Fine-tuning through training

Introduction to LLMs in Python

Jasmin Ludolf

Senior Data Science Content Developer, DataCamp

Training Arguments

from transformers import Trainer, 
TrainingArguments

training_args = TrainingArguments(

  output_dir="./finetuned",

  evaluation_strategy="epoch",

  num_train_epochs=3,

  learning_rate=2e-5,







)

TrainingArguments(): customize training settings
See documentation for all parameters
Values depend on use, dataset, speed
output_dir: output directory
eval_strategy: when to evaluate "epoch", "steps", or "none"
num_train_epochs: number of training epochs
learning_rate: for optimizer

Training Arguments

from transformers import Trainer, 
TrainingArguments

training_args = TrainingArguments(
  output_dir="./finetuned",
  evaluation_strategy="epoch",
  num_train_epochs=3,
  learning_rate=2e-5,

  per_device_train_batch_size=8,
  per_device_eval_batch_size=8,

  weight_decay=0.01,

)

per_device_train_batch_size and per_device_eval_batch_size define the batch size
weight_decay: applied to the optimizer to avoid overfitting

Trainer class

from transformers import Trainer, 
TrainingArguments

training_args = TrainingArguments(...)

trainer = Trainer(

    model=model,

    args=training_args,

    train_dataset=tokenized_training_data,

    eval_dataset=tokenized_test_data,

    tokenizer=tokenizer

)


trainer.train()

model: the model to fine-tune
args: the training arguments
train_dataset: the data used for training
eval_dataset: the data used for evaluation
tokenizer: the tokenizer

Number of training loops: Dataset size, num_train_epochs, per_device_train_batch_size and per_device_eval_batch_size

Trainer output

{'eval_loss': 0.398524671792984, 'eval_runtime': 33.3145, 'eval_samples_per_second': 46.916, 
'eval_steps_per_second': 5.883, 'epoch': 1.0}
{'eval_loss': 0.1745782047510147, 'eval_runtime': 33.5202, 'eval_samples_per_second': 46.629, 
'eval_steps_per_second': 5.847, 'epoch': 2.0}
{'loss': 0.4272, 'grad_norm': 15.558795928955078, 'learning_rate': 2.993197278911565e-06, 
'epoch': 2.5510204081632653}
{'eval_loss': 0.12216147780418396, 'eval_runtime': 33.2238, 'eval_samples_per_second': 47.045, 
'eval_steps_per_second': 5.899, 'epoch': 3.0}
{'train_runtime': 673.0528, 'train_samples_per_second': 6.967, 'train_steps_per_second': 0.874, 
'train_loss': 0.40028538347101533, 'epoch': 3.0}
TrainOutput(global_step=588, training_loss=0.40028538347101533, metrics={'train_runtime': 673.0528, 
'train_samples_per_second': 6.967, 'train_steps_per_second': 0.874, 
'train_loss': 0.40028538347101533, 'epoch': 3.0})

Using the fine-tuned model

new_data = ["This is movie was disappointing!", "This is the best movie ever!"]


new_input = tokenizer(new_data, return_tensors="pt", padding=True, truncation=True, max_length=64)


with torch.no_grad():
    outputs = model(**new_input)


predicted_labels = torch.argmax(outputs.logits, dim=1).tolist()

label_map = {0: "NEGATIVE", 1: "POSITIVE"}
for i, predicted_label in enumerate(predicted_labels):
    sentiment = label_map[predicted_label]
    print(f"\nInput Text {i + 1}: {new_data[i]}")
    print(f"Predicted Label: {sentiment}")

Fine-tuning results

Input Text 1: This is movie was disappointing!
Predicted Sentiment: NEGATIVE

Input Text 2: This is the best movie ever!
Predicted Sentiment: POSITIVE

Saving models and tokenizers

model.save_pretrained("my_finetuned_files")

tokenizer.save_pretrained("my_finetuned_files")

# Loading a saved model
model = AutoModelForSequenceClassification.from_pretrained("my_finetuned_files")
tokenizer = AutoTokenizer.from_pretrained("my_finetuned_files")

Let's practice!

Introduction to LLMs in Python