I tried and got the same outcome.
Hm, I've tried it just now and got this: https://i.imgur.com/01MIEWD.png
Note: Wait more time to repost.
Thanks, I'll remember that.
I saw the previous post gaining no traction, and personally had fun testing the service with C / Audio code before, so I wanted it to stay afloat a bit longer, at least.
It's hard to believe that the system can detect that it's the "mean squared error".
I won't be surprised ML model detected it, statistics-wise, associating "Tensorflow" and "mean" with "mean squared error". Happened when I fed it CLIP-related code.
Don't you want to use the average instead of "0.5"?
It was tested on the real-life code, looks like a Colab I have previously recommended.
https://colab.research.google.com/drive/1WWo2L7ZMdslbZyOg-pq...