An Introduction To HuggingFace Transformers for NLP

from transformers import pipeline

task_pipe = pipeline(f"{SELECTED_TASK}")

text = "Hello world! Hugging Face + Weights and Biases = [MASK]"

output = task_pipe(text)

# AutoModelForTASK should be replaced with any desired task type
from transformers import AutoTokenizer, AutoModelForTASK

tokenizer = AutoTokenizer.from_pretrained(f"{SELECTED_MODEL_TOKENIZER}")
model = AutoModelForTASK.from_pretrained(f"{SELECTED_MODEL}")

text = "Hello my friends! How are you doing today?"
tokenized_text = tokenizer.encode([text])

output = model.generate(**tokenized_text)

from transformers import TrainingArguments, Trainer
training_args = TrainingArguments(
		  output_dir="test_trainer", 
		  evaluation_strategy="epoch",
		  report_to="wandb") #🪄🐝
trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=small_train_dataset,
    eval_dataset=small_eval_dataset,
    compute_metrics=compute_metrics,
)
trainer.train()

SELECTED_TASK = "sentiment_analysis"
INPUT = "I hate you"

## From Above ##
from transformers import pipeline

task_pipe = pipeline(f"{SELECTED_TASK}")

OUTPUT = task_pipe(INPUT)
print(f"label: {OUTPUT['label']}, with score: {round(OUTPUT['score'], 4)}")

not paraphrase: 10%
is paraphrase: 90%

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

tokenizer = AutoTokenizer.from_pretrained("bert-base-cased-finetuned-mrpc")
model = AutoModelForSequenceClassification.from_pretrained("bert-base-cased-finetuned-mrpc")

classes = ["not paraphrase", "is paraphrase"]

sequence_0 = "The company HuggingFace is based in New York City"
sequence_1 = "Apples are especially bad for your health"
sequence_2 = "HuggingFace's headquarters are situated in Manhattan"

# The tokenizer will automatically add any model specific separators (i.e. <CLS> and <SEP>) and tokens to
# the sequence, as well as compute the attention masks.
paraphrase = tokenizer(sequence_0, sequence_2, return_tensors="pt")
not_paraphrase = tokenizer(sequence_0, sequence_1, return_tensors="pt")

paraphrase_classification_logits = model(**paraphrase).logits
not_paraphrase_classification_logits = model(**not_paraphrase).logits

paraphrase_results = torch.softmax(paraphrase_classification_logits, dim=1).tolist()[0]
not_paraphrase_results = torch.softmax(not_paraphrase_classification_logits, dim=1).tolist()[0]

# Should be paraphrase
for i in range(len(classes)):
    print(f"{classes[i]}: {int(round(paraphrase_results[i] * 100))}%")

# Should not be paraphrase
for i in range(len(classes)):
    print(f"{classes[i]}: {int(round(not_paraphrase_results[i] * 100))}%")

SELECTED_TASK = "question-answering"
CONTEXT = "Extractive Question Answering is the task of extracting an answer from a text given a question."
QUERY = "What is extractive question answering?"

## From Above ##
from transformers import pipeline

task_pipe = pipeline(f"{SELECTED_TASK}")

OUTPUT = task_pipe(question=QUERY, context=CONTEXT)
print(
    f"Answer: '{result['answer']}', score: {round(result['score'], 4)}, start: {result['start']}, end: {result['end']}"
)

## From Above ##
from transformers import pipeline
from pprint import pprint

SELECTED_TASK = "text-generation"
MODEL = "gpt2"
task_pipe = pipeline(f"{SELECTED_TASK}", model = MODEL)

INPUT = "Hello, I'm a language model"
OUTPUT = task_pipe(INPUT, max_length = 30, num_return_sequences=3)

pprint(OUTPUT)

## From Above ##
from transformers import pipeline
from pprint import pprint

SELECTED_TASK = "summarization"
task_pipe = pipeline(f"{SELECTED_TASK}")

INPUT = "Paris is the capital and most populous city of France, with an estimated population of 2,175,601 residents as of 2018, in an area of more than 105 square kilometres (41 square miles). The City of Paris is the centre and seat of government of the region and province of Île-de-France, or Paris Region, which has an estimated population of 12,174,880, or about 18 percent of the population of France as of 2017."
OUTPUT = task_pipe(INPUT)

pprint(OUTPUT)

## From Above ##
from transformers import pipeline
from pprint import pprint

lang1 = "en"
lang2 = "fr"
SELECTED_TASK = f"translation_{lang1}_to_{lang2}"

task_pipe = pipeline(f"{SELECTED_TASK}")

INPUT = "How are you?"
OUTPUT = task_pipe(INPUT)

pprint(OUTPUT)

from transformers import TrainingArguments, Trainer

# Load model similar to what was shown above
# model = AutoModelForTASK.from_pretrained("{SELECTED_MODEL}")
# (...)
training_args = TrainingArguments(
		  output_dir="test_trainer", 
		  evaluation_strategy="epoch",
		  report_to="wandb") #🪄🐝
trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=small_train_dataset,
    eval_dataset=small_eval_dataset,
    compute_metrics=compute_metrics,
)
trainer.train()