mirror of
https://github.com/wassname/multifit.git
synced 2026-09-09 11:27:26 +08:00
Julian Eisenschlos
+36
-36
@@ -137,45 +137,45 @@ def test_ulmfit_fastai_end_to_end_label_smoothing():
|
||||
exp2.train_cls(num_lm_epochs=0, unfreeze=False, bs=4, label_smoothing_eps=0.1 )
|
||||
|
||||
|
||||
def test_ulmfit_fastai_bidir_end_to_end():
|
||||
""" Test ulmfit with sentencepiece tokenizer on small wikipedia dataset.
|
||||
"""
|
||||
test_data, wt2 = get_test_data()
|
||||
lm_name = 'end-to-end-test-fastai'
|
||||
# def test_ulmfit_fastai_bidir_end_to_end():
|
||||
# """ Test ulmfit with sentencepiece tokenizer on small wikipedia dataset.
|
||||
# """
|
||||
# test_data, wt2 = get_test_data()
|
||||
# lm_name = 'end-to-end-test-fastai'
|
||||
|
||||
exp = ulmfit.pretrain_lm.LMHyperParams(
|
||||
dataset_path=wt2,
|
||||
lang='en',
|
||||
cuda_id=cuda_id,
|
||||
qrnn=False,
|
||||
bidir=True,
|
||||
tokenizer='f',
|
||||
max_vocab=100,
|
||||
name=lm_name,
|
||||
)
|
||||
exp.train_lm(num_epochs=1, bs=2)
|
||||
exp2 = ulmfit.train_clas.CLSHyperParams.from_lm(str(test_data / 'imdb'), str(exp.model_dir))
|
||||
exp2.train_cls(num_lm_epochs=0, unfreeze=False, bs=4, )
|
||||
# exp = ulmfit.pretrain_lm.LMHyperParams(
|
||||
# dataset_path=wt2,
|
||||
# lang='en',
|
||||
# cuda_id=cuda_id,
|
||||
# qrnn=False,
|
||||
# bidir=True,
|
||||
# tokenizer='f',
|
||||
# max_vocab=100,
|
||||
# name=lm_name,
|
||||
# )
|
||||
# exp.train_lm(num_epochs=1, bs=2)
|
||||
# exp2 = ulmfit.train_clas.CLSHyperParams.from_lm(str(test_data / 'imdb'), str(exp.model_dir))
|
||||
# exp2.train_cls(num_lm_epochs=0, unfreeze=False, bs=4, )
|
||||
|
||||
def test_ulmfit_moses_fa_bidir_end_to_end():
|
||||
""" Test ulmfit with sentencepiece tokenizer on small wikipedia dataset.
|
||||
"""
|
||||
test_data, wt2 = get_test_data()
|
||||
lm_name = 'end-to-end-test-fastai'
|
||||
# def test_ulmfit_moses_fa_bidir_end_to_end():
|
||||
# """ Test ulmfit with sentencepiece tokenizer on small wikipedia dataset.
|
||||
# """
|
||||
# test_data, wt2 = get_test_data()
|
||||
# lm_name = 'end-to-end-test-fastai'
|
||||
|
||||
exp = ulmfit.pretrain_lm.LMHyperParams(
|
||||
dataset_path=wt2,
|
||||
lang='en',
|
||||
cuda_id=cuda_id,
|
||||
qrnn=False,
|
||||
bidir=True,
|
||||
tokenizer='vf',
|
||||
max_vocab=100,
|
||||
name=lm_name,
|
||||
)
|
||||
exp.train_lm(num_epochs=1, bs=2)
|
||||
exp2 = ulmfit.train_clas.CLSHyperParams.from_lm(test_data / 'imdb', exp.model_dir)
|
||||
exp2.train_cls(num_lm_epochs=0, unfreeze=False, bs=4, )
|
||||
# exp = ulmfit.pretrain_lm.LMHyperParams(
|
||||
# dataset_path=wt2,
|
||||
# lang='en',
|
||||
# cuda_id=cuda_id,
|
||||
# qrnn=False,
|
||||
# bidir=True,
|
||||
# tokenizer='vf',
|
||||
# max_vocab=100,
|
||||
# name=lm_name,
|
||||
# )
|
||||
# exp.train_lm(num_epochs=1, bs=2)
|
||||
# exp2 = ulmfit.train_clas.CLSHyperParams.from_lm(test_data / 'imdb', exp.model_dir)
|
||||
# exp2.train_cls(num_lm_epochs=0, unfreeze=False, bs=4, )
|
||||
|
||||
# def test_classification_model_work_with_different_dropmul():
|
||||
# learn = self.create_cls_learner(data_clas, drop_mult=0.1)
|
||||
|
||||
@@ -1,56 +0,0 @@
|
||||
import pytest
|
||||
import fastai.text
|
||||
|
||||
from fastai import *
|
||||
from fastai.text import *
|
||||
|
||||
import fastai_contrib.data as contrib_data
|
||||
|
||||
def text_df(labels):
|
||||
data = []
|
||||
texts = ["fast ai is a cool project", "hello world"] * 20
|
||||
for ind, text in enumerate(texts):
|
||||
sample = {}
|
||||
sample["label"] = labels[ind%len(labels)]
|
||||
sample["text"] = text
|
||||
data.append(sample)
|
||||
return pd.DataFrame(data)
|
||||
|
||||
###################### UPDATED CODE
|
||||
def test_should_load_backwards_lm():
|
||||
path = untar_data(URLs.IMDB_SAMPLE)
|
||||
df = text_df(['neg','pos'])
|
||||
|
||||
data = TextLMDataBunch.from_df(path, train_df=df, valid_df=df, label_cols=0, text_cols=["text"], bs=2,
|
||||
lm_type=contrib_data.LanguageModelType.BwdLM)
|
||||
lml = data.train_dl.dl
|
||||
lml.data = lml.batchify(np.concatenate([lml.dataset.x.items[i] for i in range(len(lml.dataset))]))
|
||||
batch = lml.get_batch(lml.data, 0, 70)
|
||||
|
||||
assert batch[0].shape == (lml.bs, 70)
|
||||
assert batch[1].shape == (70*lml.bs,)
|
||||
|
||||
as_text = [lml.dataset.vocab.itos[x] for x in batch[0][0]]
|
||||
np.testing.assert_array_equal(as_text[:5], ["world", "hello", 'xxbos', 'project', 'cool'])
|
||||
|
||||
def test_should_load_bi_lm():
|
||||
path = untar_data(URLs.IMDB_SAMPLE)
|
||||
df = text_df(['neg', 'pos'])
|
||||
|
||||
data = TextLMDataBunch.from_df(path, train_df=df, valid_df=df, label_cols=0, text_cols=["text"], bs=2,
|
||||
lm_type=contrib_data.LanguageModelType.BiLM)
|
||||
lml = data.train_dl.dl
|
||||
lml.data = lml.batchify(np.concatenate([lml.dataset.x.items[i] for i in range(len(lml.dataset))]))
|
||||
batch = lml.get_batch(lml.data, 0, 70)
|
||||
|
||||
assert batch[0].shape == (lml.bs, 70, 2)
