Skip to content

Improved finetuned transformers accuracy.#717

Merged
jklaise merged 4 commits intoSeldonIO:masterfrom
RobertSamoilescu:ig_transformers_finetune
Jul 20, 2022
Merged

Improved finetuned transformers accuracy.#717
jklaise merged 4 commits intoSeldonIO:masterfrom
RobertSamoilescu:ig_transformers_finetune

Conversation

@RobertSamoilescu
Copy link
Contributor

This PR improves the accuracy of the Integrated Gradients example finetuned transformes models from 87% to > 91%. It also contains minor fixes.

@review-notebook-app
Copy link

Check out this pull request on  ReviewNB

See visual diffs & provide feedback on Jupyter Notebooks.


Powered by ReviewNB

@RobertSamoilescu RobertSamoilescu force-pushed the ig_transformers_finetune branch from 88bd751 to 9522b2c Compare July 17, 2022 20:22
@codecov
Copy link

codecov bot commented Jul 17, 2022

Codecov Report

Merging #717 (ae7a3d1) into master (e790f52) will not change coverage.
The diff coverage is n/a.

Impacted file tree graph

@@           Coverage Diff           @@
##           master     #717   +/-   ##
=======================================
  Coverage   81.06%   81.06%           
=======================================
  Files         105      105           
  Lines       11794    11794           
=======================================
  Hits         9561     9561           
  Misses       2233     2233           

@RobertSamoilescu RobertSamoilescu force-pushed the ig_transformers_finetune branch from 8c805ea to 616a024 Compare July 17, 2022 22:40
@@ -43,18 +43,22 @@
"import numpy as np\n",
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Line #80.        # see transformers issue #14404: https://github.com/huggingface/transformers/issues/14404

It seem the issue says this has been resolved now? Although probably would need a minimum version bump (not sure which one) in alibi for transformers to avoid people using numpy arrays with older versions of transformers .


Reply via ReviewNB

@@ -43,18 +43,22 @@
"import numpy as np\n",
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Line #1.    # choose whether to use the BERT pr distilBERT model by selecting the appropriate name

pr -> or


Reply via ReviewNB

@jklaise
Copy link
Contributor

jklaise commented Jul 19, 2022

Nice one! Left 2 comments. Good idea to replace the manual classifier with the auto-classifier from the transformers library.

@jklaise jklaise merged commit b0df37c into SeldonIO:master Jul 20, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants