eladven commited on
Commit
09f02c8
1 Parent(s): 25781a3

Evaluation results for SetFit/deberta-v3-base__sst2__all-train model as a base model for other tasks

Browse files

As part of a research effort to identify high quality models in Huggingface that can serve as base models for further finetuning, we evaluated this by finetuning on 36 datasets. The model ranks 3rd among all tested models for the microsoft/deberta-v3-base architecture as of 09/01/2023.


To share this information with others in your model card, please add the following evaluation results to your README.md page.

For more information please see https://ibm.github.io/model-recycling/ or contact me.

Best regards,
Elad Venezian
[email protected]
IBM Research AI

Files changed (1) hide show
  1. README.md +14 -0
README.md CHANGED
@@ -61,3 +61,17 @@ The following hyperparameters were used during training:
61
  - Pytorch 1.10.2+cu102
62
  - Datasets 1.18.2
63
  - Tokenizers 0.10.3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
61
  - Pytorch 1.10.2+cu102
62
  - Datasets 1.18.2
63
  - Tokenizers 0.10.3
64
+
65
+ ## Model Recycling
66
+
67
+ [Evaluation on 36 datasets](https://ibm.github.io/model-recycling/model_gain_chart?avg=0.10&mnli_lp=nan&20_newsgroup=0.06&ag_news=0.36&amazon_reviews_multi=0.08&anli=0.63&boolq=1.45&cb=3.57&cola=0.39&copa=-1.40&dbpedia=0.57&esnli=-0.53&financial_phrasebank=1.52&imdb=-0.04&isear=-0.22&mnli=-0.19&mrpc=0.99&multirc=2.00&poem_sentiment=0.77&qnli=-0.19&qqp=0.21&rotten_tomatoes=-0.18&rte=-0.76&sst2=-0.34&sst_5bins=-0.60&stsb=-0.32&trec_coarse=0.24&trec_fine=-0.22&tweet_ev_emoji=0.82&tweet_ev_emotion=0.50&tweet_ev_hate=-3.92&tweet_ev_irony=-0.99&tweet_ev_offensive=-0.17&tweet_ev_sentiment=-0.96&wic=1.20&wnli=-2.61&wsc=2.26&yahoo_answers=-0.27&model_name=SetFit%2Fdeberta-v3-base__sst2__all-train&base_name=microsoft%2Fdeberta-v3-base) using SetFit/deberta-v3-base__sst2__all-train as a base model yields average score of 79.14 in comparison to 79.04 by microsoft/deberta-v3-base.
68
+
69
+ The model is ranked 3rd among all tested models for the microsoft/deberta-v3-base architecture as of 09/01/2023
70
+ Results:
71
+
72
+ | 20_newsgroup | ag_news | amazon_reviews_multi | anli | boolq | cb | cola | copa | dbpedia | esnli | financial_phrasebank | imdb | isear | mnli | mrpc | multirc | poem_sentiment | qnli | qqp | rotten_tomatoes | rte | sst2 | sst_5bins | stsb | trec_coarse | trec_fine | tweet_ev_emoji | tweet_ev_emotion | tweet_ev_hate | tweet_ev_irony | tweet_ev_offensive | tweet_ev_sentiment | wic | wnli | wsc | yahoo_answers |
73
+ |---------------:|----------:|-----------------------:|--------:|--------:|--------:|--------:|-------:|----------:|--------:|-----------------------:|-------:|--------:|--------:|--------:|----------:|-----------------:|--------:|--------:|------------------:|--------:|--------:|------------:|-------:|--------------:|------------:|-----------------:|-------------------:|----------------:|-----------------:|---------------------:|---------------------:|--------:|--------:|--------:|----------------:|
74
+ | 86.4711 | 90.8 | 66.94 | 59.4063 | 84.4343 | 78.5714 | 86.9607 | 57 | 80 | 91.3986 | 86 | 94.452 | 71.6428 | 89.5952 | 90.1961 | 64.2533 | 87.5 | 93.3187 | 91.9936 | 90.2439 | 81.5884 | 94.7248 | 56.3801 | 89.96 | 98 | 90.8 | 47.014 | 84.4476 | 52.2896 | 78.8265 | 84.8837 | 70.8401 | 72.4138 | 67.6056 | 66.3462 | 71.7667 |
75
+
76
+
77
+ For more information, see: [Model Recycling](https://ibm.github.io/model-recycling/)