This commit is contained in:
LixayTF 2024-06-10 18:03:51 +02:00
parent b0a7744214
commit 134b585d49
2 changed files with 3353 additions and 0 deletions

102
main.py Normal file
View File

@ -0,0 +1,102 @@
import pandas as pd
from datasets import load_dataset, load_metric
from transformers import AutoTokenizer, AutoModelForTokenClassification, TrainingArguments, Trainer, DataCollatorForTokenClassification
import numpy as np
# Load the CoNLL-2003 dataset with trust_remote_code
dataset = load_dataset("conll2003", trust_remote_code=True)
# Load the tokenizer
tokenizer = AutoTokenizer.from_pretrained("bert-base-cased")
# Define label list and map labels to IDs
label_list = dataset['train'].features['ner_tags'].feature.names
# Tokenize and align labels function
def tokenize_and_align_labels(examples):
tokenized_inputs = tokenizer(examples['tokens'], truncation=True, padding='max_length', is_split_into_words=True)
labels = []
for i, label in enumerate(examples['ner_tags']):
word_ids = tokenized_inputs.word_ids(batch_index=i) # Map tokens to their respective word.
previous_word_idx = None
label_ids = []
for word_idx in word_ids: # Set the special tokens to -100.
if word_idx is None:
label_ids.append(-100)
elif word_idx != previous_word_idx: # Only label the first token of a given word.
label_ids.append(label[word_idx])
else:
label_ids.append(-100)
previous_word_idx = word_idx
labels.append(label_ids)
tokenized_inputs["labels"] = labels
return tokenized_inputs
# Tokenize the datasets
tokenized_datasets = dataset.map(tokenize_and_align_labels, batched=True)
# Split the dataset into training and evaluation sets
train_dataset = tokenized_datasets["train"]
eval_dataset = tokenized_datasets["validation"]
# Load the model
model = AutoModelForTokenClassification.from_pretrained("bert-base-cased", num_labels=len(label_list))
# Data collator for token classification
data_collator = DataCollatorForTokenClassification(tokenizer)
# Training arguments
training_args = TrainingArguments(
output_dir='./results',
evaluation_strategy="epoch",
learning_rate=2e-5,
per_device_train_batch_size=16,
per_device_eval_batch_size=16,
num_train_epochs=3,
weight_decay=0.01,
)
# Define the trainer
trainer = Trainer(
model=model,
args=training_args,
train_dataset=train_dataset,
eval_dataset=eval_dataset,
data_collator=data_collator,
compute_metrics=lambda p: {
"accuracy": (p.predictions.argmax(-1) == p.label_ids).astype(np.float32).mean().item(),
"f1": load_metric("seqeval").compute(predictions=np.argmax(p.predictions, axis=2), references=p.label_ids)['overall_f1']
},
)
# Train the model
trainer.train()
# Evaluate the model
results = trainer.evaluate()
# Print the results
print("Evaluation results:", results)
# Predict on the evaluation set
predictions, labels, _ = trainer.predict(eval_dataset)
predictions = np.argmax(predictions, axis=2)
# Convert the predictions and labels to the original tags
true_labels = [[label_list[l] for l in label if l != -100] for label in labels]
true_predictions = [
[label_list[p] for (p, l) in zip(prediction, label) if l != -100]
for prediction, label in zip(predictions, labels)
]
# Create a DataFrame for the results
results_df = pd.DataFrame({
'tokens': eval_dataset['tokens'],
'true_labels': true_labels,
'predicted_labels': true_predictions
})
# Save the results to a CSV file
results_df.to_csv('mnt/data/ner_results.csv', index=False)
print("Wyniki analizy NER zostały zapisane do pliku 'mnt/data/ner_results.csv'.")

3251
mnt/data/ner_results.csv Normal file

File diff suppressed because it is too large Load Diff