I am trying to use trax as taught in the course. But it keep on popping error. I also use help from community and ChatGPT but couldn’t fixed it please look my code.
There have been some other threads about that and it looks like it’s a problem with no easy solution.
@paulinpaloalto and @Zonunmawia_Zadeng that thread from last year focused on challenges specifically related to installing on the Apple M1 silicon and running with the Apple-specific version of TensorFlow. I haven’t tried since then, but my belief is that trax still only runs natively on Ubuntu.
I can’t tell from information provided above if that is the issue here, though from the fact that there seems to be a …/site_packages/trax/tf_numpy/… directory perhaps not?
If the install is indeed on Ubuntu, it should in theory work, so it could just be version incompatibility. Python, TensorFlow and trax all need to be at compatible levels in your environment, and those versions also have to be compatible with the code from the exercise. When I took the class, I think trax was still at a 0.x release level, so wouldn’t surprise me if there had been some refactoring, renaming, etc such that the latest default install by pip breaks the course code.
So I could not follow the course ?
I’m not familiar with that course, but in all the other courses here they give you the ability to run everything on the course website. Is that not possible in NLP C3?
My understanding is that it will run like a champ in Colab or locally on Ubuntu. Outside of that, it’s a slog to try and build your own from source. I tried for a bit on the Mac but lost interest pretty quickly.
NOTE: before investing too much time building and learning trax, you might want to take a look at this…
If I read that correctly, there hasn’t been an update in 2 1/2 years.