amp" ]; then
set_amp_config="Global.use_amp=True Global.scale_loss=1024.0 Global.use_dynamic_loss_scaling=True"
else
set_amp_config=" "
fi
for trainer in ${trainer_list[*]}; do
flag_quant=False
if [ ${trainer} = ${pact_key} ]; then
......@@ -290,7 +297,6 @@ else
if [ ${run_train} = "null" ]; then
continue
fi
set_autocast=$(func_set_params "${autocast_key}" "${autocast}")
set_epoch=$(func_set_params "${epoch_key}" "${epoch_num}")
set_pretrain=$(func_set_params "${pretrain_model_key}" "${pretrain_model_value}")
......@@ -306,11 +312,11 @@ else
set_save_model=$(func_set_params "${save_model_key}" "${save_log}")
if [ ${#gpu} -le 2 ];then # train with cpu or single gpu
cmd="${python} ${run_train} ${set_use_gpu} ${set_save_model} ${set_epoch} ${set_pretrain} ${set_autocast} ${set_batchsize} ${set_train_params1} "
elif [ ${#gpu} -le 15 ];then # train with multi-gpu
cmd="${python} -m paddle.distributed.launch --gpus=${gpu} ${run_train} ${set_save_model} ${set_epoch} ${set_pretrain} ${set_autocast} ${set_batchsize} ${set_train_params1}"
cmd="${python} ${run_train} ${set_use_gpu} ${set_save_model} ${set_epoch} ${set_pretrain} ${set_autocast} ${set_batchsize} ${set_train_params1} ${set_amp_config} "
elif [ ${#ips} -le 26 ];then # train with multi-gpu
cmd="${python} -m paddle.distributed.launch --gpus=${gpu} ${run_train} ${set_use_gpu} ${set_save_model} ${set_epoch} ${set_pretrain} ${set_autocast} ${set_batchsize} ${set_train_params1} ${set_amp_config}"
else # train with multi-machine
cmd="${python} -m paddle.distributed.launch --ips=${ips} --gpus=${gpu} ${run_train} ${set_save_model} ${set_pretrain} ${set_epoch} ${set_autocast} ${set_batchsize} ${set_train_params1}"
cmd="${python} -m paddle.distributed.launch --ips=${ips} --gpus=${gpu} ${run_train} ${set_use_gpu} ${set_save_model} ${set_pretrain} ${set_epoch} ${set_autocast} ${set_batchsize} ${set_train_params1} ${set_amp_config}"
fi
# run train
eval "unset CUDA_VISIBLE_DEVICES"
......@@ -342,7 +348,13 @@ else
#run inference
eval $env
save_infer_path="${save_log}"
func_inference "${python}" "${inference_py}" "${save_infer_path}" "${LOG_PATH}" "${train_infer_img_dir}" "${flag_quant}"
if [ ${inference_dir} != "null" ] && [ ${inference_dir} != '##' ]; then
infer_model_dir="${save_infer_path}/${inference_dir}"
else
infer_model_dir=${save_infer_path}
fi
func_inference "${python}" "${inference_py}" "${infer_model_dir}" "${LOG_PATH}" "${train_infer_img_dir}" "${flag_quant}"
eval "unset CUDA_VISIBLE_DEVICES"
fi
done # done with: for trainer in ${trainer_list[*]}; do
......
