104 lines
3.8 KiB
Python
104 lines
3.8 KiB
Python
import torch
|
|
from transformers.file_utils import is_tf_available, is_torch_available, is_torch_tpu_available
|
|
from transformers import BertTokenizerFast, BertForSequenceClassification
|
|
from transformers import Trainer, TrainingArguments
|
|
import numpy as np
|
|
import random
|
|
from sklearn.metrics import accuracy_score
|
|
import pandas as pd
|
|
|
|
|
|
class Dataset(torch.utils.data.Dataset):
|
|
def __init__(self, encodings, labels):
|
|
self.encodings = encodings
|
|
self.labels = labels
|
|
|
|
def __getitem__(self, idx):
|
|
item = {k: torch.tensor(v[idx]) for k, v in self.encodings.items()}
|
|
item["labels"] = torch.tensor([self.labels[idx]])
|
|
return item
|
|
|
|
def __len__(self):
|
|
return len(self.labels)
|
|
|
|
|
|
def set_seed(seed: int):
|
|
random.seed(seed)
|
|
np.random.seed(seed)
|
|
if is_torch_available():
|
|
torch.manual_seed(seed)
|
|
torch.cuda.manual_seed_all(seed)
|
|
if is_tf_available():
|
|
import tensorflow as tf
|
|
|
|
tf.random.set_seed(seed)
|
|
|
|
|
|
def compute_metrics(pred):
|
|
labels = pred.label_ids
|
|
preds = pred.predictions.argmax(-1)
|
|
acc = accuracy_score(labels, preds)
|
|
return {
|
|
'accuracy': acc,
|
|
}
|
|
|
|
|
|
def get_prediction(text):
|
|
inputs = tokenizer(text, padding=True, truncation=True, max_length=max_length, return_tensors="pt").to("cuda")
|
|
outputs = model(**inputs)
|
|
probs = outputs[0].softmax(1)
|
|
return probs.argmax()
|
|
|
|
|
|
set_seed(1)
|
|
|
|
train_texts = \
|
|
pd.read_csv('train/in.tsv.xz', compression='xz', sep='\t', header=None, error_bad_lines=False, quoting=3)[0].tolist()
|
|
train_labels = pd.read_csv('train/expected.tsv', sep='\t', header=None, quoting=3)[0].tolist()
|
|
dev_texts = pd.read_csv('dev-0/in.tsv.xz', compression='xz', sep='\t', header=None, quoting=3)[0].tolist()
|
|
dev_labels = pd.read_csv('dev-0/expected.tsv', sep='\t', header=None, quoting=3)[0].tolist()
|
|
# test_texts = pd.read_table('test-A/in.tsv.xz', compression='xz', sep='\t', header=None, quoting=3)
|
|
|
|
model_name = "bert-base-uncased"
|
|
max_length = 512
|
|
tokenizer = BertTokenizerFast.from_pretrained(model_name, do_lower_case=True)
|
|
|
|
train_encodings = tokenizer(train_texts, truncation=True, padding=True, max_length=max_length)
|
|
valid_encodings = tokenizer(dev_texts, truncation=True, padding=True, max_length=max_length)
|
|
|
|
train_dataset = Dataset(train_encodings, train_labels)
|
|
valid_dataset = Dataset(valid_encodings, dev_labels)
|
|
|
|
model = BertForSequenceClassification.from_pretrained(
|
|
model_name, num_labels=len(pd.unique(train_labels))).to("cuda")
|
|
|
|
training_args = TrainingArguments(
|
|
output_dir='./results', # output directory
|
|
num_train_epochs=3, # total number of training epochs
|
|
per_device_train_batch_size=1, # batch size per device during training
|
|
per_device_eval_batch_size=1, # batch size for evaluation
|
|
warmup_steps=500, # number of warmup steps for learning rate scheduler
|
|
weight_decay=0.01, # strength of weight decay
|
|
logging_dir='./logs', # directory for storing logs
|
|
load_best_model_at_end=True, # load the best model when finished training (default metric is loss)
|
|
# but you can specify `metric_for_best_model` argument to change to accuracy or other metric
|
|
logging_steps=200, # log & save weights each logging_steps
|
|
evaluation_strategy="steps", # evaluate each `logging_steps`
|
|
)
|
|
|
|
trainer = Trainer(
|
|
model=model, # the instantiated Transformers model to be trained
|
|
args=training_args, # training arguments, defined above
|
|
train_dataset=train_dataset, # training dataset
|
|
eval_dataset=valid_dataset, # evaluation dataset
|
|
compute_metrics=compute_metrics, # the callback that computes metrics of interest
|
|
)
|
|
|
|
trainer.train()
|
|
|
|
trainer.evaluate()
|
|
|
|
model_path = "bert-base-uncased"
|
|
model.save_pretrained(model_path)
|
|
tokenizer.save_pretrained(model_path)
|