I have installed RASA-1.10.16 on Jetson nano with JetPack-4.3, python-3.6.9 and Tensorflow-2.1.0.
However, when I tried to train the model of rasa-demo chatbot, Sara, I met the following problem:
Module for policy ‘TEDPolicy’ could not be loaded.
Is this a library-depdency problem?
What version of RASA should I use to avoid this problem?
1 Like
Tanja
October 29, 2020, 1:47pm
2
TEDPolicy should be known in Rasa 1.10.16. Do you see anything else? Any other error that might give a hint where exactly this comes from? Does it work if you use some other policies?
Hi @Tanja ,
Yes, I met 2 problems actually.
ImportError: /usr/lib/aarch64-linux-gnu/libgomp.so.1: cannot allocate memory in static TLS block
Module for policy ‘TEDPolicy’ could not be loaded
After searching on Google, I found the answer.
opened 02:44PM - 06 Jun 20 UTC
closed 05:16AM - 17 May 22 UTC
Hi,
Importing kerastuner fails on my my Jetson TX2.
Here is the complete tr… ace:
Python 3.6.9 (default, Apr 18 2020, 01:56:04)
[GCC 8.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import kerastuner
2020-06-06 15:41:05.390226: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libcudart.so.10.2
2020-06-06 15:41:07.658713: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libnvinfer.so.7
2020-06-06 15:41:07.660570: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully opened dynamic library libnvinfer_plugin.so.7
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/home/oliver/Python/keras-tuner/kerastuner/__init__.py", line 20, in <module>
from . import oracles
File "/home/oliver/Python/keras-tuner/kerastuner/oracles/__init__.py", line 16, in <module>
from ..tuners.bayesian import BayesianOptimizationOracle as BayesianOptimization
File "/home/oliver/Python/keras-tuner/kerastuner/tuners/__init__.py", line 17, in <module>
from .bayesian import BayesianOptimization
File "/home/oliver/Python/keras-tuner/kerastuner/tuners/bayesian.py", line 5, in <module>
from sklearn import exceptions
File "/home/oliver/.local/lib/python3.6/site-packages/sklearn/__init__.py", line 81, in <module>
from .utils._show_versions import show_versions
File "/home/oliver/.local/lib/python3.6/site-packages/sklearn/utils/_show_versions.py", line 12, in <module>
from ._openmp_helpers import _openmp_parallelism_enabled
ImportError: /usr/lib/aarch64-linux-gnu/libgomp.so.1: cannot allocate memory in static TLS block
>>>
Oliver
The problem is gone after I execute “export LD_PRELOAD=/usr/lib/aarch64-linux-gnu/libgomp.so.1”.
1 Like
Hey @Joseph.Huang , any chance you have written down the needed steps to install RASA v1.10.16 on the jetson nano?
Edit:
I was able to install rasa v1.10.16 on the jetson nano.
Using the pre-compiled tensorflow 2.1 from nvidia. Official TensorFlow for Jetson Nano! - Jetson Nano - NVIDIA Developer Forums
For tensorflow-addons i build v0.9.0 from source using only cpu ops. GitHub - tensorflow/addons at v0.9.0
Best regards,
Alex