julien-c's picture
julien-c HF staff
Add evaluation results on glue dataset (#6)
83dbcb2
metadata
language: en
license: apache-2.0
datasets:
  - sst2
  - glue
model-index:
  - name: distilbert-base-uncased-finetuned-sst-2-english
    results:
      - task:
          type: text-classification
          name: Text Classification
        dataset:
          name: glue
          type: glue
          config: sst2
          split: validation
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.9105504587155964
            verified: true
          - name: Precision
            type: precision
            value: 0.8978260869565218
            verified: true
          - name: Recall
            type: recall
            value: 0.9301801801801802
            verified: true
          - name: AUC
            type: auc
            value: 0.9716626673402374
            verified: true
          - name: F1
            type: f1
            value: 0.9137168141592922
            verified: true
          - name: loss
            type: loss
            value: 0.39013850688934326
            verified: true

DistilBERT base uncased finetuned SST-2

This model is a fine-tune checkpoint of DistilBERT-base-uncased, fine-tuned on SST-2. This model reaches an accuracy of 91.3 on the dev set (for comparison, Bert bert-base-uncased version reaches an accuracy of 92.7).

For more details about DistilBERT, we encourage users to check out this model card.

Fine-tuning hyper-parameters

  • learning_rate = 1e-5
  • batch_size = 32
  • warmup = 600
  • max_seq_length = 128
  • num_train_epochs = 3.0

Bias

Based on a few experimentations, we observed that this model could produce biased predictions that target underrepresented populations.

For instance, for sentences like This film was filmed in COUNTRY, this binary classification model will give radically different probabilities for the positive label depending on the country (0.89 if the country is France, but 0.08 if the country is Afghanistan) when nothing in the input indicates such a strong semantic shift. In this colab, Aurélien Géron made an interesting map plotting these probabilities for each country.

Map of positive probabilities per country.

We strongly advise users to thoroughly probe these aspects on their use-cases in order to evaluate the risks of this model. We recommend looking at the following bias evaluation datasets as a place to start: WinoBias, WinoGender, Stereoset.