tools/infer/predict_cls.py
View file @ 26c16324
......@@ -47,6 +47,7 @@ class TextClassifier(object):
self.postprocess_op = build_post_process(postprocess_params)
self.predictor, self.input_tensor, self.output_tensors, _ = \
utility.create_predictor(args, 'cls', logger)
self.use_onnx = args.use_onnx
def resize_norm_img(self, img):
imgC, imgH, imgW = self.cls_image_shape
......@@ -100,10 +101,16 @@ class TextClassifier(object):
norm_img_batch = np.concatenate(norm_img_batch)
norm_img_batch = norm_img_batch.copy()
self.input_tensor.copy_from_cpu(norm_img_batch)
self.predictor.run()
prob_out = self.output_tensors[0].copy_to_cpu()
self.predictor.try_shrink_memory()
if self.use_onnx:
input_dict = {}
input_dict[self.input_tensor.name] = norm_img_batch
outputs = self.predictor.run(self.output_tensors, input_dict)
prob_out = outputs[0]
else:
self.input_tensor.copy_from_cpu(norm_img_batch)
self.predictor.run()
prob_out = self.output_tensors[0].copy_to_cpu()
self.predictor.try_shrink_memory()
cls_result = self.postprocess_op(prob_out)
elapse += time.time() - starttime
for rno in range(len(cls_result)):
......
tools/infer/predict_det.py
View file @ 26c16324
......@@ -38,6 +38,7 @@ class TextDetector(object):
def __init__(self, args):
self.args = args
self.det_algorithm = args.det_algorithm
self.use_onnx = args.use_onnx
pre_process_list = [{
'DetResizeForTest': {
'limit_side_len': args.det_limit_side_len,
......@@ -100,7 +101,12 @@ class TextDetector(object):
else:
logger.info("unknown det_algorithm:{}".format(self.det_algorithm))
sys.exit(0)
if self.use_onnx:
pre_process_list[0] = {
'DetResizeForTest': {
'image_shape': [640, 640]
}
}
self.preprocess_op = create_operators(pre_process_list)
self.postprocess_op = build_post_process(postprocess_params)
self.predictor, self.input_tensor, self.output_tensors, self.config = utility.create_predictor(
......@@ -198,15 +204,19 @@ class TextDetector(object):
if self.args.benchmark:
self.autolog.times.stamp()
self.input_tensor.copy_from_cpu(img)
self.predictor.run()
outputs = []
for output_tensor in self.output_tensors:
output = output_tensor.copy_to_cpu()
outputs.append(output)
if self.args.benchmark:
self.autolog.times.stamp()
if self.use_onnx:
input_dict = {}
input_dict[self.input_tensor.name] = img
outputs = self.predictor.run(self.output_tensors, input_dict)
else:
self.input_tensor.copy_from_cpu(img)
self.predictor.run()
outputs = []
for output_tensor in self.output_tensors:
output = output_tensor.copy_to_cpu()
outputs.append(output)
if self.args.benchmark:
self.autolog.times.stamp()
preds = {}
if self.det_algorithm == "EAST":
......
tools/infer/predict_rec.py
View file @ 26c16324
......@@ -73,6 +73,7 @@ class TextRecognizer(object):
self.predictor, self.input_tensor, self.output_tensors, self.config = \
utility.create_predictor(args, 'rec', logger)
self.benchmark = args.benchmark
self.use_onnx = args.use_onnx
if args.benchmark:
import auto_log
pid = os.getpid()
......@@ -106,8 +107,9 @@ class TextRecognizer(object):
return norm_img.astype(np.float32) / 128. - 1.
