Training: "'Trainer' object has no attribute 'epoch'"

🤗Transformers
1

def finetune():

progress = widgets.FloatProgress(value=0.1, min=0.0, max=1.0, bar_style = ‘info’)

block_size = tokenizer.max_len

train_dataset = TextDataset(tokenizer=tokenizer, file_path=’/content/Armenian Para.txt’, block_size=block_size)

data_collator = DataCollatorForLanguageModeling(tokenizer=tokenizer, mlm=False)

epochs = 2.5 # limit before overfitting

training_args = TrainingArguments(

                output_dir= '/content/ParaCorpus',

                overwrite_output_dir=True,

                do_train=True,

                num_train_epochs=2.5,

                per_gpu_train_batch_size=1,

                prediction_loss_only=True,

                logging_steps=5,

                save_steps=0,

                seed=random.randint(0,2**32-1))

trainer = Trainer(

                model=model,

                tokenizer=tokenizer,

                args=training_args,

                data_collator=data_collator,

                train_dataset=train_dataset,)

progress.value = 0.4

p_start, p_end = 0.4, 1.

def progressify(f):

def inner(*args, **kwargs):

  if trainer.epoch is not None:

    progress.value = p_start + trainer.epoch / epochs * (p_end - p_start)

    return f(*args, **kwargs)

return inner

try:

trainer.training_step = progressify(trainer.training_step)

trainer.train()

except KeyboardInterrupt:

print('Model will be saved')

finally:

trainer.save_model('/content/ParaCorpus')

tokenizer.save_pretrained('/content/ParaCorpus')

def clean_prediction(text):

token = ‘<|endoftext|>’

while len(token)>1:

text = text.replace(token, '')

token = token[:-1]

text = text.strip()

if len(text) != 0:

if text[-1] == '"' and text.count('"') % 2: text = text[:-1]

return text.strip()

finetune()