search

RamyE / SFU_ML

SFU MSE ML Course Labs and Tools
MIT License
0 stars 3 forks source link

Error occurred with lab selection while script is running #12

Closed hpoonian closed 3 years ago

hpoonian commented 3 years ago

Hi, we're getting the error occurred with lab selection message even though we have the auto run setup for the script. We manually ran the script in the pi as well but we still get the same error. There are also two other error messages we get at the same time: Processing has failed or was interrupted, and Resetting the serial state of RPi Failed, please try again or power cycle the RPi if the problem persists.

gui error

gui ssh error

RamyE commented 3 years ago

can you confirm you're connected to the right serial port? (maybe check the port names in the device manager)

hpoonian commented 3 years ago

There's only one serial port for me to select. COM 1 is the only port showing under device manager whether I have the pi connected or not.

RamyE commented 3 years ago

if it shows up when the pi is not connected then it is not the pi's serial port. Did u use the same pi & cable & PC when u did lab 1?

hpoonian commented 3 years ago

Yes, I had the same error messages from this pc in lab 1. I connected the pi to different ports until I got one showing up as working but now I'm getting some error messages that don't really seem consistent.

The first error I got is this buffer dtype error when I wasn't in the 32 bit conda setup I had.

buffer dtype error

After I switched to ml_course_32 I got this error.

failed to load required model

When I try running it again now I get the buffer dtype error happening and it switches randomly between the two error messages.

RamyE commented 3 years ago

Sounds like the buffer dtype is still the issue here, please confirm that you trained the model using a 32-bit architecture. You can try something like this in your script or NB where you save the trained model import platform; print(platform.architecture())

hpoonian commented 3 years ago

We fiexed the issue. For some reason visual studio code wasn't outputting the model as 32 bit from a 32 bit environment but running it through cmd did.