FATAL ERROR mkl_vml_avx2.dll

311
3
03-12-2020 04:45 PM
Highlighted
New Contributor

I'm new in arcgis pro, I want to train a model for deep learning with python on jupyter notebooks, when running epoch 7, I have problems about with models. The model will restart automatically after epoch 7 out of 20, all progress running is still 45%, when I check in cmd, "MKL FATAL ERROR: Cannot load mkl_vml_avx2.dll", what should I do? any suggestion?

Reply
0 Kudos
3 Replies
Highlighted
New Contributor II

I also have this issue - I was able to solve it by locating mkl_vml_avx2.dll from another python environment and copying it to my current python environment, so for me it was copy mkl_vml_avx2.dll from

C:\Users\[username]\Anaconda3\envs\fastai_v1\Library\bin
to here...
C:\ArcGIS\Pro_\bin\Python\envs\deeplearn\Library\bin

I have no idea why it wasn't there and why it had to be done manually.

FYI.....The Jupyter notebook log that this fixes is as below....

INTEL MKL ERROR: The specified module could not be found. mkl_vml_avx2.dll.
Intel MKL FATAL ERROR: cannot load mkl_vml_avx2.dll.
[I 09:52:22.077 NotebookApp] KernelRestarter: restarting kernel (1/5), keep random ports

I also think this error has the following behavior...... when using 'Classify Pixel for Deep Learning' and it hits this error, I believe it crashes ArcGIS Pro and completely shuts the application down without any notification or warnings.

Highlighted
New Contributor

I have a couple follow-up questions if you happen to still have access to this information:

  • I'm assuming you are using a machine with an AMD CPU?
    • You can find this by searching for the `System Information` app on your machine, it is listed under "Processor"
  • For your `fastai_v1` environment where did you install `fastai` from (which channel) and what version is it? 
    • You can find this information by activating the environment from a terminal and running `conda list`
Reply
0 Kudos
Highlighted
New Contributor II

Hi Brian

Yes a AMD Ryzen 9 3950X 16-Core Processor, 3493 Mhz, 16 Core(s), 32 Logical Processor(s)

I believe the install was:

conda install -c esri -c fastai -c pytorch arcgis pillow scikit-image fastai=1.0.54 pytorch=1.1.0

and the 'conda list' is

fastai 1.0.54 py36_1 esri

Regards

Tim

Reply
0 Kudos