Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add distilBERT with batches and test it #4

Open
MastafaF opened this issue Apr 8, 2020 · 3 comments
Open

Add distilBERT with batches and test it #4

MastafaF opened this issue Apr 8, 2020 · 3 comments

Comments

@MastafaF
Copy link
Owner

MastafaF commented Apr 8, 2020

No description provided.

@MastafaF
Copy link
Owner Author

MastafaF commented Apr 8, 2020

Test of distilBERT with batch and GPU support:
Input:

! sh similarity_distilBERT_batch.sh 40 cls True 

Output:
Confusion matrix:

langs   cs       de       en       es       fr       avg     
cs     0.00%   92.64%   98.43%   92.57%   94.44%   94.52%
de    77.66%    0.00%   91.08%   76.99%   80.25%   81.49%
en    83.82%   71.46%    0.00%   54.08%   61.44%   67.70%
es    79.65%   82.88%   88.68%    0.00%   69.70%   80.23%
fr    79.39%   82.32%   87.01%   60.81%    0.00%   77.38%
avg   80.13%   82.33%   91.30%   71.11%   76.46%   80.26%

Previously, we had:

Input:

! sh similarity_distilBERT.sh 40 cls

Output:

langs   de       en       es       fr       ru       avg     
de     0.00%   91.08%   76.99%   80.25%   84.75%   83.27%
en    71.46%    0.00%   54.08%   61.44%   67.83%   63.70%
es    82.88%   88.68%    0.00%   69.70%   78.75%   80.00%
fr    82.32%   87.01%   60.81%    0.00%   78.95%   77.27%
ru    86.15%   91.91%   76.26%   80.82%    0.00%   83.78%
avg   80.70%   89.67%   67.03%   73.05%   77.57%   77.61%

@MastafaF
Copy link
Owner Author

MastafaF commented Apr 8, 2020

Looking at pairs like ('en', 'es'), we can see that we yield the same results when using batch and GPU than when using one sentence at a time with CPU.

@MastafaF
Copy link
Owner Author

MastafaF commented Apr 8, 2020

Input

! sh similarity_distilBERT_batch.sh 100 cls True 

Output:

Confusion matrix:
langs   cs       de       en       es       fr       avg     
cs     0.00%   89.84%   98.07%   89.88%   91.94%   92.43%
de    72.86%    0.00%   88.68%   71.33%   73.63%   76.62%
en    81.55%   64.47%    0.00%   42.99%   51.85%   60.21%
es    77.46%   78.82%   85.55%    0.00%   60.11%   75.48%
fr    75.42%   76.76%   83.82%   50.18%    0.00%   71.55%
avg   76.82%   77.47%   89.03%   63.59%   69.38%   75.26%

Input:

! sh similarity_distilBERT_batch.sh 40 cls True 

Output:

Confusion matrix:
langs   cs       de       en       es       fr       avg     
cs     0.00%   92.64%   98.43%   92.57%   94.44%   94.52%
de    77.66%    0.00%   91.08%   76.99%   80.25%   81.49%
en    83.82%   71.46%    0.00%   54.08%   61.44%   67.70%
es    79.65%   82.88%   88.68%    0.00%   69.70%   80.23%
fr    79.39%   82.32%   87.01%   60.81%    0.00%   77.38%
avg   80.13%   82.33%   91.30%   71.11%   76.46%   80.26%

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

No branches or pull requests

1 participant