Fixed: #1722 Run out of memory

[Anselmoo]

Update utils.py and test by using break instead of raise

Which issue(s) does this Pull Request fix?

resolves #1722

Details of the Pull Request

According to discussion #1721, theraise will be replaced by break for running out of memory in the case of batch_size==1
.

[codecov]

Codecov Report

Merging #1724 (f477215) into master (c51da2d) will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##            master     #1724   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           51        51           
  Lines         3411      3413    +2     
=========================================
+ Hits          3411      3413    +2     

Impacted Files	Coverage Δ
autokeras/utils/utils.py	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c51da2d...f477215. Read the comment docs.

[haifeng-jin]

Thanks for the fix!

If history is None, it should trigger this error. To your test, does it trigger it?
We can first add a if to check its None, return an empty dict.

Check for empty dict in Oracle.update_trial to return invalid status.

I can make the change if you prefer?

[Anselmoo]

Thanks for the fix!

If history is None, it should trigger this error. To your test, does it trigger it? We can first add a if to check its None, return an empty dict.

Check for empty dict in Oracle.update_trial to return invalid status.

I can make the change if you prefer?

Will be the changes small, then I would prefer if you would do it because I have to first get into the code. Otherwise, I will take later a look.

[haifeng-jin]

@Anselmoo I will make the fix then, it may take some time. Thanks.

[haifeng-jin]

This is blocked by:
keras-team/keras-tuner#704

[auvipy]

I am not sure if using print a right approach here?