deepinsight
Not sure if insightface is using CPU or GPU?
Hi, all:
I can run my test.py (using insightface) for real-time face swapping. However:
- the swapping for camera streaming is extremly slow. It cannot even see my mouth open and close
- I ONLY installed
onnxruntime-gpu - But when I run my code, I got
Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}}
as follows:
➜ python test.py
~/.local/lib/python3.10/site-packages/numpy/core/getlimits.py:518: UserWarning: The value of the smallest subnormal for <class 'numpy.float64'> type is zero.
setattr(self, word, getattr(machar, word).flat[0])
~/.local/lib/python3.10/site-packages/numpy/core/getlimits.py:89: UserWarning: The value of the smallest subnormal for <class 'numpy.float64'> type is zero.
return self._float_to_str(self.smallest_subnormal)
~/.local/lib/python3.10/site-packages/numpy/core/getlimits.py:518: UserWarning: The value of the smallest subnormal for <class 'numpy.float32'> type is zero.
setattr(self, word, getattr(machar, word).flat[0])
~/.local/lib/python3.10/site-packages/numpy/core/getlimits.py:89: UserWarning: The value of the smallest subnormal for <class 'numpy.float32'> type is zero.
return self._float_to_str(self.smallest_subnormal)
Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}}
find model: ~/.insightface/models/buffalo_l/1k3d68.onnx landmark_3d_68 ['None', 3, 192, 192] 0.0 1.0
Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}}
find model: ~/.insightface/models/buffalo_l/2d106det.onnx landmark_2d_106 ['None', 3, 192, 192] 0.0 1.0
Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}}
find model: ~/.insightface/models/buffalo_l/det_10g.onnx detection [1, 3, '?', '?'] 127.5 128.0
Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}}
find model: ~/.insightface/models/buffalo_l/genderage.onnx genderage ['None', 3, 96, 96] 0.0 1.0
Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}}
find model: ~/.insightface/models/buffalo_l/w600k_r50.onnx recognition ['None', 3, 112, 112] 127.5 127.5
set det-size: (640, 640)
Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}}
inswapper-shape: [1, 3, 128, 128]
~/.local/lib/python3.10/site-packages/insightface/utils/transform.py:68: FutureWarning: `rcond` parameter will change to the default of machine precision times ``max(M, N)`` where M and N are the input matrix dimensions.
To use the future default and silence this warning we advise to pass `rcond=None`, to keep using the old, explicitly pass `rcond=-1`.
P = np.linalg.lstsq(X_homo, Y)[0].T # Affine matrix. 3 x 4
[ WARN:[email protected]] global cap_gstreamer.cpp:1728 open OpenCV | GStreamer warning: Cannot query video position: status=0, value=-1, duration=-1
QFactoryLoader::QFactoryLoader() checking directory path "/usr/lib/x86_64-linux-gnu/qt6/plugins/platforms" ...
QFactoryLoader::QFactoryLoader() looking at "/usr/lib/x86_64-linux-gnu/qt6/plugins/platforms/libqeglfs.so"
Found metadata in lib /usr/lib/x86_64-linux-gnu/qt6/plugins/platforms/libqeglfs.so, metadata=
{
"IID": "org.qt-project.Qt.QPA.QPlatformIntegrationFactoryInterface.5.3",
"MetaData": {
"Keys": [
"eglfs"
]
},
"archreq": 0,
"className": "QEglFSIntegrationPlugin",
"debug": false,
"version": 393728
}
......
I think the Applied providers should be one of TensorrtExecutionProvider or CUDAExecutionProvider, for I got so many providers.
['TensorrtExecutionProvider', 'CUDAExecutionProvider', 'MIGraphXExecutionProvider', 'ROCMExecutionProvider', 'OpenVINOExecutionProvider', 'DnnlExecutionProvider', 'TvmExecutionProvider', 'VitisAIExecutionProvider', 'QNNExecutionProvider', 'NnapiExecutionProvider', 'JsExecutionProvider', 'CoreMLExecutionProvider', 'ArmNNExecutionProvider', 'ACLExecutionProvider', 'DmlExecutionProvider', 'RknpuExecutionProvider', 'XnnpackExecutionProvider', 'CANNExecutionProvider', 'AzureExecutionProvider', 'CPUExecutionProvider']
Why the actual provider I obtained is CPUExecutionProvider ?? Is this the reason that makes my test.py very slow???
Cheers
import insightface
insightface.app.FaceAnalysis(name='buffalo_l', providers=['CUDAExecutionProvider'])
@phineas-pta
Thank you ...
