The TensorFlow library was compiled to use AVX instructions, but these aren't available on your machine.

[alishhde]

Hey there!

I have installed the dependencies (poetry and yarn). I have also run the following commands without any error:

• pip install gdown
• make download_demo

However, when I reach the third command,

• make CFG_PATH=/config/development/clinc/conf.json launch
  I get the error "The TensorFlow library was compiled to use AVX instructions, but these aren't available on your machine."

Here's the exact output of this last command:

---

the attribute version is obsolete, it will be ignored, please remove it to avoid potential confusion
[+] Running 2/0
✔ Container azimuth-web-1 Created 0.0s
✔ Container azimuth-backend-1 Created 0.0s
Attaching to backend-1, web-1
web-1 | [HPM] Proxy created: / -> http://backend:8091
web-1 | [HPM] Proxy rewrite rule created: "^/api/[^/]+/" ~> "/"
web-1 |
web-1 | API server listening on port 8080
web-1 | Server PID: 1
web-1 | (following appears after a couple of seconds)
backend-1 | The TensorFlow library was compiled to use AVX instructions, but these aren't available on your machine.
backend-1 | Fatal Python error: Aborted
backend-1 |
backend-1 | Current thread 0x00007fffff1b6740 (most recent call first):
backend-1 | File "", line 228 in _call_with_frames_removed
...
backend-1 | ...
backend-1 | Aborted
backend-1 exited with code 134

---

I am on Macbook M3 Pro Sequoia 15.3

I would appreciate it if you could guide me how I can tackle the problem of CPU/GPU on this system.

Regards,
Ali

[Dref360]

Hello,

The "The TensorFlow library was compiled to use AVX instructions, but these aren't available on your machine." seems to be caused by MacOS Silicon. Looking into it

I'll try to launch the app on my new M4 and I'll report back

[Dref360]

Could you check that you allow for enough memory/cpu in Docker -> Settings -> Resources

12GB mem should be enough and 8 cores

[Dref360]

I got the same error on my M4, I think it's because the Docker image is not built for arm64.
I could add the flag, but I can't rebuild the Docker image because Python 3.9 is now deprecated so that will have to be done over the weekend. In the mean time, you can run Azimuth on a Linux machine.