def test_multibatch(self):
Vi, Ei, Hi, Vo, Eo, Ho, Ha, Hl = 29, 37, 13, 17, 7, 12, 19, 33
encdec = EncoderDecoderNaive(Vi, Ei, Hi, Vo, Eo, Ho, Ha, Hl)
eos_idx = Vo - 1
raw_seq1 = [2, 5, 0, 3], [4, 6]
raw_seq2 = [2, 5, 4, 3, 0, 0, 1, 11, 3], [4, 8, 9, 12, 0]
raw_seq3 = [2, 5, 4, 3, 0, 11,
3], [5, 7, 1, 4, 4, 1, 0, 0, 5, 5, 3, 4, 6, 7, 8]
raw_seq4 = [5, 3, 0, 0, 1, 11, 3], [0, 0, 1, 1]
trg_data = [raw_seq1, raw_seq2, raw_seq3, raw_seq4]
src_batch, tgt_batch, src_mask = utils.make_batch_src_tgt(
trg_data, eos_idx=eos_idx)
loss, attn = encdec(src_batch, tgt_batch, src_mask)
total_loss_naive = 0
total_length = 0
for i in range(len(trg_data)):
raw_s_raw = trg_data[i]
raw_s = [raw_s_raw[0], raw_s_raw[1] + [eos_idx]]
input_seq = [
Variable(np.array([v], dtype=np.int32)) for v in raw_s[0]
]
tgt_seq = [
Variable(np.array([v], dtype=np.int32)) for v in raw_s[1]
]
loss_naive, attn_naive = encdec.naive_call(input_seq, tgt_seq,
None)
total_loss_naive += float(loss_naive.data) * len(raw_s[1])
total_length += len(raw_s[1])
assert abs(total_loss_naive / total_length - float(loss.data)) < 1e-6
def batch_align(encdec, eos_idx, src_tgt_data, batch_size=80, gpu=None):
nb_ex = len(src_tgt_data)
nb_batch = nb_ex / batch_size + (1 if nb_ex % batch_size != 0 else 0)
sum_loss = 0
attn_all = []
for i in range(nb_batch):
current_batch_raw_data = src_tgt_data[i * batch_size:(i + 1) *
batch_size]
# print current_batch_raw_data
src_batch, tgt_batch, src_mask, arg_sort = make_batch_src_tgt(
current_batch_raw_data,
eos_idx=eos_idx,
gpu=gpu,
volatile="on",
need_arg_sort=True)
loss, attn_list = encdec(src_batch,
tgt_batch,
src_mask,
keep_attn_values=True)
deb_attn = de_batch(attn_list,
mask=None,
eos_idx=None,
is_variable=True,
raw=True)
assert len(arg_sort) == len(deb_attn)
de_sorted_attn = [None] * len(deb_attn)
for xpos in xrange(len(arg_sort)):
original_pos = arg_sort[xpos]
de_sorted_attn[original_pos] = deb_attn[xpos]
attn_all += de_sorted_attn
sum_loss += float(loss.data)
return sum_loss, attn_all
def sample_extension(trainer):
encdec = trainer.updater.get_optimizer("main").target
iterator = trainer.updater.get_iterator("main")
mb_raw = iterator.peek()
def s_unk_tag(num, utag):
return "S_UNK_%i" % utag
def t_unk_tag(num, utag):
return "T_UNK_%i" % utag
try:
if encdec.encdec_type() == "ff":
src_seqs, tgt_seqs = list(six.moves.zip(*mb_raw))
sample_once_ff(encdec, src_seqs, tgt_seqs, src_indexer, tgt_indexer, max_nb=20,
s_unk_tag=s_unk_tag, t_unk_tag=t_unk_tag)
else:
src_batch, tgt_batch, src_mask = make_batch_src_tgt(mb_raw, eos_idx=eos_idx, padding_idx=0, gpu=gpu, need_arg_sort=False,
use_chainerx = use_chainerx)
sample_once(encdec, src_batch, tgt_batch, src_mask, src_indexer, tgt_indexer, eos_idx,
max_nb=20,
s_unk_tag=s_unk_tag, t_unk_tag=t_unk_tag)
except CudaException:
log.warn("CUDARuntimeError during sample. Skipping sample")
def convert_mb(mb_raw, device):
return make_batch_src_tgt(mb_raw,
eos_idx=eos_idx,
padding_idx=0,
gpu=device,
volatile="off",
need_arg_sort=False)
def sample_extension(trainer):
encdec = trainer.updater.get_optimizer("main").target
iterator = trainer.updater.get_iterator("main")
mb_raw = iterator.peek()
src_batch, tgt_batch, src_mask = make_batch_src_tgt(
mb_raw,
eos_idx=eos_idx,
padding_idx=0,
gpu=gpu,
volatile="on",
need_arg_sort=False)
def s_unk_tag(num, utag):
return "S_UNK_%i" % utag
def t_unk_tag(num, utag):
return "T_UNK_%i" % utag
sample_once(encdec,
src_batch,
tgt_batch,
src_mask,
src_indexer,
tgt_indexer,
eos_idx,
max_nb=20,
s_unk_tag=s_unk_tag,
t_unk_tag=t_unk_tag)
def convert_mb(mb_raw, device):
return make_batch_src_tgt(mb_raw, eos_idx=eos_idx, padding_idx=0, gpu=device, need_arg_sort=False,
use_chainerx=use_chainerx)