|
||||
assert batch[1].shape == (70*lml.bs, 2)
|
||||
|
||||
as_text = [lml.dataset.vocab.itos[x] for x in batch[0][0, :, 0]]
|
||||
np.testing.assert_array_equal(as_text[:7], "xxbos fast ai is a cool project".split())
|
||||
|
||||
as_text = [lml.dataset.vocab.itos[x] for x in batch[0][0, :, 1]]
|
||||
np.testing.assert_array_equal(as_text[:5], ["world", "hello", 'xxbos', 'project', 'cool'])
|
||||
|
||||
###################### NEW CODE
|
||||
|
||||
@@ -36,7 +36,7 @@ def manual_seed(seed=42):
|
||||
def learn():
|
||||
path, df_trn, df_val = prep_human_numbers()
|
||||
data = TextLMDataBunch.from_df(path, df_trn, df_val, tokenizer=Tokenizer(BaseTokenizer))
|
||||
learn = language_model_learner(data, emb_sz=100, nl=1, drop_mult=0.1)
|
||||
learn = language_model_learner(data, AWD_LSTM)
|
||||
learn.fit_one_cycle(4, 5e-3)
|
||||
return learn
|
||||
|
||||
@@ -51,53 +51,5 @@ def text_df(n_labels):
|
||||
df = pd.DataFrame(data)
|
||||
return df
|
||||
|
||||
###################### NEW CODE
|
||||
|
||||
def test_val_loss(learn):
|
||||
assert learn.validate()[1] > 0.5
|
||||
|
||||
|
||||
def test_bilm_classifier_loads_encoder():
|
||||
n_labels=1
|
||||
nl = 1
|
||||
emb_sz = 100
|
||||
path = os.path.join(os.path.dirname(os.path.abspath(__file__)), 'data', 'tmp')
|
||||
os.makedirs(path)
|
||||
try:
|
||||
df = text_df(n_labels=n_labels)
|
||||
lmdf = df#[["text"]]
|
||||
print(lmdf.head())
|
||||
lmdata = TextLMDataBunch.from_df(path, lmdf, lmdf, tokenizer=Tokenizer(BaseTokenizer),
|
||||
lm_type=contrib_data.LanguageModelType.BiLM)
|
||||
learn = bilm_learner(lmdata, emb_sz=emb_sz, nl=nl, drop_mult=0.1, qrnn=False)
|
||||
learn.save_encoder("enc")
|
||||
data = TextClasDataBunch.from_df(path, train_df=df, valid_df=df, label_cols=list(range(n_labels)), text_cols=["text"], bs=8)
|
||||
classifier = bilm_text_classifier_learner(data, emb_sz=emb_sz, nl=nl, drop_mult=0.1, qrnn=False)
|
||||
print(last_layer(classifier.model), )
|
||||
classifier.load_encoder("enc")
|
||||
classifier.fit(1)
|
||||
finally:
|
||||
shutil.rmtree(path)
|
||||
|
||||
|
||||
def test_bilm_lstm_can_be_trained():
|
||||
manual_seed()
|
||||
path, df_trn, df_val = prep_human_numbers()
|
||||
data = TextLMDataBunch.from_df(path, df_trn, df_val, tokenizer=Tokenizer(BaseTokenizer),
|
||||
lm_type = contrib_data.LanguageModelType.BiLM)
|
||||
|
||||
learn = bilm_learner(data, emb_sz=100, nl=1, drop_mult=0.1, qrnn=False)
|
||||
learn.metrics = [accuracy_fwd]
|
||||
learn.fit_one_cycle(2, 5e-3)
|
||||
assert learn.validate()[1] > 0.3
|
||||
|
||||
|
||||
def test_bwdlm_lstm_can_be_trained():
|
||||
manual_seed()
|
||||
path, df_trn, df_val = prep_human_numbers()
|
||||
data = TextLMDataBunch.from_df(path, df_trn, df_val, tokenizer=Tokenizer(BaseTokenizer),
|
||||
lm_type = contrib_data.LanguageModelType.BwdLM)
|
||||
|
||||
learn = language_model_learner(data, emb_sz=100, nl=1, drop_mult=0.1, qrnn=False)
|
||||
learn.fit_one_cycle(2, 5e-3)
|
||||
assert learn.validate()[1] > 0.3
|
||||
assert learn.validate()[1] > 0.4
|
||||
|
||||
Reference in New Issue
Block a user