Evaluation results for devpranjal/deberta-v3-base-devrev-data model as a base model for other tasks
Browse filesAs 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 07/02/2023.
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
eladv@il.ibm.com
IBM Research AI
README.md
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# devpranjal/deberta-v3-base-devrev-data model
|
| 2 |
+
This model is based on microsoft/deberta-v3-base pretrained model.
|
| 3 |
+
|
| 4 |
+
|
| 5 |
+
## Model Recycling
|
| 6 |
+
|
| 7 |
+
[Evaluation on 36 datasets](https://ibm.github.io/model-recycling/model_gain_chart?avg=0.54&mnli_lp=nan&20_newsgroup=-0.92&ag_news=-0.51&amazon_reviews_multi=-0.14&anli=-0.59&boolq=2.27&cb=5.36&cola=-0.28&copa=11.60&dbpedia=-0.40&esnli=-0.74&financial_phrasebank=3.22&imdb=-0.48&isear=-0.61&mnli=-0.30&mrpc=0.75&multirc=1.38&poem_sentiment=-3.08&qnli=0.30&qqp=0.24&rotten_tomatoes=-0.18&rte=1.77&sst2=-0.34&sst_5bins=0.66&stsb=1.16&trec_coarse=-0.56&trec_fine=0.38&tweet_ev_emoji=0.18&tweet_ev_emotion=-0.70&tweet_ev_hate=1.46&tweet_ev_irony=-2.40&tweet_ev_offensive=0.87&tweet_ev_sentiment=-0.87&wic=-0.37&wnli=1.62&wsc=-0.63&yahoo_answers=0.47&model_name=devpranjal%2Fdeberta-v3-base-devrev-data&base_name=microsoft%2Fdeberta-v3-base) using devpranjal/deberta-v3-base-devrev-data as a base model yields average score of 79.58 in comparison to 79.04 by microsoft/deberta-v3-base.
|
| 8 |
+
|
| 9 |
+
The model is ranked 3rd among all tested models for the microsoft/deberta-v3-base architecture as of 07/02/2023
|
| 10 |
+
Results:
|
| 11 |
+
|
| 12 |
+
| 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 |
|
| 13 |
+
|---------------:|----------:|-----------------------:|--------:|--------:|--------:|--------:|-------:|----------:|--------:|-----------------------:|-------:|--------:|--------:|-------:|----------:|-----------------:|--------:|--------:|------------------:|--------:|--------:|------------:|--------:|--------------:|------------:|-----------------:|-------------------:|----------------:|-----------------:|---------------------:|---------------------:|--------:|-------:|--------:|----------------:|
|
| 14 |
+
| 85.4886 | 89.9333 | 66.72 | 58.1875 | 85.2599 | 80.3571 | 86.2895 | 70 | 79.0333 | 91.1849 | 87.7 | 94.012 | 71.2516 | 89.4833 | 89.951 | 63.6345 | 83.6538 | 93.8129 | 92.0257 | 90.2439 | 84.1155 | 94.7248 | 57.6471 | 91.4423 | 97.2 | 91.4 | 46.374 | 83.2512 | 57.6768 | 77.4235 | 85.9302 | 70.9297 | 70.8464 | 71.831 | 63.4615 | 72.5 |
|
| 15 |
+
|
| 16 |
+
|
| 17 |
+
For more information, see: [Model Recycling](https://ibm.github.io/model-recycling/)
|