logger = logging.getLogger()
# on met le niveau du logger à DEBUG, comme ça il écrit tout
logger.setLevel(logging.INFO)
fmt = logging.Formatter('%(asctime)s: %(message)s', '%m/%d/%Y %I:%M:%S %p')
console = logging.StreamHandler()
console.setFormatter(fmt)
console.setLevel(logging.INFO)
logger.addHandler(console)
# Here we load only a small chunk of the embeddings (100k most common words)
# You can change it if you want
all_words = set(line.strip() for line in open('all_sst_words.txt'))
emb_dict = EmbeddingsDictionary(word_whitelist=all_words)
data = SifDataset()
train_exs, train_labels, train_freq = dataset.preprocess_dataset(
data.train, emb_dict.dictionary)
logging.info('Loaded train, size={}, npos={}'.format(len(train_exs),
sum(train_labels).sum()))
dev_exs, dev_labels, dev_freq = dataset.preprocess_dataset(
data.dev, emb_dict.dictionary)
logging.info('Loaded dev, size={}, npos={}'.format(len(dev_exs),
sum(dev_labels).sum()))
model = BowModel(emb_dict.emb, train_freq)
loss_fn = nn.NLLLoss()
optimized_params = filter(lambda p: p.requires_grad, model.parameters())
optimizer = optim.Adam(optimized_params, lr=0.003)
import torch
from torch.autograd import Variable
import dataset
from dataset import SifDataset
data = SifDataset()
checkpoint = torch.load('model.pth')
model = checkpoint['net'].eval()
dictionary = checkpoint['dict']
# print('Loaded model {}, reported w/ accuracy {}'.format(model, checkpoint['score']))
test_loss = 0
correct = 0
exs, labels = dataset.preprocess_dataset(data.dev, dictionary)
for data, target in zip(exs, labels):
data, target = Variable(data.unsqueeze(0), volatile=True), Variable(target)
output = model(data)
pred = output.data.max(1)[1] # get the index of the max log-probability
correct += pred.eq(target.data).cpu().sum()
test_loss /= len(exs)
print("%.1f" % (100 * correct / len(exs)))