def evaluate_metrics(model, test_iter):
model.eval()
fields = test_iter.dataset.fields
greedy = Greedy(use_stop=True)
df = pd.DataFrame(columns=[
"Answer", "Target Question", "Predited Question", "bleu1", "bleu2",
"bleu3", "bleu4"
])
pbar = tqdm(test_iter,
total=len(test_iter),
unit=' batches',
disable=hp.tqdm)
for i, batch in enumerate(pbar):
greedy.set_maxlen(len(batch.que[1:]))
outputs, attention = model(batch.ans, greedy)
seq_len, batch_size, vocab_size = outputs.size()
preds = outputs.topk(1)[1]
source = sequence_to_text(batch.ans[:, 0].data, fields['ans'])
prediction = sequence_to_text(preds[:, 0].data, fields['que'])
target = sequence_to_text(batch.que[1:, 0].data, fields['que'])
bleu = calc_bleu([clean(target)], clean(prediction))
df = df.append(
{
"Answer": " ".join(clean(source)),
"Target Question": " ".join(clean(target)),
"Predited Question": " ".join(clean(prediction)),
"bleu1": bleu["bleu1"],
"bleu2": bleu["bleu2"],
"bleu3": bleu["bleu3"],
"bleu4": bleu["bleu4"]
},
ignore_index=True)
df.to_csv("eval_results.csv")
def main():
valid_dataset = TranslateDataset('valid')
with open(Config.vocab_file, 'rb') as file:
data = pickle.load(file)
# 加载id转为文本 的词表
src_idx2char = data['dict']['src_idx2char']
tgt_idx2char = data['dict']['tgt_idx2char']
src_text, tgt_text = valid_dataset[0]
src_text = sequence_to_text(src_text, src_idx2char) # 将id转为文本
src_text = ' '.join(src_text)
print('src_text: ' + src_text)
tgt_text = sequence_to_text(tgt_text, tgt_idx2char) # 将id转为文本
tgt_text = ' '.join(tgt_text)
print('tgt_text: ' + tgt_text)
def main():
valid_dataset = Qingyun11wChatDataset('valid')
# print(valid_dataset[0])
# 加载词表
with open(Config.vocab_file, 'rb') as file:
data = pickle.load(file)
idx2char = data['dict']['idx2char']
for i in range(10):
src_text, tgt_text = valid_dataset[i]
src_text = sequence_to_text(src_text, idx2char)
src_text = ''.join(src_text)
print('src_text: ' + src_text)
tgt_text = sequence_to_text(tgt_text, idx2char)
tgt_text = ''.join(tgt_text)
print('tgt_text: ' + tgt_text)
def main():
from utils import sequence_to_text
valid_dataset = AiChallenger2017Dataset('valid')
print(valid_dataset[0])
with open(vocab_file, 'rb') as file:
data = pickle.load(file)
src_idx2char = data['dict']['src_idx2char']
tgt_idx2char = data['dict']['tgt_idx2char']
src_text, tgt_text = valid_dataset[0]
src_text = sequence_to_text(src_text, src_idx2char)
src_text = ' '.join(src_text)
print('src_text: ' + src_text)
tgt_text = sequence_to_text(tgt_text, tgt_idx2char)
tgt_text = ''.join(tgt_text)
print('tgt_text: ' + tgt_text)
import config
from tqdm import tqdm
from utils import parse_args, sequence_to_text
args = parse_args()
collate_fn = TextMelCollate(config.n_frames_per_step)
train_dataset = TextMelLoader('train', config)
print('len(train_dataset): ' + str(len(train_dataset)))
dev_dataset = TextMelLoader('dev', config)
print('len(dev_dataset): ' + str(len(dev_dataset)))
text, mel = train_dataset[0]
print('text: ' + str(text))
text = sequence_to_text(text.numpy().tolist())
text = ''.join(text)
print('text: ' + str(text))
print('type(mel): ' + str(type(mel)))
text_lengths = []
mel_lengths = []
for data in tqdm(dev_dataset):
text, mel = data
text = sequence_to_text(text.numpy().tolist())
text = ''.join(text)
mel = mel.numpy()
# print('text: ' + str(text))
# print('mel.size: ' + str(mel.size))
char2idx = data['char2idx']
idx2char = data['idx2char']
args = parse_args()
collate_fn = TextMelCollate(config.n_frames_per_step)
train_dataset = TextMelLoader(training_files, config)
print('len(train_dataset): ' + str(len(train_dataset)))
dev_dataset = TextMelLoader(validation_files, config)
print('len(dev_dataset): ' + str(len(dev_dataset)))
text, mel = train_dataset[0]
print('text: ' + str(text))
text = sequence_to_text(text.numpy().tolist(), idx2char)
text = ''.join(text)
print('text: ' + str(text))
print('type(mel): ' + str(type(mel)))
text_lengths = []
mel_lengths = []
for data in tqdm(dev_dataset):
text, mel = data
text = sequence_to_text(text.numpy().tolist(), idx2char)
text = ''.join(text)
mel = mel.numpy()
# print('text: ' + str(text))
# print('mel.size: ' + str(mel.size))