def test_sequence_classification_viz():
explainer_string = "I love you , I like you"
seq_explainer = SequenceClassificationExplainer(explainer_string,
DISTILBERT_MODEL,
DISTILBERT_TOKENIZER)
seq_explainer()
seq_explainer.visualize()
import pandas as pd
import numpy as np
import transformers
#We can't use our initial dataset (3GB), too heavy, so instead... :
file_name = "clothes.csv"
ds = pd.read_csv(file_name)
example = ds.iloc[
15, 4] #Getting the right column (4), i.e. the one with the review
classifier = transformers.pipeline('sentiment-analysis',
return_all_scores=True)
tokenizer = classifier.tokenizer
model = classifier.model
#Explaining using Sequence Classification Explainer from the transformers explainer
from transformers_interpret import SequenceClassificationExplainer
cls_explainer = SequenceClassificationExplainer(model, tokenizer)
word_attributions = cls_explainer(example)
cls_explainer.visualize("visualize.html")
# "0": "natural",
# "1": "positive",
# "2": "negative"
from transformers import AutoModelForSequenceClassification, AutoTokenizer
from transformers_interpret import SequenceClassificationExplainer
if __name__ == "__main__":
model_name = "avichr/heBERT"
model = AutoModelForSequenceClassification.from_pretrained(model_name)
tokenizer = AutoTokenizer.from_pretrained(model_name)
# With both the model and tokenizer initialized we are now able to get explanations on an example text.
cls_explainer = SequenceClassificationExplainer(
"I love you, I like you",
model,
tokenizer)
attributions = cls_explainer()
# print(attributions.predicted_class_name)
cls_explainer.visualize("outputs/explainer-viz.html")