assert imgC == img.shape[2]
max_wh_ratio = max(max_wh_ratio, imgW / imgH)
imgW = int((32 * max_wh_ratio))
if self.use_onnx:
imgW = 100
h, w = img.shape[:2]
ratio = w / float(h)
if math.ceil(imgH * ratio) > imgW:
......@@ -297,51 +299,72 @@ class TextRecognizer(object):
gsrm_slf_attn_bias1_list,
gsrm_slf_attn_bias2_list,
]
input_names = self.predictor.get_input_names()
for i in range(len(input_names)):
input_tensor = self.predictor.get_input_handle(input_names[
i])
input_tensor.copy_from_cpu(inputs[i])
self.predictor.run()
outputs = []
for output_tensor in self.output_tensors:
output = output_tensor.copy_to_cpu()
outputs.append(output)
if self.benchmark:
self.autolog.times.stamp()
preds = {"predict": outputs[2]}
if self.use_onnx:
input_dict = {}
input_dict[self.input_tensor.name] = norm_img_batch
outputs = self.predictor.run(self.output_tensors,
input_dict)
preds = {"predict": outputs[2]}
else:
input_names = self.predictor.get_input_names()
for i in range(len(input_names)):
input_tensor = self.predictor.get_input_handle(
input_names[i])
input_tensor.copy_from_cpu(inputs[i])
self.predictor.run()
outputs = []
for output_tensor in self.output_tensors:
output = output_tensor.copy_to_cpu()
outputs.append(output)
if self.benchmark:
self.autolog.times.stamp()
preds = {"predict": outputs[2]}
elif self.rec_algorithm == "SAR":
valid_ratios = np.concatenate(valid_ratios)
inputs = [
norm_img_batch,
valid_ratios,
]
input_names = self.predictor.get_input_names()
for i in range(len(input_names)):
input_tensor = self.predictor.get_input_handle(input_names[
i])
input_tensor.copy_from_cpu(inputs[i])
self.predictor.run()
outputs = []
for output_tensor in self.output_tensors:
output = output_tensor.copy_to_cpu()
outputs.append(output)
if self.benchmark:
self.autolog.times.stamp()
preds = outputs[0]
else:
self.input_tensor.copy_from_cpu(norm_img_batch)
self.predictor.run()
outputs = []
for output_tensor in self.output_tensors:
output = output_tensor.copy_to_cpu()
outputs.append(output)
if self.benchmark:
self.autolog.times.stamp()
if len(outputs) != 1:
preds = outputs
if self.use_onnx:
input_dict = {}
input_dict[self.input_tensor.name] = norm_img_batch
outputs = self.predictor.run(self.output_tensors,
input_dict)
preds = outputs[0]
else:
input_names = self.predictor.get_input_names()
for i in range(len(input_names)):
input_tensor = self.predictor.get_input_handle(
input_names[i])
input_tensor.copy_from_cpu(inputs[i])
self.predictor.run()
outputs = []
for output_tensor in self.output_tensors:
output = output_tensor.copy_to_cpu()
outputs.append(output)
if self.benchmark:
self.autolog.times.stamp()
preds = outputs[0]
else:
if self.use_onnx:
input_dict = {}
input_dict[self.input_tensor.name] = norm_img_batch
outputs = self.predictor.run(self.output_tensors,
input_dict)
preds = outputs[0]
else:
self.input_tensor.copy_from_cpu(norm_img_batch)
self.predictor.run()
outputs = []
for output_tensor in self.output_tensors:
output = output_tensor.copy_to_cpu()
outputs.append(output)
if self.benchmark:
self.autolog.times.stamp()
if len(outputs) != 1:
preds = outputs
else:
preds = outputs[0]
rec_result = self.postprocess_op(preds)
for rno in range(len(rec_result)):
rec_res[indices[beg_img_no + rno]] = rec_result[rno]
......
