06 - Canada

[erwangranger]

This is the thread where you can paste your notes based on the scenario presented.

[thom-at-redhat]

• Due to time constraints, brain shuts down and blindly click cells

• Add Big warning in SKIP ME if user runs this notebook
  Warning: metadata.finalizers: "inferenceservice.finalizers": prefer a domain-qualified finalizer name to avoid accidental conflicts with other finalizer writers

• Ran out of time, was on step 03

/opt/app-root/lib64/python3.9/site-packages/transformers/models/vit/feature_extraction_vit.py:28: FutureWarning: The class ViTFeatureExtractor is deprecated and will be removed in version 5 of Transformers. Please use ViTImageProcessor instead.
warnings.warn(
/opt/app-root/lib64/python3.9/site-packages/transformers/models/vit/modeling_vit.py:172: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
if num_channels != self.num_channels:
/opt/app-root/lib64/python3.9/site-packages/transformers/models/vit/modeling_vit.py:178: TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
if height != self.image_size[0] or width != self.image_size[1]:

• Prediction <Response [200]>
  Do we expect all users understand this is a HTTP status code?

• Error starting Kernel / Kernel died unexpectedly

[KiriGould]

I'm a bit confused on the problem solving part, since I just clicked through and expected an error that didn't really come. ^Agreed with Thom here that maybe some visual display on the input and output of the model would be helpful.

[erwangranger]

Thanks both. Kernel was not supposed to die out.

the gist of the problem here is not solve errors (there should not be any), it's to try to find a way to speed up the response time of the model. (see hints for suggestion).

the 7-8 notebooks I provide should be enough to run everything and measure response time, and then, adjusting various knobs in the RHOAI dashboard interface should make (or not) a difference in the response time.

but this is great feedback, because I can see how fresh eyes look at this and how the important info is buried in details.
gimme a week and I'll re-adjust that.

Thanks both.