Evaluate LM with financial phrasebank

%config InlineBackend.figure_format='retina'
import logging
from ekorpkit import eKonf

logging.basicConfig(level=logging.WARNING)
print(eKonf.__version__)

0.1.32+4.gfd15d82.dirty

Prepare financial_phrasebank dataset

from ekorpkit import eKonf

ds_cfg = eKonf.compose(config_group="dataset")
ds_cfg.name = "financial_phrasebank"
ds_cfg.path.cache.uri = "gd://15TUjIfQstCYXU82s0pcaxh2QcSWf1zyG:financial_phrasebank.zip"
ds_cfg.data_dir = ds_cfg.path.cached_path
ds = eKonf.instantiate(ds_cfg)
ds.INFO

{'name': 'financial_phrasebank',
 'domain': 'econ',
 'task': 'classification',
 'lang': 'en',
 'column_info': {'keys': {'id': 'id', 'text': 'text'},
  'data': {'id': 'int', 'labels': 'str', 'text': 'str'}},
 'splits': {'train': {'name': 'train',
   'dataset_name': 'financial_phrasebank',
   'data_file': 'financial_phrasebank-train.csv',
   'num_examples': 1445,
   'num_bytes': 176485,
   'num_bytes_median': 112.0,
   'num_bytes_max': 315,
   'num_bytes_min': 15,
   'human_bytes': '172.35 KiB'},
  'test': {'name': 'test',
   'dataset_name': 'financial_phrasebank',
   'data_file': 'financial_phrasebank-test.csv',
   'num_examples': 452,
   'num_bytes': 55518,
   'num_bytes_median': 110.0,
   'num_bytes_max': 296,
   'num_bytes_min': 11,
   'human_bytes': '54.22 KiB'},
  'dev': {'name': 'dev',
   'dataset_name': 'financial_phrasebank',
   'data_file': 'financial_phrasebank-dev.csv',
   'num_examples': 362,
   'num_bytes': 43404,
   'num_bytes_median': 109.0,
   'num_bytes_max': 296,
   'num_bytes_min': 9,
   'human_bytes': '42.39 KiB'}},
 'num_examples': 2259,
 'size_in_bytes': 275407,
 'size_in_human_bytes': '268.95 KiB',
 'data_files': {'train': 'financial_phrasebank-train.csv',
  'test': 'financial_phrasebank-test.csv',
  'dev': 'financial_phrasebank-dev.csv'},
 'meta_files': {},
 'data_files_modified': '2022-04-22 03:18:20',
 'info_updated': '2022-04-22 03:18:20'}

Instantiating a sentiment analyser with financial_phrasebank dataset

model_cfg = eKonf.compose(config_group="model/sentiment=lm")

cfg = eKonf.compose(config_group='pipeline')
cfg.data.dataset = ds_cfg
cfg._pipeline_ = ['predict']
cfg.predict.model = model_cfg
cfg.predict.output_dir = "./data/predict"
cfg.predict.output_file = f'{ds_cfg.name}.parquet'
cfg.num_workers = 100
data = eKonf.instantiate(cfg)

Predicting [text]: 100%|██████████| 97/97 [00:25<00:00,  3.82it/s]
Predicting [text]: 100%|██████████| 91/91 [00:17<00:00,  5.14it/s]
Predicting [text]: 100%|██████████| 91/91 [00:17<00:00,  5.27it/s]

# print(cfg.predict.output_dir)
print(cfg.predict.path.output.suffix)

None

eval_cfg = eKonf.compose(config_group='model/eval=classification')
eval_cfg.to_eval.actual = 'labels'
eval_cfg.to_eval.predicted = 'polarity_label'
eval_cfg.to_eval.labels = ['positive','neutral','negative']
eval_cfg.data_dir = './data/predict'
eval_cfg.data_file = 'financial_phrasebank-*.parquet'
eval_cfg.output_dir = './data/eval'
# eval_cfg.method.average = None
# eKonf.print(eval_cfg)
eKonf.instantiate(eval_cfg)

Accuracy:  0.6423196104471005
Precison:  0.6335214835254505
Recall:  0.6423196104471005
F1 Score:  0.5979784064056457
Model Report: 
___________________________________________________
              precision    recall  f1-score   support

    negative       0.38      0.36      0.37       303
     neutral       0.68      0.89      0.77      1386
    positive       0.65      0.19      0.29       570

    accuracy                           0.64      2259
   macro avg       0.57      0.48      0.48      2259
weighted avg       0.63      0.64      0.60      2259