About the labels
# 0 negative
# 1 neutral
# 2 positive
In [1]:
#!pip install transformers[torch]
#!pip install torch transformers datasets
#!pip install accelerate -U
# Getting ready the work environment. Importing libraries and modules:
import matplotlib.pyplot as plt
import numpy as np
import seaborn as sns
import torch
import time
import pandas as pd
from datasets import load_dataset
from sklearn.metrics import accuracy_score, classification_report, confusion_matrix
from transformers import BertTokenizer, BertForSequenceClassification, Trainer, TrainingArguments
C:\Users\nonox\anaconda3\lib\site-packages\transformers\utils\generic.py:260: UserWarning: torch.utils._pytree._register_pytree_node is deprecated. Please use torch.utils._pytree.register_pytree_node instead. torch.utils._pytree._register_pytree_node( C:\Users\nonox\anaconda3\lib\site-packages\transformers\utils\generic.py:260: UserWarning: torch.utils._pytree._register_pytree_node is deprecated. Please use torch.utils._pytree.register_pytree_node instead. torch.utils._pytree._register_pytree_node(
WARNING:tensorflow:From C:\Users\nonox\anaconda3\lib\site-packages\keras\src\losses.py:2976: The name tf.losses.sparse_softmax_cross_entropy is deprecated. Please use tf.compat.v1.losses.sparse_softmax_cross_entropy instead.
In [2]:
# Time consumed (starts)
start_time = time.time()
# Loading the pre-trained FinBERT model and tokenizer
model_ = "ProsusAI/finbert"
tokenizer = BertTokenizer.from_pretrained(model_)
model = BertForSequenceClassification.from_pretrained(model_)
# We directly load the Financial PhraseBank dataset from The Hugging Face
dataset = load_dataset("financial_phrasebank", "sentences_allagree") # https://huggingface.co/datasets/financial_phrasebank/viewer/sentences_allagree
#"sentences_50agree"
#"sentences_66agree"
#"sentences_75agree"
Found cached dataset financial_phrasebank (C:/Users/nonox/.cache/huggingface/datasets/financial_phrasebank/sentences_allagree/1.0.0/550bde12e6c30e2674da973a55f57edde5181d53f5a5a34c1531c53f93b7e141)
0%| | 0/1 [00:00<?, ?it/s]
In [3]:
# Preprocess the dataset
def preprocess_function(examples):
return tokenizer(examples["sentence"],
padding="longest",
max_length= 45,
truncation = True)
dataset = dataset.map(preprocess_function, batched=True)
# Split the dataset into training and validation sets
train_test_split = dataset['train'].train_test_split(test_size=0.2, seed=42)
train_dataset = train_test_split['train']
eval_dataset = train_test_split['test']
Loading cached processed dataset at C:\Users\nonox\.cache\huggingface\datasets\financial_phrasebank\sentences_allagree\1.0.0\550bde12e6c30e2674da973a55f57edde5181d53f5a5a34c1531c53f93b7e141\cache-ab52e02081578aae.arrow Loading cached split indices for dataset at C:\Users\nonox\.cache\huggingface\datasets\financial_phrasebank\sentences_allagree\1.0.0\550bde12e6c30e2674da973a55f57edde5181d53f5a5a34c1531c53f93b7e141\cache-10a3db2b34327c77.arrow and C:\Users\nonox\.cache\huggingface\datasets\financial_phrasebank\sentences_allagree\1.0.0\550bde12e6c30e2674da973a55f57edde5181d53f5a5a34c1531c53f93b7e141\cache-f1ca108c82a0a6c4.arrow
In [4]:
# Training arguments
training_args = TrainingArguments(
output_dir = "./results",
evaluation_strategy = "epoch",
learning_rate = 0.0001,
per_device_train_batch_size = 25,
per_device_eval_batch_size = 25,
num_train_epochs = 2,
warmup_steps = 500,
save_strategy = "epoch"
)
# Define the trainer
trainer = Trainer(
model = model,
args = training_args,
train_dataset = train_dataset,
eval_dataset = eval_dataset,
)
# Fine-tune the model
trainer.train()
