config = args
config.n_embed = len(inputs.vocab)
config.d_out = len(answers.vocab)
config.n_cells = config.n_layers
# double the number of cells for bidirectional networks
if config.birnn:
config.n_cells *= 2
if args.resume_snapshot:
model = torch.load(
args.resume_snapshot,
map_location=lambda storage, location: storage.cuda(args.gpu))
else:
model = LSTMSentiment(config)
if args.word_vectors:
model.embed.weight.data = inputs.vocab.vectors
model.cuda()
criterion = nn.CrossEntropyLoss()
opt = O.Adam(model.parameters())
# model.embed.requires_grad = False
iterations = 0
start_time = time.time()
best_dev_acc = -1
train_iter.repeat = False
header = ' Time Epoch Loss Dev/Loss CD Loss Accuracy Dev/Accuracy'
config = args
config.n_embed = len(inputs.vocab)
config.d_out = len(answers.vocab)
config.n_cells = config.n_layers
# double the number of cells for bidirectional networks
if config.birnn:
config.n_cells *= 2
if args.resume_snapshot:
model = torch.load(
args.resume_snapshot,
map_location=lambda storage, location: storage.cuda(args.gpu))
else:
model = LSTMSentiment(config)
if args.word_vectors:
model.embed.weight.data = inputs.vocab.vectors
model.cuda()
criterion = nn.CrossEntropyLoss()
opt = O.Adam(model.parameters()) # , lr=args.lr)
# model.embed.requires_grad = False
iterations = 0
start_time = time.time()
best_dev_acc = -1
train_iter.repeat = False
header = ' Time Epoch Loss Dev/Loss CD Loss Accuracy Dev/Accuracy'
def test_load_save_state():
model = LSTMSentiment(0, 2, 300, 150, 3, 'lstm', nn.NLLLoss())
model2 = LSTMSentiment(0, 2, 300, 150, 3, 'lstm', nn.NLLLoss())
model.save_state_files('.')
model2.load_state_files('.')
model3 = LSTMSentiment(0, 1, 300, 150, 3, 'lstm', nn.NLLLoss())
with pytest.raises(Exception):
model3.load_state_files('.')
answers.build_vocab(train)
train_iter, dev_iter, test_iter = data.BucketIterator.splits(
(train, dev, test), batch_size=args.batch_size, device=args.gpu)
config = args
config.n_embed = len(inputs.vocab)
config.d_out = len(answers.vocab)
config.n_cells = config.n_layers
if args.resume_snapshot:
model = torch.load(
args.resume_snapshot,
map_location=lambda storage, location: storage.cuda(args.gpu))
else:
model = LSTMSentiment(config)
if args.word_vectors:
model.embed.weight.data = inputs.vocab.vectors
model.cuda()
criterion = nn.CrossEntropyLoss()
opt = O.Adam(model.parameters())
iterations = 0
start = time.time()
best_dev_acc = -1
train_iter.repeat = False
header = ' Time Epoch Iteration Progress (%Epoch) Loss Dev/Loss Accuracy Dev/Accuracy'
dev_log_template = ' '.join(
'{:>6.0f},{:>5.0f},{:>9.0f},{:>5.0f}/{:<5.0f} {:>7.0f}%,{:>8.6f},{:8.6f},{:12.4f},{:12.4f}'
.split(','))
answers,
fine_grained=False,
train_subtrees=True,
filter_pred=lambda ex: ex.label != 'neutral')
inputs.build_vocab(train, dev, test)
inputs.vocab.load_vectors('glove.6B.300d')
answers.build_vocab(train)
train_iter, dev_iter, test_iter = data.BucketIterator.splits(
(train, dev, test), batch_size=50, repeat=False, device=DEVICE)
############################################
# CREATING MODEL
print("Creating model")
model = LSTMSentiment(embedding_dim=300,
hidden_dim=168,
vocab_size=300,
label_size=2,
gpu_device=DEVICE)
model.word_embeddings.weight.data = inputs.vocab.vectors
model.cuda(device=DEVICE)
# Load previously checkpointed model if it exists
model_path = "model.pt"
if os.path.exists(model_path) and RESUME_CKPT:
print("Loading previously stored model")
model = torch.load(model_path)
loss_function = nn.CrossEntropyLoss()
optimizer = optim.Adam(model.parameters(), lr=0.001)
#############################################
# build a new model
class Config(object):
d_hidden = model.hidden_dim
n_embed = model.vocab_size + 1
d_embed = model.emb_dim
d_out = model.num_out
batch_size = model.batch_size
birnn = False
if USE_RANDOM_MODEL_BASELINE:
from model import LSTMSentiment
config = Config()
new_model = LSTMSentiment(config)
new_model.embed.weight.data.copy_(model.embed.weight.data) # glove data
new_model.use_gpu = False
model = new_model # completely random new model
# [l, r], a = previous line, b = current line
# children: [num_words]
# tree_dict: {}
def build_tree(l, r, a, b, score, tree_dict):
ret = []
i = l
while i <= r:
if a[i] != 0:
j = nonzero(i, a)
ret.append(tree_dict[(i, j)])