def evaluate_input(searcher, word2idx, idx2word, device):
tokenizer = DialogSpacyTokenizer(lower=True, specials=HRED_SPECIAL_TOKENS)
to_token_ids = ToTokenIds(word2idx, specials=HRED_SPECIAL_TOKENS)
to_tensor = ToTensor()
transforms = [tokenizer, to_token_ids, to_tensor]
previous = None
while True:
try:
# Get input sentence
input_sentence1 = input('> ')
if input_sentence1 == 'q' or input_sentence1 == 'quit': break
# Normalize sentence
input_sentence1 = normalizeString(input_sentence1)
# Evaluate sentence
for t in transforms:
input_sentence1 = t(input_sentence1)
output_words = evaluate(searcher, idx2word, previous,
input_sentence1, device)
previous = input_sentence1
print(output_words)
output_words[:] = [
x for x in output_words if not (x == 'EOS' or x == 'PAD')
]
print('Bot:', ' '.join(output_words))
except KeyError:
print("Error: Encountered unknown word.")
def setup(self, stage=None):
if self.setup_has_run:
return
super(PLDataModuleFromCorpus, self).setup(stage=stage)
train_corpus, train_labels = zip(*self.train) # type: ignore
val_corpus, val_labels = zip(*self.val) # type: ignore
if not self.no_test_set:
test_corpus, test_labels = zip(*self.test) # type: ignore
self.train_corpus, self.val_corpus, self.test_corpus = self._create_corpora(
train_corpus, val_corpus, test_corpus, self.corpus_args)
to_tensor = ToTensor(device="cpu")
if self.language_model:
self.train = CorpusLMDataset(self.train_corpus).map(to_tensor)
self.val = CorpusLMDataset(self.val_corpus).map(to_tensor)
if not self.no_test_set:
self.test = CorpusLMDataset(self.test_corpus).map(to_tensor)
else:
self.train = CorpusDataset(self.train_corpus,
train_labels).map(to_tensor)
self.val = CorpusDataset(self.val_corpus,
val_labels).map(to_tensor)
if not self.no_test_set:
self.test = CorpusDataset(self.test_corpus,
test_labels).map(to_tensor)
def __init__(
self,
data: List[Dict[str, Any]],
modalities: Union[List[str], Set[str]] = {"text", "audio", "visual"},
text_is_tokens: bool = False,
label_selector: Optional[Callable] = None,
):
super(MOSEI, self).__init__(data, modalities)
def default_label_selector(l):
return l[0][0]
if label_selector is None:
label_selector = default_label_selector
self.map(label_selector, "label", lazy=True)
for m in self.modalities:
if m == "text" and text_is_tokens:
self.map(ToTensor(dtype=torch.long), m, lazy=True)
else:
self.map(ToTensor(dtype=torch.float), m, lazy=True)
def __init__(
self,
data: List[Dict[str, Any]],
modalities: Union[List[str], Set[str]] = {"text", "audio", "visual"},
text_is_tokens: bool = False,
binary: bool = False,
):
super(MOSI, self).__init__(data, modalities)
def label_selector(l):
return l.item()
self.map(label_selector, "label", lazy=True)
if binary:
self.map(binarize, "label", lazy=True)
for m in self.modalities:
if m == "text" and text_is_tokens:
self.map(ToTensor(dtype=torch.long), m, lazy=True)
else:
self.map(ToTensor(dtype=torch.float), m, lazy=True)
else:
word2idx, idx2word = word2idx_from_dataset(
vocab_dict, most_freq=10000, extra_tokens=HRED_SPECIAL_TOKENS)
embeddings = None
emb_dim = options.emb_dim
vocab_size = len(word2idx)
print("Vocabulary size: {}".format(vocab_size))
# --- set dataset transforms ---
tokenizer = DialogSpacyTokenizer(lower=True,
prepend_sos=True,
append_eos=True,
specials=HRED_SPECIAL_TOKENS)
to_token_ids = ToTokenIds(word2idx, specials=HRED_SPECIAL_TOKENS)
to_tensor = ToTensor()
dataset = dataset.map(tokenizer).map(to_token_ids).map(to_tensor)
print("Dataset size: {}".format(len(dataset)))
import ipdb
ipdb.set_trace()
# --- make train and val loaders ---
collator_fn = HRED_Collator(device='cpu')
train_loader, val_loader = train_test_split(dataset,
batch_train=BATCH_TRAIN_SIZE,
batch_val=BATCH_VAL_SIZE,
collator_fn=collator_fn,
test_size=0.2)
pad_index = word2idx[HRED_SPECIAL_TOKENS.PAD.value]
sos_index = word2idx[HRED_SPECIAL_TOKENS.SOS.value]
directory='data/',
train=True,
dev=True,
test=True,
extracted_name='wikitext-2',
url=
'https://s3.amazonaws.com/research.metamind.io/wikitext/wikitext-2-v1.zip', # noqa: E501
unknown_token=SPECIAL_TOKENS.UNK.value,
eos_token=SPECIAL_TOKENS.EOS.value)
vocab = create_vocab(train + dev,
vocab_size=vocab_size,
extra_tokens=SPECIAL_TOKENS.to_list())
replace_unk = ReplaceUnknownToken()
to_token_ids = ToTokenIds(vocab)
to_tensor = ToTensor(device='cpu')
def create_dataloader(base):
wrapped = (LMDataset(base, max_len=max_len).map(replace_unk).map(
to_token_ids).map(to_tensor).apply_transforms())
return DataLoader(wrapped,
batch_size=128,
num_workers=1,
pin_memory=True,
collate_fn=collate_fn)
train_loader = create_dataloader(train[:1000])
dev_loader = create_dataloader(dev[:1000])
test_loader = create_dataloader(test[:1000])
device = 'cuda' if torch.cuda.is_available() else 'cpu'
hidden_size = 300
epochs = 40
lexicons = False
lex_size = 99
loader = EmbeddingsLoader('/data/embeddings/glove.840B.300d.txt', 300)
word2idx, idx2word, embeddings = loader.load()
embeddings = torch.tensor(embeddings)
with open("avec.pkl", "rb") as handle:
_file = pickle.load(handle)
tokenizer = SpacyTokenizer()
replace_unknowns = ReplaceUnknownToken()
to_token_ids = ToTokenIds(word2idx)
to_tensor = ToTensor(device=DEVICE)
train = AVECDataset(_file,
max_word_length,
transforms=Compose([
tokenizer, replace_unknowns, to_token_ids,
to_tensor
]),
split='train')
dev = AVECDataset(_file,
max_word_length,
transforms=Compose([
tokenizer, replace_unknowns, to_token_ids, to_tensor
]),
split='dev')
test = AVECDataset(_file,
from slp.modules.transformer import Transformer
from slp.util.pytorch import pad_mask, subsequent_mask
simplefilter(action="ignore")
pl.utilities.seed.seed_everything(42)
collate_fn = Seq2SeqCollator(device="cpu")
# All tokens are different. Should get 100% accuracy
sentence = "The big brown fox jumps over the lazy dog".split(" ")
vocab = create_vocab([sentence], vocab_size=-1, special_tokens=SPECIAL_TOKENS)
vocab = dict(zip(vocab.keys(), itertools.count()))
to_token_ids = ToTokenIds(vocab)
to_tensor = ToTensor(device="cpu")
class DummyDataset(Dataset):
def __init__(self):
self.data = [(sentence[0:-1], sentence[1:])]
def __len__(self):
return 1
def __getitem__(self, i):
s, t = self.data[i]
return to_tensor(to_token_ids(s)), to_tensor(to_token_ids(t))