tools/infer/utility.py
View file @ 26c16324
......@@ -121,6 +121,7 @@ def init_args():
parser.add_argument("--save_log_path", type=str, default="./log_output/")
parser.add_argument("--show_log", type=str2bool, default=True)
parser.add_argument("--use_onnx", type=str2bool, default=False)
return parser
......@@ -144,152 +145,163 @@ def create_predictor(args, mode, logger):
if model_dir is None:
logger.info("not find {} model file path {}".format(mode, model_dir))
sys.exit(0)
model_file_path = model_dir + "/inference.pdmodel"
params_file_path = model_dir + "/inference.pdiparams"
if not os.path.exists(model_file_path):
raise ValueError("not find model file path {}".format(model_file_path))
if not os.path.exists(params_file_path):
raise ValueError("not find params file path {}".format(
params_file_path))
config = inference.Config(model_file_path, params_file_path)
if hasattr(args, 'precision'):
if args.precision == "fp16" and args.use_tensorrt:
precision = inference.PrecisionType.Half
elif args.precision == "int8":
precision = inference.PrecisionType.Int8
else:
precision = inference.PrecisionType.Float32
if args.use_onnx:
import onnxruntime as ort
model_file_path = model_dir
if not os.path.exists(model_file_path):
raise ValueError("not find model file path {}".format(
model_file_path))
sess = ort.InferenceSession(model_file_path)
return sess, sess.get_inputs()[0], None, None
else:
precision = inference.PrecisionType.Float32
if args.use_gpu:
gpu_id = get_infer_gpuid()
if gpu_id is None:
raise ValueError(
"Not found GPU in current device. Please check your device or set args.use_gpu as False"
)
config.enable_use_gpu(args.gpu_mem, 0)
if args.use_tensorrt:
config.enable_tensorrt_engine(
precision_mode=precision,
max_batch_size=args.max_batch_size,
min_subgraph_size=args.min_subgraph_size)
# skip the minmum trt subgraph
if mode == "det":
min_input_shape = {
"x": [1, 3, 50, 50],
"conv2d_92.tmp_0": [1, 120, 20, 20],
"conv2d_91.tmp_0": [1, 24, 10, 10],
"conv2d_59.tmp_0": [1, 96, 20, 20],
"nearest_interp_v2_1.tmp_0": [1, 256, 10, 10],
"nearest_interp_v2_2.tmp_0": [1, 256, 20, 20],
"conv2d_124.tmp_0": [1, 256, 20, 20],
"nearest_interp_v2_3.tmp_0": [1, 64, 20, 20],
"nearest_interp_v2_4.tmp_0": [1, 64, 20, 20],
"nearest_interp_v2_5.tmp_0": [1, 64, 20, 20],
"elementwise_add_7": [1, 56, 2, 2],
"nearest_interp_v2_0.tmp_0": [1, 256, 2, 2]
}
max_input_shape = {
"x": [1, 3, 2000, 2000],
"conv2d_92.tmp_0": [1, 120, 400, 400],
"conv2d_91.tmp_0": [1, 24, 200, 200],
"conv2d_59.tmp_0": [1, 96, 400, 400],
"nearest_interp_v2_1.tmp_0": [1, 256, 200, 200],
"conv2d_124.tmp_0": [1, 256, 400, 400],
"nearest_interp_v2_2.tmp_0": [1, 256, 400, 400],
"nearest_interp_v2_3.tmp_0": [1, 64, 400, 400],
"nearest_interp_v2_4.tmp_0": [1, 64, 400, 400],
"nearest_interp_v2_5.tmp_0": [1, 64, 400, 400],
"elementwise_add_7": [1, 56, 400, 400],
"nearest_interp_v2_0.tmp_0": [1, 256, 400, 400]
}
opt_input_shape = {
"x": [1, 3, 640, 640],
"conv2d_92.tmp_0": [1, 120, 160, 160],
"conv2d_91.tmp_0": [1, 24, 80, 80],
"conv2d_59.tmp_0": [1, 96, 160, 160],
"nearest_interp_v2_1.tmp_0": [1, 256, 80, 80],
"nearest_interp_v2_2.tmp_0": [1, 256, 160, 160],
"conv2d_124.tmp_0": [1, 256, 160, 160],
"nearest_interp_v2_3.tmp_0": [1, 64, 160, 160],
