-
Notifications
You must be signed in to change notification settings - Fork 121
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.
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
Openvino #228
base: master
Are you sure you want to change the base?
Conversation
OpenVINO backend for Intel Architectures
@dkurt first hurdle - have you seen this? /tmp/test$ python3.8 -m venv venv3.8
/tmp/test$ source venv3.8/bin/activate
(venv3.8) /tmp/test$ pip install --upgrade pip
Collecting pip
Downloading https://files.pythonhosted.org/packages/a4/6d/6463d49a933f547439d6b5b98b46af8742cc03ae83543e4d7688c2420f8b/pip-21.3.1-py3-none-any.whl (1.7MB)
|████████████████████████████████| 1.7MB 12.6MB/s
Installing collected packages: pip
Found existing installation: pip 19.2.3
Uninstalling pip-19.2.3:
Successfully uninstalled pip-19.2.3
Successfully installed pip-21.3.1
(venv3.8) /tmp/test$ pip install openvino==2021.4.2
Collecting openvino==2021.4.2
Downloading openvino-2021.4.2-3976-cp38-cp38-manylinux2014_x86_64.whl (28.9 MB)
|████████████████████████████████| 28.9 MB 15.8 MB/s
Collecting numpy<1.20,>=1.16.6
Downloading numpy-1.19.5-cp38-cp38-manylinux2010_x86_64.whl (14.9 MB)
|████████████████████████████████| 14.9 MB 47.4 MB/s
Installing collected packages: numpy, openvino
Successfully installed numpy-1.19.5 openvino-2021.4.2
(venv3.8) /tmp/test$ python3.8 -c "from openvino.inference_engine import IECore, StatusCode"
Traceback (most recent call last):
File "<string>", line 1, in <module>
File "/tmp/test/venv3.8/lib/python3.8/site-packages/openvino/inference_engine/__init__.py", line 30, in <module>
from .ie_api import *
ImportError: libpython3.8.so.1.0: cannot open shared object file: No such file or directory
(venv3.8) /tmp/test$ ldd .../ie_api.cpython-38-x86_64-linux-gnu.so
linux-vdso.so.1 => (0x00007ffce0dd8000)
libpython3.8.so.1.0 => not found
libinference_engine.so => /tmp/test/venv3.8/lib/python3.8/site-packages/openvino/inference_engine/./../../openvino/libs/libinference_engine.so (0x00007fd035826000)
libstdc++.so.6 => /usr/lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007fd0354a4000)
libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007fd03519b000)
libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007fd034f85000)
libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007fd034bbb000)
libtbb.so.2 => /tmp/test/venv3.8/lib/python3.8/site-packages/openvino/inference_engine/./../../openvino/libs/libtbb.so.2 (0x00007fd03494d000)
libtbbmalloc.so.2 => /tmp/test/venv3.8/lib/python3.8/site-packages/openvino/inference_engine/./../../openvino/libs/libtbbmalloc.so.2 (0x00007fd0346f1000)
libdl.so.2 => /lib/x86_64-linux-gnu/libdl.so.2 (0x00007fd0344ed000)
libinference_engine_transformations.so => /tmp/test/venv3.8/lib/python3.8/site-packages/openvino/inference_engine/./../../openvino/libs/libinference_engine_transformations.so (0x00007fd033f20000)
libngraph.so => /tmp/test/venv3.8/lib/python3.8/site-packages/openvino/inference_engine/./../../openvino/libs/libngraph.so (0x00007fd033562000)
libpthread.so.0 => /lib/x86_64-linux-gnu/libpthread.so.0 (0x00007fd033345000)
/lib64/ld-linux-x86-64.so.2 (0x00007fd035f3c000)
librt.so.1 => /lib/x86_64-linux-gnu/librt.so.1 (0x00007fd03313d000)
|
@iiSeymour, do you use default Python or Anaconda env? For default, |
Yes, that has resolved it @dkurt. Basecalling results are now correct for the CRF models but I'm finding the performance much slower than expected even with the smaller models (note: I haven't done any profiling yet).
|
Most of execution time (up to 95%) take LSTM layers. I know that OpenVINO team works on improvements. I've tried to think up some workaround but no progress. |
Thanks @dkurt let me know if there is any progress w.r.t LSTM performance. |
Use OpenVINO 2022
Revert "Use OpenVINO 2022"
Batching with OpenVINO
Use OpenVINO 2022 with new API
@iiSeymour, can you please add an empty file BTW, is that possible to merge this PR? So users of pip package could use OpenVINO as well. |
No description provided.