metadata

language: en
tags:
  - autotrain
  - DEV
widget:
  - text: Operating profit jumped to EUR 47 million from EUR 6.6 million
datasets:
  - rajistics/autotrain-data-auditor-sentiment
  - FinanceInc/auditor_sentiment
co2_eq_emissions: 3.165771608457648
model-index:
  - name: auditor_sentiment_finetuned
    results:
      - task:
          type: text-classification
          name: Text Classification
        dataset:
          name: FinanceInc/auditor_sentiment
          type: glue
          split: validation
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.862
            verified: true
          - name: F1
            type: f1
            value: 0.845
            verified: true
          - name: Recall
            type: recall
            value: 0.846
            verified: true
          - name: Precision
            type: precision
            value: 0.844
            verified: true
      - task:
          type: text-classification
          name: Text Classification
        dataset:
          name: FinanceInc/auditor_sentiment_2021
          type: glue
          split: validation
        metrics:
          - name: Accuracy
            type: accuracy
            value: 0.848937
            verified: true
          - name: F1
            type: f1
            value: 0.848282
            verified: true
          - name: Recall
            type: recall
            value: 0.808937
            verified: true
          - name: Precision
            type: precision
            value: 0.818542
            verified: true

Auditor Review Sentiment Model

This model has been finetuned from the proprietary version of FinBERT trained internally using demo.org proprietary dataset of auditor evaluation of sentiment.

FinBERT is a BERT model pre-trained on a large corpora of financial texts. The purpose is to enhance financial NLP research and practice in the financial domain, hoping that financial practitioners and researchers can benefit from this model without the necessity of the significant computational resources required to train the model.

Training Data

This model was fine-tuned using Autotrain from the demo-org/auditor_review review dataset.

Model Status

This model is currently being evaluated in development until the end of the quarter. Based on the results, it may be elevated to production.

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0002
train_batch_size: 16
eval_batch_size: 8
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 4
mixed_precision_training: Native AMP

Model Trained Using AutoTrain

Problem type: Multi-class Classification
Model ID: 1167143226
CO2 Emissions (in grams): 3.165771608457648

Validation Metrics

Loss: 0.3418470025062561
Accuracy: 0.8617131062951496
Macro F1: 0.8448284352912685
Micro F1: 0.8617131062951496
Weighted F1: 0.8612696670395574
Macro Precision: 0.8440532616584138
Micro Precision: 0.8617131062951496
Weighted Precision: 0.8612762332366959
Macro Recall: 0.8461980005490884
Micro Recall: 0.8617131062951496
Weighted Recall: 0.8617131062951496

Usage

You can use cURL to access this model:

$ curl -X POST -H "Authorization: Bearer YOUR_API_KEY" -H "Content-Type: application/json" -d '{"inputs": "I love AutoTrain"}' https://api-inference.huggingface.co/models/rajistics/autotrain-auditor-sentiment-1167143226

Or Python API:

from transformers import AutoModelForSequenceClassification, AutoTokenizer

model = AutoModelForSequenceClassification.from_pretrained("rajistics/autotrain-auditor-sentiment-1167143226", use_auth_token=True)

tokenizer = AutoTokenizer.from_pretrained("rajistics/autotrain-auditor-sentiment-1167143226", use_auth_token=True)

inputs = tokenizer("I love AutoTrain", return_tensors="pt")

outputs = model(**inputs)