"nearest_interp_v2_4.tmp_0": [1, 64, 160, 160],
"nearest_interp_v2_5.tmp_0": [1, 64, 160, 160],
"elementwise_add_7": [1, 56, 40, 40],
"nearest_interp_v2_0.tmp_0": [1, 256, 40, 40]
}
min_pact_shape = {
"nearest_interp_v2_26.tmp_0": [1, 256, 20, 20],
"nearest_interp_v2_27.tmp_0": [1, 64, 20, 20],
"nearest_interp_v2_28.tmp_0": [1, 64, 20, 20],
"nearest_interp_v2_29.tmp_0": [1, 64, 20, 20]
}
max_pact_shape = {
"nearest_interp_v2_26.tmp_0": [1, 256, 400, 400],
"nearest_interp_v2_27.tmp_0": [1, 64, 400, 400],
"nearest_interp_v2_28.tmp_0": [1, 64, 400, 400],
"nearest_interp_v2_29.tmp_0": [1, 64, 400, 400]
}
opt_pact_shape = {
"nearest_interp_v2_26.tmp_0": [1, 256, 160, 160],
"nearest_interp_v2_27.tmp_0": [1, 64, 160, 160],
"nearest_interp_v2_28.tmp_0": [1, 64, 160, 160],
"nearest_interp_v2_29.tmp_0": [1, 64, 160, 160]
}
min_input_shape.update(min_pact_shape)
max_input_shape.update(max_pact_shape)
opt_input_shape.update(opt_pact_shape)
elif mode == "rec":
min_input_shape = {"x": [1, 3, 32, 10]}
max_input_shape = {"x": [args.rec_batch_num, 3, 32, 2000]}
opt_input_shape = {"x": [args.rec_batch_num, 3, 32, 320]}
elif mode == "cls":
min_input_shape = {"x": [1, 3, 48, 10]}
max_input_shape = {"x": [args.rec_batch_num, 3, 48, 2000]}
opt_input_shape = {"x": [args.rec_batch_num, 3, 48, 320]}
model_file_path = model_dir + "/inference.pdmodel"
params_file_path = model_dir + "/inference.pdiparams"
if not os.path.exists(model_file_path):
raise ValueError("not find model file path {}".format(
model_file_path))
if not os.path.exists(params_file_path):
raise ValueError("not find params file path {}".format(
params_file_path))
config = inference.Config(model_file_path, params_file_path)
if hasattr(args, 'precision'):
if args.precision == "fp16" and args.use_tensorrt:
precision = inference.PrecisionType.Half
elif args.precision == "int8":
precision = inference.PrecisionType.Int8
else:
precision = inference.PrecisionType.Float32
else:
min_input_shape = {"x": [1, 3, 10, 10]}
max_input_shape = {"x": [1, 3, 1000, 1000]}
opt_input_shape = {"x": [1, 3, 500, 500]}
config.set_trt_dynamic_shape_info(min_input_shape, max_input_shape,
opt_input_shape)
precision = inference.PrecisionType.Float32
if args.use_gpu:
gpu_id = get_infer_gpuid()
if gpu_id is None:
raise ValueError(
"Not found GPU in current device. Please check your device or set args.use_gpu as False"
)
config.enable_use_gpu(args.gpu_mem, 0)
if args.use_tensorrt:
config.enable_tensorrt_engine(
precision_mode=precision,
max_batch_size=args.max_batch_size,
min_subgraph_size=args.min_subgraph_size)
# skip the minmum trt subgraph
if mode == "det":
min_input_shape = {
"x": [1, 3, 50, 50],
"conv2d_92.tmp_0": [1, 120, 20, 20],
"conv2d_91.tmp_0": [1, 24, 10, 10],
"conv2d_59.tmp_0": [1, 96, 20, 20],
"nearest_interp_v2_1.tmp_0": [1, 256, 10, 10],
"nearest_interp_v2_2.tmp_0": [1, 256, 20, 20],
"conv2d_124.tmp_0": [1, 256, 20, 20],
"nearest_interp_v2_3.tmp_0": [1, 64, 20, 20],
"nearest_interp_v2_4.tmp_0": [1, 64, 20, 20],
"nearest_interp_v2_5.tmp_0": [1, 64, 20, 20],
"elementwise_add_7": [1, 56, 2, 2],
"nearest_interp_v2_0.tmp_0": [1, 256, 2, 2]
}
max_input_shape = {
"x": [1, 3, 2000, 2000],
"conv2d_92.tmp_0": [1, 120, 400, 400],
