Open
Description
`import numpy as np
from rouge import Rouge
rouge = Rouge()
def test_loop(dataloader, model):
preds, labels = [], []
model.eval()
for batch_data in tqdm(dataloader):
batch_data = batch_data.to(device)
with torch.no_grad():
generated_tokens = model.generate(
batch_data["input_ids"],
attention_mask=batch_data["attention_mask"],
max_length=max_target_length,
num_beams=4,
no_repeat_ngram_size=2,
).cpu().numpy()
if isinstance(generated_tokens, tuple):
generated_tokens = generated_tokens[0]
label_tokens = batch_data["labels"].cpu().numpy()
decoded_preds = tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)
label_tokens = np.where(label_tokens != -100, label_tokens, tokenizer.pad_token_id)
decoded_labels = tokenizer.batch_decode(label_tokens, skip_special_tokens=True)
preds += [' '.join(pred.strip()) for pred in decoded_preds]
labels += [' '.join(label.strip()) for label in decoded_labels]
scores = rouge.get_scores(hyps=preds, refs=labels)[0]
result = {key: value['f'] * 100 for key, value in scores.items()}
result['avg'] = np.mean(list(result.values()))
print(f"Rouge1: {result['rouge-1']:>0.2f} Rouge2: {result['rouge-2']:>0.2f} RougeL: {result['rouge-l']:>0.2f}\n")
return result`
Metadata
Assignees
Labels
No labels