def batch_train_increment(dataset, input_model=None, output_model=None, lang='en',
factor=1, dropout=0.2, n_iter=1, batch_size=10,
eval_id=None, eval_split=None, long_text=False, silent=False,shuffle=False,gpu_id = None):
"""
Batch train a new text classification model from annotations. Prodigy will
export the best result to the output directory, and include a JSONL file of
the training and evaluation examples. You can either supply a dataset ID
containing the evaluation data, or choose to split off a percentage of
examples for evaluation.
"""
#log("RECIPE: Starting recipe textcat.batch-train", locals())
if(gpu_id):
spacy.util.use_gpu(gpu_id)
if(n_iter ==1):
print("one pass mode")
print("batch_size",batch_size)
print(factor,type(factor))
DB = connect()
print_ = get_print(silent)
random.seed(0)
if input_model is not None:
nlp = spacy.load(input_model, disable=['ner'])
print_('\nLoaded model {}'.format(input_model))
else:
print("build your customized model")
nlp = spacy.load('en_core_web_lg')
pt_model = FastText(vocab_size=684831, emb_dim = 300)
pt_model.embeds.weight.data.copy_(torch.from_numpy(nlp.vocab.vectors.data))
model = PyTorchWrapper(pt_model)
#textcat = TextCategorizer(nlp.vocab,model)
textcat = Loss_TextCategorizer(nlp.vocab,model)
nlp.add_pipe(textcat)
examples = DB.get_dataset(dataset)
labels = {eg['label'] for eg in examples}
labels = list(sorted(labels))
print(labels)
model = TextClassifier(nlp, labels, long_text=long_text,
low_data=len(examples) < 1000)
if shuffle:
print("it's shuffling")
random.shuffle(examples)
else:
print("it's not shuffling")
if eval_id:
evals = DB.get_dataset(eval_id)
print_("Loaded {} evaluation examples from '{}'"
.format(len(evals), eval_id))
else:
examples, evals, eval_split = split_evals(examples, eval_split)
print_("Using {}% of examples ({}) for evaluation"
.format(round(eval_split * 100), len(evals)))
if shuffle:
random.shuffle(examples)
examples = examples[:int(len(examples) * factor)]
print_(printers.trainconf(dropout, n_iter, batch_size, factor,
len(examples)))
if len(evals) > 0:
print_(printers.tc_update_header())
# best_acc = {'accuracy': 0}
# best_model = None
if long_text:
examples = list(split_sentences(nlp, examples, min_length=False))
batch_idx = 0
start_time = datetime.now()
for batch in cytoolz.partition_all(batch_size,
tqdm.tqdm(examples, leave=False)):
batch = list(batch)
for i in range(n_iter):
loss = model.update(batch, revise=False, drop=dropout)
if len(evals) > 0:
#print("optimizer averages",model.optimizer.averages)
with nlp.use_params(model.optimizer.averages):
acc = model.evaluate(tqdm.tqdm(evals, leave=False))
#print_(printers.tc_update(i, loss, acc))
end_time = datetime.now() -start_time
print('Time:[{0} seconds], Epoch: [{1}/{2}], batch: [{3}/{4}], Loss:{5}, Accuracy:{6}'.format(
end_time.seconds,i+1, n_iter, batch_idx+1, len(examples)//batch_size, loss, acc['accuracy']))
batch_idx += 1
return acc
def batch_train(dataset, input_model=None, output_model=None, lang='en',
factor=1, dropout=0.2, n_iter=10, batch_size=10,
eval_id=None, eval_split=None, long_text=False, silent=False,shuffle=False):
"""
Batch train a new text classification model from annotations. Prodigy will
export the best result to the output directory, and include a JSONL file of
the training and evaluation examples. You can either supply a dataset ID
containing the evaluation data, or choose to split off a percentage of
examples for evaluation.
"""
#log("RECIPE: Starting recipe textcat.batch-train", locals())
print("batch_size",batch_size)
print(factor,type(factor))
DB = connect()
print_ = get_print(silent)
random.seed(0)
if input_model is not None:
nlp = spacy.load(input_model, disable=['ner'])
print_('\nLoaded model {}'.format(input_model))
else:
print("build your customized model")
nlp = spacy.load('en_core_web_lg')
pt_model = FastText(vocab_size=684831, emb_dim = 300)
pt_model.embeds.weight.data.copy_(torch.from_numpy(nlp.vocab.vectors.data))
model = PyTorchWrapper(pt_model)
textcat = TextCategorizer(nlp.vocab,model)
nlp.add_pipe(textcat)
#pt_model = LSTMSentiment(embedding_dim = 100, hidden_dim =100, vocab_size=259136, label_size=2, batch_size=3, dropout=0.5)
#model = PyTorchWrapper(pt_model)
#nlp = spacy.load('/home/ysun/pytorchprodigy/')
#textcat = TextCategorizer(nlp.vocab,model)
#nlp.add_pipe(textcat)
examples = DB.get_dataset(dataset)
labels = {eg['label'] for eg in examples}
labels = list(sorted(labels))
print(labels)
model = TextClassifier(nlp, labels, long_text=long_text,
low_data=len(examples) < 1000)
#log('RECIPE: Initialised TextClassifier with model {}'
# .format(input_model), model.nlp.meta)
if shuffle:
print("it's shuffling")
random.shuffle(examples)
else:
print("it's not shuffling")
if eval_id:
evals = DB.get_dataset(eval_id)
print_("Loaded {} evaluation examples from '{}'"
.format(len(evals), eval_id))
else:
examples, evals, eval_split = split_evals(examples, eval_split)
print_("Using {}% of examples ({}) for evaluation"
.format(round(eval_split * 100), len(evals)))
if shuffle:
random.shuffle(examples)
examples = examples[:int(len(examples) * factor)]
print_(printers.trainconf(dropout, n_iter, batch_size, factor,
len(examples)))
if len(evals) > 0:
print_(printers.tc_update_header())
best_acc = {'accuracy': 0}
best_model = None
if long_text:
examples = list(split_sentences(nlp, examples, min_length=False))
for i in range(n_iter):
loss = 0.
random.shuffle(examples)
for batch in cytoolz.partition_all(batch_size,
tqdm.tqdm(examples, leave=False)):
batch = list(batch)
loss += model.update(batch, revise=False, drop=dropout)
if len(evals) > 0:
with nlp.use_params(model.optimizer.averages):
acc = model.evaluate(tqdm.tqdm(evals, leave=False))
if acc['accuracy'] > best_acc['accuracy']:
best_acc = dict(acc)
best_model = nlp.to_bytes()
print_(printers.tc_update(i, loss, acc))
if len(evals) > 0:
print_(printers.tc_result(best_acc))
if output_model is not None:
if best_model is not None:
nlp = nlp.from_bytes(best_model)
msg = export_model_data(output_model, nlp, examples, evals)
print_(msg)
return best_acc['accuracy']