- But, how can I test if this particilar provider is applied? After adding
providers=['CUDAExecutionProvider']), I still got the following:
Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}}
find model: /home/lvision/.insightface/models/buffalo_l/1k3d68.onnx landmark_3d_68 ['None', 3, 192, 192] 0.0 1.0
Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}}
find model: /home/lvision/.insightface/models/buffalo_l/2d106det.onnx landmark_2d_106 ['None', 3, 192, 192] 0.0 1.0
Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}}
find model: /home/lvision/.insightface/models/buffalo_l/det_10g.onnx detection [1, 3, '?', '?'] 127.5 128.0
Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}}
find model: /home/lvision/.insightface/models/buffalo_l/genderage.onnx genderage ['None', 3, 96, 96] 0.0 1.0
Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}}
find model: /home/lvision/.insightface/models/buffalo_l/w600k_r50.onnx recognition ['None', 3, 112, 112] 127.5 127.5
set det-size: (640, 640)
Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}}
How to test if CPUExecutionProvider is now working, or if CUDAExecutionProvider is working?
- BTW, the speed of my program is still NOT fast enough. It still cannot catch up my mouth's speed even with a slower speaking speed. For now,
- frame size: 640x480
- the following time differences are in seconds.
...
Time difference is: 0.752118
Time difference is: 0.833245
Time difference is: 0.748658
Time difference is: 0.6013
Time difference is: 0.720522
Time difference is: 0.658189
Time difference is: 0.608403
Time difference is: 0.763706
Time difference is: 0.792851
Time difference is: 0.84137
...
well it means u didnt set up env correctly: need cuda 11.8 (not 12 yet), remove anything onnxruntime and reinstall onnxruntime-gpu, recent nvidia gpu
official wheels onnxruntime-gpu dont support cuda 12 yet, if u want u have to build from source
onnxruntime-gpu now suport CUDA 12, just uninstall and reinstall the new one:
Must install with extra-index-url to work with CUDA 12
pip uninstall onnxruntime-gpu onnxruntime
pip install onnxruntime-gpu --extra-index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-12/pypi/simple/
I can't ran model on GPU when I ran the below code I get GPU output import onnxruntime as ort print( ort.get_device() )
but when I ran the below code just cpu is available.
Can you help me with that?
model = insightface.app.FaceAnalysis(providers=['CUDAExecutionProvider']) model.prepare(ctx_id = 1) # use CPU and set NMS threshold
Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}}
@phineas-pta
@phineas-pta
@ phineas-pta
Thank you ...
谢谢..。
- But, how can I test if this particilar provider is applied? After adding 但是，如何测试是否应用了这个特定的提供程序
providers=['CUDAExecutionProvider']), I still got the following:，我仍然得到以下:Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}} find model: /home/lvision/.insightface/models/buffalo_l/1k3d68.onnx landmark_3d_68 ['None', 3, 192, 192] 0.0 1.0 Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}} find model: /home/lvision/.insightface/models/buffalo_l/2d106det.onnx landmark_2d_106 ['None', 3, 192, 192] 0.0 1.0 Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}} find model: /home/lvision/.insightface/models/buffalo_l/det_10g.onnx detection [1, 3, '?', '?'] 127.5 128.0 Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}} find model: /home/lvision/.insightface/models/buffalo_l/genderage.onnx genderage ['None', 3, 96, 96] 0.0 1.0 Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}} find model: /home/lvision/.insightface/models/buffalo_l/w600k_r50.onnx recognition ['None', 3, 112, 112] 127.5 127.5 set det-size: (640, 640) Applied providers: ['CPUExecutionProvider'], with options: {'CPUExecutionProvider': {}}How to test if CPUExecutionProvider is now working, or if CUDAExecutionProvider is working?
如何测试 CPUExectionProvider 是否正在工作，或 CUDAExectionProvider 是否正在工作？
- BTW, the speed of my program is still NOT fast enough. It still cannot catch up my mouth's speed even with a slower speaking speed. For now,
- 顺便说一下，我的程序的速度还是不够快。它仍然不能赶上我的嘴的速度，即使说话速度较慢。现在,
- frame size: 640x480
- 画面大小: 640x480
- the following time differences are in seconds.
- 以下时差单位为秒。
... Time difference is: 0.752118 Time difference is: 0.833245 Time difference is: 0.748658 Time difference is: 0.6013 Time difference is: 0.720522 Time difference is: 0.658189 Time difference is: 0.608403 Time difference is: 0.763706 Time difference is: 0.792851 Time difference is: 0.84137 ...
My question is the same as yours. My current environment is cudnn=8.2.4 cuda=11.4 onnxruntime gpu=1.12
