ap.add_argument("--w-featdrop", type=float, default = 0.2)
ap.add_argument("--rnn", default = 'GRU')
opt = ap.parse_args()
o = opt
data_trn = load(opt.input_prefix)
data_tst = load(opt.test_prefix)
trn_labels = to_categorical(np.array(data_trn.labels))
if not o.c_maxlen:
o.c_maxlen = np.max(data_trn.len_char)
c_vocab = Counter({k:v for k,v in data_trn.chars.items() if v > o.c_cutoff})
c_trn, _ = doc_to_numseq(np.array(data_trn.docs), vocab=c_vocab,
pad=o.c_maxlen)
c_tst, _ = doc_to_numseq(np.array(data_tst.docs), vocab=c_vocab,
pad=o.c_maxlen)
if not o.w_maxlen:
o.w_maxlen = np.max(data_trn.len_word)
w_vocab = Counter({k:v for k,v in data_trn.words.items() if v > o.w_cutoff})
w_trn, _ = doc_to_numseq(np.array(data_trn.docs), vocab=w_vocab,
tokenizer="word", pad=o.w_maxlen)
w_tst, _ = doc_to_numseq(np.array(data_tst.docs), vocab=w_vocab,
tokenizer="word", pad=o.w_maxlen)
c_inp = Input(shape=(o.c_maxlen, ), name='char_input')
w_inp = Input(shape=(o.w_maxlen, ), name='word_input')
trn_labels = to_categorical(np.array(data.labels)[trn_idx])
dev_labels = to_categorical(np.array(data.labels)[dev_idx])
search_iter = 1000
search_done = set()
for _ in range(search_iter):
o, h = Options.sample()
if h in search_done: continue
search_done.add(h)
if not o.c_maxlen:
o.c_maxlen = np.max(data.len_char)
c_vocab = Counter({k: v for k, v in data.chars.items() if v > o.c_cutoff})
c_trn, _ = doc_to_numseq(np.array(data.docs)[trn_idx],
vocab=c_vocab,
pad=o.c_maxlen)
c_dev, _ = doc_to_numseq(np.array(data.docs)[dev_idx],
vocab=c_vocab,
pad=o.c_maxlen)
if not o.w_maxlen:
o.w_maxlen = np.max(data.len_word)
w_vocab = Counter({k: v for k, v in data.words.items() if v > o.w_cutoff})
w_trn, _ = doc_to_numseq(np.array(data.docs)[trn_idx],
vocab=w_vocab,
tokenizer="word",
pad=o.w_maxlen)
w_dev, _ = doc_to_numseq(np.array(data.docs)[dev_idx],
vocab=w_vocab,
tokenizer="word",
trn_idx, dev_idx = list(ssp.split(data.docs, data.labels))[0]
trn_labels = to_categorical(np.array(data.labels)[trn_idx])
dev_labels = to_categorical(np.array(data.labels)[dev_idx])
search_iter = 1000
search_done = set()
for _ in range(search_iter):
o, h = Options.sample()
if h in search_done: continue
search_done.add(h)
if not o.c_maxlen:
o.c_maxlen = np.max(data.len_char)
c_vocab = Counter({k:v for k,v in data.chars.items() if v > o.c_cutoff})
c_trn, _ = doc_to_numseq(np.array(data.docs)[trn_idx], vocab=c_vocab,
pad=o.c_maxlen)
c_dev, _ = doc_to_numseq(np.array(data.docs)[dev_idx], vocab=c_vocab,
pad=o.c_maxlen)
if not o.w_maxlen:
o.w_maxlen = np.max(data.len_word)
w_vocab = Counter({k:v for k,v in data.words.items() if v > o.w_cutoff})
w_trn, _ = doc_to_numseq(np.array(data.docs)[trn_idx], vocab=w_vocab,
tokenizer="word", pad=o.w_maxlen)
w_dev, _ = doc_to_numseq(np.array(data.docs)[dev_idx], vocab=w_vocab,
tokenizer="word", pad=o.w_maxlen)
c_inp = Input(shape=(o.c_maxlen, ), name='char_input')
w_inp = Input(shape=(o.w_maxlen, ), name='word_input')
c_emb = Embedding(len(c_vocab) + 4, o.c_embdim, mask_zero=True,