"conv2d_91.tmp_0": [1, 24, 200, 200],
"conv2d_59.tmp_0": [1, 96, 400, 400],
"nearest_interp_v2_1.tmp_0": [1, 256, 200, 200],
"conv2d_124.tmp_0": [1, 256, 400, 400],
"nearest_interp_v2_2.tmp_0": [1, 256, 400, 400],
"nearest_interp_v2_3.tmp_0": [1, 64, 400, 400],
"nearest_interp_v2_4.tmp_0": [1, 64, 400, 400],
"nearest_interp_v2_5.tmp_0": [1, 64, 400, 400],
"elementwise_add_7": [1, 56, 400, 400],
"nearest_interp_v2_0.tmp_0": [1, 256, 400, 400]
}
opt_input_shape = {
"x": [1, 3, 640, 640],
"conv2d_92.tmp_0": [1, 120, 160, 160],
"conv2d_91.tmp_0": [1, 24, 80, 80],
"conv2d_59.tmp_0": [1, 96, 160, 160],
"nearest_interp_v2_1.tmp_0": [1, 256, 80, 80],
"nearest_interp_v2_2.tmp_0": [1, 256, 160, 160],
"conv2d_124.tmp_0": [1, 256, 160, 160],
"nearest_interp_v2_3.tmp_0": [1, 64, 160, 160],
"nearest_interp_v2_4.tmp_0": [1, 64, 160, 160],
"nearest_interp_v2_5.tmp_0": [1, 64, 160, 160],
"elementwise_add_7": [1, 56, 40, 40],
"nearest_interp_v2_0.tmp_0": [1, 256, 40, 40]
}
min_pact_shape = {
"nearest_interp_v2_26.tmp_0": [1, 256, 20, 20],
"nearest_interp_v2_27.tmp_0": [1, 64, 20, 20],
"nearest_interp_v2_28.tmp_0": [1, 64, 20, 20],
"nearest_interp_v2_29.tmp_0": [1, 64, 20, 20]
}
max_pact_shape = {
"nearest_interp_v2_26.tmp_0": [1, 256, 400, 400],
"nearest_interp_v2_27.tmp_0": [1, 64, 400, 400],
"nearest_interp_v2_28.tmp_0": [1, 64, 400, 400],
"nearest_interp_v2_29.tmp_0": [1, 64, 400, 400]
}
opt_pact_shape = {
"nearest_interp_v2_26.tmp_0": [1, 256, 160, 160],
"nearest_interp_v2_27.tmp_0": [1, 64, 160, 160],
"nearest_interp_v2_28.tmp_0": [1, 64, 160, 160],
"nearest_interp_v2_29.tmp_0": [1, 64, 160, 160]
}
min_input_shape.update(min_pact_shape)
max_input_shape.update(max_pact_shape)
opt_input_shape.update(opt_pact_shape)
elif mode == "rec":
min_input_shape = {"x": [1, 3, 32, 10]}
max_input_shape = {"x": [args.rec_batch_num, 3, 32, 2000]}
opt_input_shape = {"x": [args.rec_batch_num, 3, 32, 320]}
elif mode == "cls":
min_input_shape = {"x": [1, 3, 48, 10]}
max_input_shape = {"x": [args.rec_batch_num, 3, 48, 2000]}
opt_input_shape = {"x": [args.rec_batch_num, 3, 48, 320]}
else:
min_input_shape = {"x": [1, 3, 10, 10]}
max_input_shape = {"x": [1, 3, 1000, 1000]}
opt_input_shape = {"x": [1, 3, 500, 500]}
config.set_trt_dynamic_shape_info(min_input_shape, max_input_shape,
opt_input_shape)
else:
config.disable_gpu()
if hasattr(args, "cpu_threads"):
config.set_cpu_math_library_num_threads(args.cpu_threads)
else:
# default cpu threads as 10
config.set_cpu_math_library_num_threads(10)
if args.enable_mkldnn:
# cache 10 different shapes for mkldnn to avoid memory leak
config.set_mkldnn_cache_capacity(10)
config.enable_mkldnn()
if args.precision == "fp16":
config.enable_mkldnn_bfloat16()
# enable memory optim
config.enable_memory_optim()
config.disable_glog_info()
config.delete_pass("conv_transpose_eltwiseadd_bn_fuse_pass")
if mode == 'table':
config.delete_pass("fc_fuse_pass") # not supported for table
config.switch_use_feed_fetch_ops(False)
config.switch_ir_optim(True)
# create predictor
predictor = inference.create_predictor(config)
input_names = predictor.get_input_names()
for name in input_names:
input_tensor = predictor.get_input_handle(name)
output_names = predictor.get_output_names()
