README.md

metadata

license: mit
dataset_info:
  features:
    - name: input
      dtype: string
    - name: output
      dtype: string
    - name: instruction
      dtype: string
  splits:
    - name: train
      num_bytes: 1906560
      num_examples: 9543
    - name: validation
      num_bytes: 479540
      num_examples: 2388
  download_size: 728648
  dataset_size: 2386100
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*

zeroshot/twitter-financial-news-sentiment prepared for LLM fine-tuning by adding an instruction column and mapping the label from numeric to string ({0:"negative", 1:'positive', 2:'neutral'}).

Source

from datasets import load_dataset
import datasets

from huggingface_hub import notebook_login
notebook_login()

ds = load_dataset('zeroshot/twitter-financial-news-sentiment')

num_to_label = {
    0: 'negative',
    1: 'positive',
    2: 'neutral',
}

instruction = 'What is the sentiment of this tweet? Please choose an answer from {negative/neutral/positive}.'

# Training split

ds_train = ds['train']
ds_train = ds_train.to_pandas()
ds_train['label'] = ds_train['label'].apply(num_to_label.get)
ds_train['instruction'] = instruction
ds_train.columns = ['input', 'output', 'instruction']
ds_train = datasets.Dataset.from_pandas(ds_train)

ds_train.push_to_hub("twitter-financial-news-sentiment")

# Validation split

ds_valid = ds['validation']
ds_valid = ds_valid.to_pandas()
ds_valid['label'] = ds_valid['label'].apply(num_to_label.get)
ds_valid['instruction'] = instruction
ds_valid.columns = ['input', 'output', 'instruction']
ds_valid = datasets.Dataset.from_pandas(ds_valid, split='validation')

ds_valid.push_to_hub("twitter-financial-news-sentiment", split='validation')