cant load deep learning model

[helplessness]

[Wapiti] Loading model: "/data1/tf_data/pro/project/citation/grobid-master/grobid-home/models/reference-segmenter/model.wapiti" ERROR [2023-11-28 09:57:51,378] org.grobid.core.jni.JEPThreadPool: JEP initialization failed ! jep.JepException: <class 'ImportError'>: Traceback (most recent call last): ! File "/home/phoenix/anaconda3/envs/grobid/lib/python3.9/site-packages/tensorflow/python/pywrap_tensorflow.py", line 62, in <module> ! from tensorflow.python._pywrap_tensorflow_internal import * ! ImportError: /home/phoenix/anaconda3/envs/grobid/lib/python3.9/site-packages/tensorflow/python/_pywrap_tensorflow_internal.so: undefined symbol: PyThread_tss_alloc ! ! ! Failed to load the native TensorFlow runtime. ! See https://www.tensorflow.org/install/errors for some common causes and solutions. ! If you need help, create an issue at https://github.com/tensorflow/tensorflow/issues and include the entire stack trace above this error message. ! at /home/phoenix/anaconda3/envs/grobid/lib/python3.9/site-packages/tensorflow/python/pywrap_tensorflow.<module>(pywrap_tensorflow.py:78) ! at /home/phoenix/anaconda3/envs/grobid/lib/python3.9/site-packages/tensorflow/python/__init__.<module>(__init__.py:36) ! at /home/phoenix/anaconda3/envs/grobid/lib/python3.9/site-packages/tensorflow/__init__.<module>(__init__.py:37) ! at /data1/tf_data/pro/project/citation/delft-master/delft/utilities/Embeddings.<module>(Embeddings.py:17) ! at <string>.<module>(<string>:1) ! at jep.Jep.eval(Native Method) ! at jep.Jep.eval(Jep.java:312) ! at org.grobid.core.jni.JEPThreadPool.initializeJepInstance(JEPThreadPool.java:89) ! at org.grobid.core.jni.JEPThreadPool.createJEPInstance(JEPThreadPool.java:113) ! at org.grobid.core.jni.JEPThreadPool.getJEPInstance(JEPThreadPool.java:151) ! at org.grobid.core.jni.DeLFTModel$InitModel.run(DeLFTModel.java:53) ! at java.base/java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:515) ! at java.base/java.util.concurrent.FutureTask.run(FutureTask.java:264) ! at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128) ! at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628) ! at java.base/java.lang.Thread.run(Thread.java:829) Model path: /data1/tf_data/pro/project/citation/grobid-master/grobid-home/models/reference-segmenter/model.wapiti [Wapiti] Loading model: "/data1/tf_data/pro/project/citation/grobid-master/grobid-home/models/figure/model.wapiti" Model path: /data1/tf_data/pro/project/citation/grobid-master/grobid-home/models/figure/model.wapiti [Wapiti] Loading model: "/data1/tf_data/pro/project/citation/grobid-master/grobid-home/models/table/model.wapiti" Model path: /data1/tf_data/pro/project/citation/grobid-master/grobid-home/models/table/model.wapiti [Wapiti] Loading model: "/data1/tf_data/pro/project/citation/grobid-master/grobid-home/models/funding-acknowledgement/model.wapiti" Model path: /data1/tf_data/pro/project/citation/grobid-master/grobid-home/models/funding-acknowledgement/model.wapiti <===========--> 91% EXECUTING [2m 28s]
python : 3.9

I have received the above error. Anyone know how to solve it？

[kermitt2]

Hi @helplessness !

Did you follow the instructions https://grobid.readthedocs.io/en/latest/Deep-Learning-models/ for installing the DL environment?

[helplessness]

yes
I follow it

Hi @helplessness !

Did you follow the instructions https://grobid.readthedocs.io/en/latest/Deep-Learning-models/ for installing the DL environment?

[helplessness]

Hi @helplessness !

Did you follow the instructions https://grobid.readthedocs.io/en/latest/Deep-Learning-models/ for installing the DL environment?

I wonder if it's a problem with a library version。but I follow it

[kermitt2]

Ok then next questions, can you describe your environment: OS, JDK version, Python & CUDA version, which virtualenv?

Did you use the script under ./grobid-home/scripts/install_jep_lib.sh to install JEP ? On my side I am still using JEP 4.0.2 as in the script, and I have not tried with more recent versions.

[helplessness]

Ok then next questions, can you describe your environment: OS, JDK version, Python & CUDA version, which virtualenv?

Did you use the script under ./grobid-home/scripts/install_jep_lib.sh to install JEP ? On my side I am still using JEP 4.0.2 as in the script, and I have not tried with more recent versions.

linux 22.04
jdk 11
python 3.9 with conda
cuda 12.2
and I use grobid-home/scripts/install_jep_lib.sh to install JEP
I suspect it's the tensorflow version, but I don't know much about it.

[helplessness]

Hi, @kermitt2 ，thank you for your reply. I solved the problem of using deep learning models using Docker. But I think my problem still exists.I want to use deep learning models because I feel that I cannot accurately identify images and attachments. But I think this problem still exists like this picture. Do I have any good ways to solve this problem?

[lfoppiano]

You can use the deep learning models using the docker image: grobid/grobid:.... see: https://grobid.readthedocs.io/en/latest/Grobid-docker/#deep-learning-and-crf-image

You can also tune more in detail what DL model to use by providing your configuration mounted as a volume when you run the container: https://grobid.readthedocs.io/en/latest/Grobid-docker/#configure-using-the-yaml-config-file