README.md

metadata

language: en
widget:
  - text: >-
      I am really upset that I have to call up to three times to the number on
      the back of my insurance card for my call to be answer
tags:
  - sagemaker
  - roberta-base
  - text classification
license: apache-2.0
datasets:
  - emotion
model-index:
  - name: sagemaker-roberta-base-emotion
    results:
      - task:
          name: Multi Class Text Classification
          type: text-classification
        dataset:
          name: emotion
          type: emotion
        metrics:
          - name: Validation Accuracy
            type: accuracy
            value: 94.1
          - name: Validation F1
            type: f1
            value: 94.13
      - task:
          type: text-classification
          name: Text Classification
        dataset:
          name: emotion
          type: emotion
          config: default
          split: test
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.931
            verified: true
          - name: Precision Macro
            type: precision
            value: 0.8833042147663716
            verified: true
          - name: Precision Micro
            type: precision
            value: 0.931
            verified: true
          - name: Precision Weighted
            type: precision
            value: 0.9337002742192515
            verified: true
          - name: Recall Macro
            type: recall
            value: 0.9087144572668905
            verified: true
          - name: Recall Micro
            type: recall
            value: 0.931
            verified: true
          - name: Recall Weighted
            type: recall
            value: 0.931
            verified: true
          - name: F1 Macro
            type: f1
            value: 0.8949974527433656
            verified: true
          - name: F1 Micro
            type: f1
            value: 0.931
            verified: true
          - name: F1 Weighted
            type: f1
            value: 0.9318434300647934
            verified: true
          - name: loss
            type: loss
            value: 0.17379647493362427
            verified: true

roberta-base

This model is a fine-tuned model that was trained using Amazon SageMaker and the new Hugging Face Deep Learning container.

• Problem type: Multi Class Text Classification (emotion detection).

It achieves the following results on the evaluation set:

• Loss: 0.1613253802061081
• f1: 0.9413321705151999

Hyperparameters

{
    "epochs": 10,
    "train_batch_size": 16,
    "learning_rate": 3e-5, 
    "weight_decay":0.01,
    "load_best_model_at_end": true,
    "model_name":"roberta-base",
    "do_eval": True,
    "load_best_model_at_end":True
}

Validation Metrics

key	value
eval_accuracy	0.941
eval_f1	0.9413321705151999
eval_loss	0.1613253802061081
eval_recall	0.941
eval_precision	0.9419519436781406