Create eval.csv
Browse files
eval.csv
ADDED
|
@@ -0,0 +1,32 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from transformers import Trainer, AutoModelForSequenceClassification, AutoTokenizer
|
| 2 |
+
from datasets import load_dataset, load_metric
|
| 3 |
+
import json
|
| 4 |
+
|
| 5 |
+
# Load configuration
|
| 6 |
+
with open('../config/config.json') as f:
|
| 7 |
+
config = json.load(f)
|
| 8 |
+
|
| 9 |
+
# Load model and tokenizer
|
| 10 |
+
model = AutoModelForSequenceClassification.from_pretrained('../model')
|
| 11 |
+
tokenizer = AutoTokenizer.from_pretrained(config['model_name'])
|
| 12 |
+
|
| 13 |
+
# Load dataset
|
| 14 |
+
dataset = load_dataset('csv', data_files={'test': '../data/test.csv'})
|
| 15 |
+
tokenized_datasets = dataset.map(lambda x: tokenizer(x['text'], padding="max_length", truncation=True), batched=True)
|
| 16 |
+
|
| 17 |
+
# Evaluation
|
| 18 |
+
metric = load_metric("accuracy")
|
| 19 |
+
|
| 20 |
+
def compute_metrics(eval_pred):
|
| 21 |
+
logits, labels = eval_pred
|
| 22 |
+
predictions = logits.argmax(axis=-1)
|
| 23 |
+
return metric.compute(predictions=predictions, references=labels)
|
| 24 |
+
|
| 25 |
+
trainer = Trainer(
|
| 26 |
+
model=model,
|
| 27 |
+
tokenizer=tokenizer,
|
| 28 |
+
compute_metrics=compute_metrics
|
| 29 |
+
)
|
| 30 |
+
|
| 31 |
+
results = trainer.evaluate(tokenized_datasets['test'])
|
| 32 |
+
print(results)
|