output_tensors = []
for output_name in output_names:
output_tensor = predictor.get_output_handle(output_name)
output_tensors.append(output_tensor)
return predictor, input_tensor, output_tensors, config
config.disable_gpu()
if hasattr(args, "cpu_threads"):
config.set_cpu_math_library_num_threads(args.cpu_threads)
else:
# default cpu threads as 10
config.set_cpu_math_library_num_threads(10)
if args.enable_mkldnn:
# cache 10 different shapes for mkldnn to avoid memory leak
config.set_mkldnn_cache_capacity(10)
config.enable_mkldnn()
if args.precision == "fp16":
config.enable_mkldnn_bfloat16()
# enable memory optim
config.enable_memory_optim()
config.disable_glog_info()
config.delete_pass("conv_transpose_eltwiseadd_bn_fuse_pass")
if mode == 'table':
config.delete_pass("fc_fuse_pass") # not supported for table
config.switch_use_feed_fetch_ops(False)
config.switch_ir_optim(True)
# create predictor
predictor = inference.create_predictor(config)
input_names = predictor.get_input_names()
for name in input_names:
input_tensor = predictor.get_input_handle(name)
output_names = predictor.get_output_names()
output_tensors = []
for output_name in output_names:
output_tensor = predictor.get_output_handle(output_name)
output_tensors.append(output_tensor)
return predictor, input_tensor, output_tensors, config
def get_infer_gpuid():
......
tools/program.py
View file @ 26c16324
......@@ -159,7 +159,8 @@ def train(config,
eval_class,
pre_best_model_dict,
logger,
vdl_writer=None):
vdl_writer=None,
scaler=None):
cal_metric_during_train = config['Global'].get('cal_metric_during_train',
False)
log_smooth_window = config['Global']['log_smooth_window']
......@@ -211,33 +212,49 @@ def train(config,
for epoch in range(start_epoch, epoch_num + 1):
train_dataloader = build_dataloader(
config, 'Train', device, logger, seed=epoch)
train_batch_cost = 0.0
train_reader_cost = 0.0
batch_sum = 0
batch_start = time.time()
train_run_cost = 0.0
total_samples = 0
reader_start = time.time()
max_iter = len(train_dataloader) - 1 if platform.system(
) == "Windows" else len(train_dataloader)
for idx, batch in enumerate(train_dataloader):
profiler.add_profiler_step(profiler_options)
train_reader_cost += time.time() - batch_start
train_reader_cost += time.time() - reader_start
if idx >= max_iter:
break
lr = optimizer.get_lr()
images = batch[0]
if use_srn:
model_average = True
if model_type == 'table' or extra_input:
preds = model(images, data=batch[1:])
train_start = time.time()
# use amp
if scaler:
with paddle.amp.auto_cast():
if model_type == 'table' or extra_input:
preds = model(images, data=batch[1:])
else:
preds = model(images)
else:
preds = model(images)
if model_type == 'table' or extra_input:
preds = model(images, data=batch[1:])
else:
preds = model(images)
loss = loss_class(preds, batch)
avg_loss = loss['loss']
avg_loss.backward()
optimizer.step()
if scaler:
scaled_avg_loss = scaler.scale(avg_loss)
scaled_avg_loss.backward()
scaler.minimize(optimizer, scaled_avg_loss)
else:
avg_loss.backward()
optimizer.step()
optimizer.clear_grad()
train_batch_cost += time.time() - batch_start
batch_sum += len(images)
train_run_cost += time.time() - train_start
total_samples += len(images)
if not isinstance(lr_scheduler, float):
lr_scheduler.step()
......@@ -268,12 +285,13 @@ def train(config,
logs = train_stats.log()
strs = 'epoch: [{}/{}], iter: {}, {}, reader_cost: {:.5f} s, batch_cost: {:.5f} s, samples: {}, ips: {:.5f}'.format(
epoch, epoch_num, global_step, logs, train_reader_cost /
print_batch_step, train_batch_cost / print_batch_step,
batch_sum, batch_sum / train_batch_cost)
print_batch_step, (train_reader_cost + train_run_cost) /
print_batch_step, total_samples,
total_samples / (train_reader_cost + train_run_cost))
logger.info(strs)
train_batch_cost = 0.0
train_reader_cost = 0.0
batch_sum = 0
train_run_cost = 0.0
total_samples = 0
# eval
if global_step > start_eval_step and \
(global_step - start_eval_step) % eval_batch_step == 0 and dist.get_rank() == 0:
......@@ -326,7 +344,7 @@ def train(config,
global_step)
global_step += 1
optimizer.clear_grad()
batch_start = time.time()
reader_start = time.time()
if dist.get_rank() == 0:
save_model(
model,
......@@ -367,7 +385,11 @@ def eval(model,
with paddle.no_grad():
total_frame = 0.0
total_time = 0.0
pbar = tqdm(total=len(valid_dataloader), desc='eval model:')
pbar = tqdm(
total=len(valid_dataloader),
desc='eval model:',
position=0,
leave=True)
max_iter = len(valid_dataloader) - 1 if platform.system(
) == "Windows" else len(valid_dataloader)
for idx, batch in enumerate(valid_dataloader):
......@@ -436,8 +458,6 @@ def get_center(model, eval_dataloader, post_process_class):
batch = [item.numpy() for item in batch]
# Obtain usable results from post-processing methods
total_time += time.time() - start
# Evaluate the results of the current batch
post_result = post_process_class(preds, batch[1])
#update char_center
......@@ -480,11 +500,6 @@ def preprocess(is_train=False):
'CLS', 'PGNet', 'Distillation', 'NRTR', 'TableAttn', 'SAR', 'PSE',
'SEED'
]
windows_not_support_list = ['PSE']
if platform.system() == "Windows" and alg in windows_not_support_list:
logger.warning('{} is not support in Windows now'.format(
windows_not_support_list))
sys.exit()
device = 'gpu:{}'.format(dist.ParallelEnv().dev_id) if use_gpu else 'cpu'
device = paddle.set_device(device)
......
tools/train.py
View file @ 26c16324
......@@ -102,10 +102,27 @@ def main(config, device, logger, vdl_writer):
if valid_dataloader is not None:
logger.info('valid dataloader has {} iters'.format(
len(valid_dataloader)))
use_amp = config["Global"].get("use_amp", False)
if use_amp:
AMP_RELATED_FLAGS_SETTING = {
'FLAGS_cudnn_batchnorm_spatial_persistent': 1,
'FLAGS_max_inplace_grad_add': 8,
}
paddle.fluid.set_flags(AMP_RELATED_FLAGS_SETTING)
scale_loss = config["Global"].get("scale_loss", 1.0)
use_dynamic_loss_scaling = config["Global"].get(
"use_dynamic_loss_scaling", False)
scaler = paddle.amp.GradScaler(
init_loss_scaling=scale_loss,
use_dynamic_loss_scaling=use_dynamic_loss_scaling)
else:
scaler = None
# start train
program.train(config, train_dataloader, valid_dataloader, device, model,
loss_class, optimizer, lr_scheduler, post_process_class,
eval_class, pre_best_model_dict, logger, vdl_writer)
eval_class, pre_best_model_dict, logger, vdl_writer, scaler)
def test_reader(config, device, logger):
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment