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

Jump to bottom

fix detection infer time statistics bug #1751

Merged
merged 1 commit into from
May 18, 2023
Merged

fix detection infer time statistics bug #1751

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
12 changes: 12 additions & 0 deletions example/auto_compression/detection/configs/rtdetr_reader.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,6 +12,18 @@ TrainDataset:
anno_path: annotations/instances_val2017.json
dataset_dir: dataset/coco/

EvalDataset:
!COCODataSet
image_dir: val2017
anno_path: annotations/instances_val2017.json
dataset_dir: dataset/coco/

TestDataset:
!COCODataSet
image_dir: val2017
anno_path: annotations/instances_val2017.json
dataset_dir: dataset/coco/

worker_num: 0

# preprocess reader in test
Expand Down
68 changes: 10 additions & 58 deletions example/auto_compression/detection/paddle_inference_eval.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -64,7 +64,8 @@ def argsparser():
"--device",
type=str,
default="GPU",
help="Choose the device you want to run, it can be: CPU/GPU/XPU, default is GPU",
help=
"Choose the device you want to run, it can be: CPU/GPU/XPU, default is GPU",
)
parser.add_argument(
"--use_dynamic_shape",
Expand Down Expand Up @@ -270,8 +271,8 @@ def load_predictor(
dynamic_shape_file = os.path.join(FLAGS.model_path,
"dynamic_shape.txt")
if os.path.exists(dynamic_shape_file):
config.enable_tuned_tensorrt_dynamic_shape(dynamic_shape_file,
True)
config.enable_tuned_tensorrt_dynamic_shape(
dynamic_shape_file, True)
print("trt set dynamic shape done!")
else:
config.collect_shape_range_info(dynamic_shape_file)
Expand All @@ -284,48 +285,6 @@ def load_predictor(
return predictor, rerun_flag


def get_current_memory_mb():
"""
It is used to Obtain the memory usage of the CPU and GPU during the running of the program.
And this function Current program is time-consuming.
"""
try:
pkg.require('pynvml')
except:
from pip._internal import main
main(['install', 'pynvml'])
try:
pkg.require('psutil')
except:
from pip._internal import main
main(['install', 'psutil'])
try:
pkg.require('GPUtil')
except:
from pip._internal import main
main(['install', 'GPUtil'])
import pynvml
import psutil
import GPUtil

gpu_id = int(os.environ.get("CUDA_VISIBLE_DEVICES", 0))

pid = os.getpid()
p = psutil.Process(pid)
info = p.memory_full_info()
cpu_mem = info.uss / 1024.0 / 1024.0
gpu_mem = 0
gpu_percent = 0
gpus = GPUtil.getGPUs()
if gpu_id is not None and len(gpus) > 0:
gpu_percent = gpus[gpu_id].load
pynvml.nvmlInit()
handle = pynvml.nvmlDeviceGetHandleByIndex(0)
meminfo = pynvml.nvmlDeviceGetMemoryInfo(handle)
gpu_mem = meminfo.used / 1024.0 / 1024.0
return round(cpu_mem, 4), round(gpu_mem, 4)


def predict_image(predictor,
image_file,
image_shape=[640, 640],
Expand Down Expand Up @@ -353,6 +312,7 @@ def predict_image(predictor,
predict_time = 0.0
time_min = float("inf")
time_max = float("-inf")
paddle.device.cuda.synchronize()
for i in range(repeats):
start_time = time.time()
predictor.run()
Expand All @@ -367,13 +327,8 @@ def predict_image(predictor,
time_min = min(time_min, timed)
time_max = max(time_max, timed)
predict_time += timed
cpu_mem, gpu_mem = get_current_memory_mb()
cpu_mems += cpu_mem
gpu_mems += gpu_mem

time_avg = predict_time / repeats
print("[Benchmark]Avg cpu_mem:{} MB, avg gpu_mem: {} MB".format(
cpu_mems / repeats, gpu_mems / repeats))
print("[Benchmark]Inference time(ms): min={}, max={}, avg={}".format(
round(time_min * 1000, 2),
round(time_max * 1000, 1), round(time_avg * 1000, 1)))
Expand Down Expand Up @@ -406,6 +361,7 @@ def eval(predictor, val_loader, metric, rerun_flag=False):
for i, _ in enumerate(input_names):
input_tensor = predictor.get_input_handle(input_names[i])
input_tensor.copy_from_cpu(data_all[input_names[i]])
paddle.device.cuda.synchronize()
start_time = time.time()
predictor.run()
np_boxes = boxes_tensor.copy_to_cpu()
Expand All @@ -418,9 +374,6 @@ def eval(predictor, val_loader, metric, rerun_flag=False):
time_min = min(time_min, timed)
time_max = max(time_max, timed)
predict_time += timed
cpu_mem, gpu_mem = get_current_memory_mb()
cpu_mems += cpu_mem
gpu_mems += gpu_mem
if not FLAGS.include_nms:
postprocess = PPYOLOEPostProcess(
score_threshold=0.3, nms_threshold=0.6)
Expand All @@ -436,8 +389,6 @@ def eval(predictor, val_loader, metric, rerun_flag=False):
map_res = metric.get_results()
metric.reset()
time_avg = predict_time / sample_nums
print("[Benchmark]Avg cpu_mem:{} MB, avg gpu_mem: {} MB".format(
cpu_mems / sample_nums, gpu_mems / sample_nums))
print("[Benchmark]Inference time(ms): min={}, max={}, avg={}".format(
round(time_min * 1000, 2),
round(time_max * 1000, 1), round(time_avg * 1000, 1)))
Expand Down Expand Up @@ -473,9 +424,10 @@ def main():

dataset = reader_cfg["EvalDataset"]
global val_loader
val_loader = create("EvalReader")(reader_cfg["EvalDataset"],
reader_cfg["worker_num"],
return_list=True)
val_loader = create("EvalReader")(
reader_cfg["EvalDataset"],
reader_cfg["worker_num"],
return_list=True)
clsid2catid = {v: k for k, v in dataset.catid2clsid.items()}
anno_file = dataset.get_anno()
metric = COCOMetric(
Expand Down