#@title
from IPython.display import HTML

HTML('<iframe width="560" height="315" src="https://www.youtube.com/embed/pUh5cGmNV8Y?rel=0&amp;controls=0&amp;showinfo=0" frameborder="0" allowfullscreen></iframe>')

! pip install datasets transformers[sentencepiece]

from huggingface_hub import notebook_login

notebook_login()

from datasets import load_dataset, load_metric

raw_datasets = load_dataset("glue", "cola")

raw_datasets

from transformers import AutoTokenizer

model_checkpoint = "bert-base-cased"
tokenizer = AutoTokenizer.from_pretrained(model_checkpoint)

def preprocess_function(examples):
    return tokenizer(examples["sentence"], truncation=True)

tokenized_datasets = raw_datasets.map(preprocess_function, batched=True)

tokenized_datasets

from transformers import DataCollatorWithPadding

collator = DataCollatorWithPadding(tokenizer=tokenizer,
                                   return_tensors='tf')

train_dataset = tokenized_datasets['train'].to_tf_dataset(
    columns=['attention_mask', 'input_ids', 'labels', 'token_type_ids'],
    collate_fn=collator,
    batch_size=32,
    shuffle=True
)
validation_dataset = tokenized_datasets['validation'].to_tf_dataset(
    columns=['attention_mask', 'input_ids', 'labels', 'token_type_ids'],
    collate_fn=collator,
    batch_size=32,
    shuffle=False
)

from transformers import TFAutoModelForSequenceClassification

model = TFAutoModelForSequenceClassification.from_pretrained(model_checkpoint)

from transformers import AdamWeightDecay

optimizer = AdamWeightDecay(2e-5, weight_decay_rate=0.01)

model.compile(optimizer=optimizer)

from transformers import PushToHubCallback

callbacks = [PushToHubCallback("model_output/", 
                               tokenizer=tokenizer,
                               hub_model_id="bert-fine-tuned-cola")]

model.fit(train_dataset, validation_data=validation_dataset, epochs=2, callbacks=callbacks)

model.push_to_hub("bert-fine-tuned-cola", commit_message="End of training")



label_names = raw_datasets["train"].features["label"].names
label_names

model.config.id2label = {str(i): lbl for i, lbl in enumerate(label_names)}
model.config.label2id = {lbl: str(i) for i, lbl in enumerate(label_names)}

repo_name = "bert-fine-tuned-cola"
model.config.push_to_hub(repo_name)

loaded_model = TFAutoModelForSequenceClassification.from_pretrained('Rocketknight1/bert-fine-tuned-cola')