C:\Users\nonox\anaconda3\lib\site-packages\accelerate\accelerator.py:436: FutureWarning: Passing the following arguments to `Accelerator` is deprecated and will be removed in version 1.0 of Accelerate: dict_keys(['dispatch_batches']). Please pass an `accelerate.DataLoaderConfiguration` instead: dataloader_config = DataLoaderConfiguration(dispatch_batches=None) warnings.warn(
[146/146 16:47, Epoch 2/2]
| Epoch | Training Loss | Validation Loss |
|---|---|---|
| 1 | No log | 0.401435 |
| 2 | No log | 0.137344 |
Out[4]:
TrainOutput(global_step=146, training_loss=1.0236330424269584, metrics={'train_runtime': 1014.7056, 'train_samples_per_second': 3.57, 'train_steps_per_second': 0.144, 'total_flos': 83759484288060.0, 'train_loss': 1.0236330424269584, 'epoch': 2.0})
In [5]:
# Evaluate the fine-tuned model
predictions = trainer.predict(eval_dataset)
preds = np.argmax(predictions.predictions, axis=1)
labels = eval_dataset["label"]
accuracy = accuracy_score(labels, preds)
report = classification_report(labels, preds)
print(f"Validation Accuracy: {accuracy:.2f}")
print(f"Classification Report:\n{report}")
Validation Accuracy: 0.96
Classification Report:
precision recall f1-score support
0 0.89 0.89 0.89 73
1 0.99 0.98 0.98 280
2 0.92 0.93 0.93 100
accuracy 0.96 453
macro avg 0.93 0.93 0.93 453
weighted avg 0.96 0.96 0.96 453
In [6]:
# Confusion matrix
cm = confusion_matrix(labels, preds)
# Plot the confusion matrix using seaborn heatmap
plt.figure(figsize=(8, 6))
sns.heatmap(cm,
annot=True,
fmt='d',
cmap='seismic',
cbar=False,
xticklabels=['Negative', 'Neutral', 'Positive'],
yticklabels=['Negative', 'Neutral', 'Positive'])
plt.title('Confusion Matrix')
plt.xlabel('Predicted Labels')
plt.ylabel('True Labels')
plt.show()
In [7]:
# Load the test dataset
test_dataset = load_dataset("financial_phrasebank", "sentences_75agree")
test_dataset = test_dataset.map(preprocess_function, batched=True)
# Evaluate the fine-tuned model on the test set
test_predictions = trainer.predict(test_dataset['train'])
test_preds = np.argmax(test_predictions.predictions, axis=1)
test_labels = test_dataset['train']["label"]
test_accuracy = accuracy_score(test_labels, test_preds)
test_report = classification_report(test_labels, test_preds)
Found cached dataset financial_phrasebank (C:/Users/nonox/.cache/huggingface/datasets/financial_phrasebank/sentences_75agree/1.0.0/550bde12e6c30e2674da973a55f57edde5181d53f5a5a34c1531c53f93b7e141)
0%| | 0/1 [00:00<?, ?it/s]
Loading cached processed dataset at C:\Users\nonox\.cache\huggingface\datasets\financial_phrasebank\sentences_75agree\1.0.0\550bde12e6c30e2674da973a55f57edde5181d53f5a5a34c1531c53f93b7e141\cache-06dc7416bca07405.arrow
In [8]:
print(f"Test Accuracy: {test_accuracy:.2f}")
print(f"Test Classification Report:\n{test_report}")
Test Accuracy: 0.93
Test Classification Report:
precision recall f1-score support
0 0.82 0.85 0.84 420
1 0.96 0.95 0.96 2146
2 0.89 0.90 0.89 887
accuracy 0.93 3453
macro avg 0.89 0.90 0.90 3453
weighted avg 0.93 0.93 0.93 3453
In [9]:
# Confusion matrix
test_cm = confusion_matrix(test_labels, test_preds)
# Plot the confusion matrix using seaborn heatmap
plt.figure(figsize=(8, 6))
sns.heatmap(test_cm,
annot=True,
fmt='d',
cmap='seismic',
cbar=False,
xticklabels=['Negative', 'Neutral', 'Positive'],
yticklabels=['Negative', 'Neutral', 'Positive'])
plt.title('Confusion Matrix')
plt.xlabel('Predicted Labels')
plt.ylabel('True Labels')
plt.show()
In [10]:
# Total Time Consumed
end_time = time.time()
execution_time = end_time - start_time
print(f"Total Execution Time: {execution_time} seconds")
Total Execution Time: 1322.466394662857 seconds
In [ ]:
In [ ]: