SUMIN/HuggingFace_transformer
1
0
Fork 0
Code Issues Pull Requests Actions 7 Packages Projects Releases Wiki Activity

Reformat source code with black.

Browse Source 
This is the result of:

    $ black --line-length 119 examples templates transformers utils hubconf.py setup.py

There's a lot of fairly long lines in the project. As a consequence, I'm
picking the longest widely accepted line length, 119 characters.

This is also Thomas' preference, because it allows for explicit variable
names, to make the code easier to understand.
This commit is contained in:
Aymeric Augustin
2019-12-21 15:46:46 +01:00
parent 63e3827c6b
commit fa84ae26d6
200 changed files with 17452 additions and 12594 deletions
Download Patch File Download Diff File Expand all files Collapse all files

37
examples/contrib/run_camembert.py
View File

@@ -10,38 +10,37 @@ from transformers.modeling_camembert import CamembertForMaskedLM
def fill_mask(masked_input, model, tokenizer, topk=5):
# Adapted from https://github.com/pytorch/fairseq/blob/master/fairseq/models/roberta/hub_interface.py
assert masked_input.count('<mask>') == 1
assert masked_input.count("<mask>") == 1
input_ids = torch.tensor(tokenizer.encode(masked_input, add_special_tokens=True)).unsqueeze(0) # Batch size 1
logits = model(input_ids)[0] # The last hidden-state is the first element of the output tuple
masked_index = (input_ids.squeeze() == tokenizer.mask_token_id).nonzero().item()
logits = logits[0, masked_index, :]
prob = logits.softmax(dim=0)
values, indices = prob.topk(k=topk, dim=0)
topk_predicted_token_bpe = ' '.join([tokenizer.convert_ids_to_tokens(indices[i].item())
for i in range(len(indices))])
topk_predicted_token_bpe = " ".join(
[tokenizer.convert_ids_to_tokens(indices[i].item()) for i in range(len(indices))]
)
masked_token = tokenizer.mask_token
topk_filled_outputs = []
for index, predicted_token_bpe in enumerate(topk_predicted_token_bpe.split(' ')):
predicted_token = predicted_token_bpe.replace('\u2581', ' ')
for index, predicted_token_bpe in enumerate(topk_predicted_token_bpe.split(" ")):
predicted_token = predicted_token_bpe.replace("\u2581", " ")
if " {0}".format(masked_token) in masked_input:
topk_filled_outputs.append((
masked_input.replace(
' {0}'.format(masked_token), predicted_token
),
values[index].item(),
predicted_token,
))
topk_filled_outputs.append(
(
masked_input.replace(" {0}".format(masked_token), predicted_token),
values[index].item(),
predicted_token,
)
)
else:
topk_filled_outputs.append((
masked_input.replace(masked_token, predicted_token),
values[index].item(),
predicted_token,
))
topk_filled_outputs.append(
(masked_input.replace(masked_token, predicted_token), values[index].item(), predicted_token,)
)
return topk_filled_outputs
tokenizer = CamembertTokenizer.from_pretrained('camembert-base')
model = CamembertForMaskedLM.from_pretrained('camembert-base')
tokenizer = CamembertTokenizer.from_pretrained("camembert-base")
model = CamembertForMaskedLM.from_pretrained("camembert-base")
model.eval()
masked_input = "Le camembert est <mask> :)"

163
examples/contrib/run_openai_gpt.py
View File

@@ -36,34 +36,42 @@ from tqdm import tqdm, trange
import numpy as np
import torch
from torch.utils.data import (DataLoader, RandomSampler, SequentialSampler,
TensorDataset)
from torch.utils.data import DataLoader, RandomSampler, SequentialSampler, TensorDataset
from transformers import (OpenAIGPTDoubleHeadsModel, OpenAIGPTTokenizer,
AdamW, cached_path, WEIGHTS_NAME, CONFIG_NAME,
get_linear_schedule_with_warmup)
from transformers import (
OpenAIGPTDoubleHeadsModel,
OpenAIGPTTokenizer,
AdamW,
cached_path,
WEIGHTS_NAME,
CONFIG_NAME,
get_linear_schedule_with_warmup,
)
ROCSTORIES_URL = "https://s3.amazonaws.com/datasets.huggingface.co/ROCStories.tar.gz"
logging.basicConfig(format = '%(asctime)s - %(levelname)s - %(name)s - %(message)s',
datefmt = '%m/%d/%Y %H:%M:%S',
level = logging.INFO)
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s", datefmt="%m/%d/%Y %H:%M:%S", level=logging.INFO
)
logger = logging.getLogger(__name__)
def accuracy(out, labels):
outputs = np.argmax(out, axis=1)
return np.sum(outputs == labels)
def load_rocstories_dataset(dataset_path):
""" Output a list of tuples(story, 1st continuation, 2nd continuation, label) """
with open(dataset_path, encoding='utf_8') as f:
with open(dataset_path, encoding="utf_8") as f:
f = csv.reader(f)
output = []
next(f) # skip the first line
next(f) # skip the first line
for line in tqdm(f):
output.append((' '.join(line[1:5]), line[5], line[6], int(line[-1])-1))
output.append((" ".join(line[1:5]), line[5], line[6], int(line[-1]) - 1))
return output
def pre_process_datasets(encoded_datasets, input_len, cap_length, start_token, delimiter_token, clf_token):
""" Pre-process datasets containing lists of tuples(story, 1st continuation, 2nd continuation, label)
@@ -80,56 +88,68 @@ def pre_process_datasets(encoded_datasets, input_len, cap_length, start_token, d
for i, (story, cont1, cont2, mc_label), in enumerate(dataset):
with_cont1 = [start_token] + story[:cap_length] + [delimiter_token] + cont1[:cap_length] + [clf_token]
with_cont2 = [start_token] + story[:cap_length] + [delimiter_token] + cont2[:cap_length] + [clf_token]
input_ids[i, 0, :len(with_cont1)] = with_cont1
input_ids[i, 1, :len(with_cont2)] = with_cont2
input_ids[i, 0, : len(with_cont1)] = with_cont1
input_ids[i, 1, : len(with_cont2)] = with_cont2
mc_token_ids[i, 0] = len(with_cont1) - 1
mc_token_ids[i, 1] = len(with_cont2) - 1
lm_labels[i, 0, :len(with_cont1)] = with_cont1
lm_labels[i, 1, :len(with_cont2)] = with_cont2
lm_labels[i, 0, : len(with_cont1)] = with_cont1
lm_labels[i, 1, : len(with_cont2)] = with_cont2
mc_labels[i] = mc_label
all_inputs = (input_ids, mc_token_ids, lm_labels, mc_labels)
tensor_datasets.append(tuple(torch.tensor(t) for t in all_inputs))
return tensor_datasets
def main():
parser = argparse.ArgumentParser()
parser.add_argument('--model_name', type=str, default='openai-gpt',
help='pretrained model name')
parser.add_argument("--do_train", action='store_true', help="Whether to run training.")
parser.add_argument("--do_eval", action='store_true', help="Whether to run eval on the dev set.")
parser.add_argument("--output_dir", default=None, type=str, required=True,
help="The output directory where the model predictions and checkpoints will be written.")
parser.add_argument('--train_dataset', type=str, default='')
parser.add_argument('--eval_dataset', type=str, default='')
parser.add_argument('--seed', type=int, default=42)
parser.add_argument('--num_train_epochs', type=int, default=3)
parser.add_argument('--train_batch_size', type=int, default=8)
parser.add_argument('--eval_batch_size', type=int, default=16)
parser.add_argument("--adam_epsilon", default=1e-8, type=float,
help="Epsilon for Adam optimizer.")
parser.add_argument('--max_grad_norm', type=int, default=1)
parser.add_argument("--max_steps", default=-1, type=int,
help="If > 0: set total number of training \
steps to perform. Override num_train_epochs.")
parser.add_argument('--gradient_accumulation_steps', type=int, default=1,
help="Number of updates steps to accumulate before\
performing a backward/update pass.")
parser.add_argument('--learning_rate', type=float, default=6.25e-5)
parser.add_argument("--warmup_steps", default=0, type=int,
help="Linear warmup over warmup_steps.")
parser.add_argument('--lr_schedule', type=str, default='warmup_linear')
parser.add_argument('--weight_decay', type=float, default=0.01)
parser.add_argument('--lm_coef', type=float, default=0.9)
parser.add_argument('--n_valid', type=int, default=374)
parser.add_argument("--model_name", type=str, default="openai-gpt", help="pretrained model name")
parser.add_argument("--do_train", action="store_true", help="Whether to run training.")
parser.add_argument("--do_eval", action="store_true", help="Whether to run eval on the dev set.")
parser.add_argument(
"--output_dir",
default=None,
type=str,
required=True,
help="The output directory where the model predictions and checkpoints will be written.",
)
parser.add_argument("--train_dataset", type=str, default="")
parser.add_argument("--eval_dataset", type=str, default="")
parser.add_argument("--seed", type=int, default=42)
parser.add_argument("--num_train_epochs", type=int, default=3)
parser.add_argument("--train_batch_size", type=int, default=8)
parser.add_argument("--eval_batch_size", type=int, default=16)
parser.add_argument("--adam_epsilon", default=1e-8, type=float, help="Epsilon for Adam optimizer.")
parser.add_argument("--max_grad_norm", type=int, default=1)
parser.add_argument(
"--max_steps",
default=-1,
type=int,
help="If > 0: set total number of training \
steps to perform. Override num_train_epochs.",
)
parser.add_argument(
"--gradient_accumulation_steps",
type=int,
default=1,
help="Number of updates steps to accumulate before\
performing a backward/update pass.",
)
parser.add_argument("--learning_rate", type=float, default=6.25e-5)
parser.add_argument("--warmup_steps", default=0, type=int, help="Linear warmup over warmup_steps.")
parser.add_argument("--lr_schedule", type=str, default="warmup_linear")
parser.add_argument("--weight_decay", type=float, default=0.01)
parser.add_argument("--lm_coef", type=float, default=0.9)
parser.add_argument("--n_valid", type=int, default=374)
parser.add_argument('--server_ip', type=str, default='', help="Can be used for distant debugging.")
parser.add_argument('--server_port', type=str, default='', help="Can be used for distant debugging.")
parser.add_argument("--server_ip", type=str, default="", help="Can be used for distant debugging.")
parser.add_argument("--server_port", type=str, default="", help="Can be used for distant debugging.")
args = parser.parse_args()
print(args)
if args.server_ip and args.server_port:
# Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
import ptvsd
print("Waiting for debugger attach")
ptvsd.enable_attach(address=(args.server_ip, args.server_port), redirect_output=True)
ptvsd.wait_for_attach()
@@ -152,7 +172,7 @@ def main():
# Load tokenizer and model
# This loading functions also add new tokens and embeddings called `special tokens`
# These new embeddings will be fine-tuned on the RocStories dataset
special_tokens = ['_start_', '_delimiter_', '_classify_']
special_tokens = ["_start_", "_delimiter_", "_classify_"]
tokenizer = OpenAIGPTTokenizer.from_pretrained(args.model_name)
tokenizer.add_tokens(special_tokens)
special_tokens_ids = tokenizer.convert_tokens_to_ids(special_tokens)
@@ -163,6 +183,7 @@ def main():
# Load and encode the datasets
if not args.train_dataset and not args.eval_dataset:
roc_stories = cached_path(ROCSTORIES_URL)
def tokenize_and_encode(obj):
""" Tokenize and encode a nested object """
if isinstance(obj, str):
@@ -170,6 +191,7 @@ def main():
elif isinstance(obj, int):
return obj
return list(tokenize_and_encode(o) for o in obj)
logger.info("Encoding dataset...")
train_dataset = load_rocstories_dataset(args.train_dataset)
eval_dataset = load_rocstories_dataset(args.eval_dataset)
@@ -178,8 +200,11 @@ def main():
# Compute the max input length for the Transformer
max_length = model.config.n_positions // 2 - 2
input_length = max(len(story[:max_length]) + max(len(cont1[:max_length]), len(cont2[:max_length])) + 3 \
for dataset in encoded_datasets for story, cont1, cont2, _ in dataset)
input_length = max(
len(story[:max_length]) + max(len(cont1[:max_length]), len(cont2[:max_length])) + 3
for dataset in encoded_datasets
for story, cont1, cont2, _ in dataset
)
input_length = min(input_length, model.config.n_positions) # Max size of input for the pre-trained model
# Prepare inputs tensors and dataloaders
@@ -198,20 +223,23 @@ def main():
if args.do_train:
if args.max_steps > 0:
t_total = args.max_steps
args.num_train_epochs = args.max_steps //\
(len(train_dataloader) // args.gradient_accumulation_steps) + 1
args.num_train_epochs = args.max_steps // (len(train_dataloader) // args.gradient_accumulation_steps) + 1
else:
t_total = len(train_dataloader)\
// args.gradient_accumulation_steps * args.num_train_epochs
t_total = len(train_dataloader) // args.gradient_accumulation_steps * args.num_train_epochs
param_optimizer = list(model.named_parameters())
no_decay = ['bias', 'LayerNorm.bias', 'LayerNorm.weight']
no_decay = ["bias", "LayerNorm.bias", "LayerNorm.weight"]
optimizer_grouped_parameters = [
{'params': [p for n, p in param_optimizer if not any(nd in n for nd in no_decay)], 'weight_decay': args.weight_decay},
{'params': [p for n, p in param_optimizer if any(nd in n for nd in no_decay)], 'weight_decay': 0.0}
]
{
"params": [p for n, p in param_optimizer if not any(nd in n for nd in no_decay)],
"weight_decay": args.weight_decay,
},
{"params": [p for n, p in param_optimizer if any(nd in n for nd in no_decay)], "weight_decay": 0.0},
]
optimizer = AdamW(optimizer_grouped_parameters, lr=args.learning_rate, eps=args.adam_epsilon)
scheduler = get_linear_schedule_with_warmup(optimizer, num_warmup_steps=args.warmup_steps, num_training_steps=t_total)
scheduler = get_linear_schedule_with_warmup(
optimizer, num_warmup_steps=args.warmup_steps, num_training_steps=t_total
)
if args.do_train:
nb_tr_steps, tr_loss, exp_average_loss = 0, 0, None
@@ -230,14 +258,16 @@ def main():
optimizer.step()
optimizer.zero_grad()
tr_loss += loss.item()
exp_average_loss = loss.item() if exp_average_loss is None else 0.7*exp_average_loss+0.3*loss.item()
exp_average_loss = (
loss.item() if exp_average_loss is None else 0.7 * exp_average_loss + 0.3 * loss.item()
)
nb_tr_steps += 1
tqdm_bar.desc = "Training loss: {:.2e} lr: {:.2e}".format(exp_average_loss, scheduler.get_lr()[0])
# Save a trained model
if args.do_train:
# Save a trained model, configuration and tokenizer
model_to_save = model.module if hasattr(model, 'module') else model # Only save the model itself
model_to_save = model.module if hasattr(model, "module") else model # Only save the model itself
# If we save using the predefined names, we can load using `from_pretrained`
output_model_file = os.path.join(args.output_dir, WEIGHTS_NAME)
@@ -260,10 +290,12 @@ def main():
batch = tuple(t.to(device) for t in batch)
input_ids, mc_token_ids, lm_labels, mc_labels = batch
with torch.no_grad():
_, mc_loss, _, mc_logits = model(input_ids, mc_token_ids=mc_token_ids, lm_labels=lm_labels, mc_labels=mc_labels)
_, mc_loss, _, mc_logits = model(
input_ids, mc_token_ids=mc_token_ids, lm_labels=lm_labels, mc_labels=mc_labels
)
mc_logits = mc_logits.detach().cpu().numpy()
mc_labels = mc_labels.to('cpu').numpy()
mc_labels = mc_labels.to("cpu").numpy()
tmp_eval_accuracy = accuracy(mc_logits, mc_labels)
eval_loss += mc_loss.mean().item()
@@ -274,10 +306,8 @@ def main():
eval_loss = eval_loss / nb_eval_steps
eval_accuracy = eval_accuracy / nb_eval_examples
train_loss = tr_loss/nb_tr_steps if args.do_train else None
result = {'eval_loss': eval_loss,
'eval_accuracy': eval_accuracy,
'train_loss': train_loss}
train_loss = tr_loss / nb_tr_steps if args.do_train else None
result = {"eval_loss": eval_loss, "eval_accuracy": eval_accuracy, "train_loss": train_loss}
output_eval_file = os.path.join(args.output_dir, "eval_results.txt")
with open(output_eval_file, "w") as writer:
@@ -286,5 +316,6 @@ def main():
logger.info(" %s = %s", key, str(result[key]))
writer.write("%s = %s\n" % (key, str(result[key])))
if __name__ == '__main__':
if __name__ == "__main__":
main()

450
examples/contrib/run_swag.py
View File

@@ -28,8 +28,7 @@ import glob
import numpy as np
import torch
from torch.utils.data import (DataLoader, RandomSampler, SequentialSampler,
TensorDataset)
from torch.utils.data import DataLoader, RandomSampler, SequentialSampler, TensorDataset
from torch.utils.data.distributed import DistributedSampler
try:
@@ -39,31 +38,23 @@ except:
from tqdm import tqdm, trange
from transformers import (WEIGHTS_NAME, BertConfig,
BertForMultipleChoice, BertTokenizer)
from transformers import WEIGHTS_NAME, BertConfig, BertForMultipleChoice, BertTokenizer
from transformers import AdamW, get_linear_schedule_with_warmup
logger = logging.getLogger(__name__)
ALL_MODELS = sum((tuple(conf.pretrained_config_archive_map.keys()) \
for conf in [BertConfig]), ())
ALL_MODELS = sum((tuple(conf.pretrained_config_archive_map.keys()) for conf in [BertConfig]), ())
MODEL_CLASSES = {
'bert': (BertConfig, BertForMultipleChoice, BertTokenizer),
"bert": (BertConfig, BertForMultipleChoice, BertTokenizer),
}
class SwagExample(object):
"""A single training/test example for the SWAG dataset."""
def __init__(self,
swag_id,
context_sentence,
start_ending,
ending_0,
ending_1,
ending_2,
ending_3,
label = None):
def __init__(self, swag_id, context_sentence, start_ending, ending_0, ending_1, ending_2, ending_3, label=None):
self.swag_id = swag_id
self.context_sentence = context_sentence
self.start_ending = start_ending
@@ -94,57 +85,49 @@ class SwagExample(object):
return ", ".join(l)
class InputFeatures(object):
def __init__(self,
example_id,
choices_features,
label
):
class InputFeatures(object):
def __init__(self, example_id, choices_features, label):
self.example_id = example_id
self.choices_features = [
{
'input_ids': input_ids,
'input_mask': input_mask,
'segment_ids': segment_ids
}
{"input_ids": input_ids, "input_mask": input_mask, "segment_ids": segment_ids}
for _, input_ids, input_mask, segment_ids in choices_features
]
self.label = label
def read_swag_examples(input_file, is_training=True):
with open(input_file, 'r', encoding='utf-8') as f:
with open(input_file, "r", encoding="utf-8") as f:
reader = csv.reader(f)
lines = []
for line in reader:
if sys.version_info[0] == 2:
line = list(unicode(cell, 'utf-8') for cell in line)
line = list(unicode(cell, "utf-8") for cell in line)
lines.append(line)
if is_training and lines[0][-1] != 'label':
raise ValueError(
"For training, the input file must contain a label column."
)
if is_training and lines[0][-1] != "label":
raise ValueError("For training, the input file must contain a label column.")
examples = [
SwagExample(
swag_id = line[2],
context_sentence = line[4],
start_ending = line[5], # in the swag dataset, the
# common beginning of each
# choice is stored in "sent2".
ending_0 = line[7],
ending_1 = line[8],
ending_2 = line[9],
ending_3 = line[10],
label = int(line[11]) if is_training else None
) for line in lines[1:] # we skip the line with the column names
swag_id=line[2],
context_sentence=line[4],
start_ending=line[5], # in the swag dataset, the
# common beginning of each
# choice is stored in "sent2".
ending_0=line[7],
ending_1=line[8],
ending_2=line[9],
ending_3=line[10],
label=int(line[11]) if is_training else None,
)
for line in lines[1:] # we skip the line with the column names
]
return examples
def convert_examples_to_features(examples, tokenizer, max_seq_length,
is_training):
def convert_examples_to_features(examples, tokenizer, max_seq_length, is_training):
"""Loads a data file into a list of `InputBatch`s."""
# Swag is a multiple choice task. To perform this task using Bert,
@@ -204,23 +187,18 @@ def convert_examples_to_features(examples, tokenizer, max_seq_length,
logger.info("swag_id: {}".format(example.swag_id))
for choice_idx, (tokens, input_ids, input_mask, segment_ids) in enumerate(choices_features):
logger.info("choice: {}".format(choice_idx))
logger.info("tokens: {}".format(' '.join(tokens)))
logger.info("input_ids: {}".format(' '.join(map(str, input_ids))))
logger.info("input_mask: {}".format(' '.join(map(str, input_mask))))
logger.info("segment_ids: {}".format(' '.join(map(str, segment_ids))))
logger.info("tokens: {}".format(" ".join(tokens)))
logger.info("input_ids: {}".format(" ".join(map(str, input_ids))))
logger.info("input_mask: {}".format(" ".join(map(str, input_mask))))
logger.info("segment_ids: {}".format(" ".join(map(str, segment_ids))))
if is_training:
logger.info("label: {}".format(label))
features.append(
InputFeatures(
example_id = example.swag_id,
choices_features = choices_features,
label = label
)
)
features.append(InputFeatures(example_id=example.swag_id, choices_features=choices_features, label=label))
return features
def _truncate_seq_pair(tokens_a, tokens_b, max_length):
"""Truncates a sequence pair in place to the maximum length."""
@@ -237,18 +215,14 @@ def _truncate_seq_pair(tokens_a, tokens_b, max_length):
else:
tokens_b.pop()
def accuracy(out, labels):
outputs = np.argmax(out, axis=1)
return np.sum(outputs == labels)
def select_field(features, field):
return [
[
choice[field]
for choice in feature.choices_features
]
for feature in features
]
return [[choice[field] for choice in feature.choices_features] for feature in features]
def set_seed(args):
@@ -258,24 +232,28 @@ def set_seed(args):
if args.n_gpu > 0:
torch.cuda.manual_seed_all(args.seed)
def load_and_cache_examples(args, tokenizer, evaluate=False, output_examples=False):
if args.local_rank not in [-1, 0]:
torch.distributed.barrier() # Make sure only the first process in distributed training process the dataset, and the others will use the cache
# Load data features from cache or dataset file
input_file = args.predict_file if evaluate else args.train_file
cached_features_file = os.path.join(os.path.dirname(input_file), 'cached_{}_{}_{}'.format(
'dev' if evaluate else 'train',
list(filter(None, args.model_name_or_path.split('/'))).pop(),
str(args.max_seq_length)))
cached_features_file = os.path.join(
os.path.dirname(input_file),
"cached_{}_{}_{}".format(
"dev" if evaluate else "train",
list(filter(None, args.model_name_or_path.split("/"))).pop(),
str(args.max_seq_length),
),
)
if os.path.exists(cached_features_file) and not args.overwrite_cache and not output_examples:
logger.info("Loading features from cached file %s", cached_features_file)
features = torch.load(cached_features_file)
else:
logger.info("Creating features from dataset file at %s", input_file)
examples = read_swag_examples(input_file)
features = convert_examples_to_features(
examples, tokenizer, args.max_seq_length, not evaluate)
features = convert_examples_to_features(examples, tokenizer, args.max_seq_length, not evaluate)
if args.local_rank in [-1, 0]:
logger.info("Saving features into cached file %s", cached_features_file)
@@ -285,21 +263,21 @@ def load_and_cache_examples(args, tokenizer, evaluate=False, output_examples=Fal
torch.distributed.barrier() # Make sure only the first process in distributed training process the dataset, and the others will use the cache
# Convert to Tensors and build dataset
all_input_ids = torch.tensor(select_field(features, 'input_ids'), dtype=torch.long)
all_input_mask = torch.tensor(select_field(features, 'input_mask'), dtype=torch.long)
all_segment_ids = torch.tensor(select_field(features, 'segment_ids'), dtype=torch.long)
all_input_ids = torch.tensor(select_field(features, "input_ids"), dtype=torch.long)
all_input_mask = torch.tensor(select_field(features, "input_mask"), dtype=torch.long)
all_segment_ids = torch.tensor(select_field(features, "segment_ids"), dtype=torch.long)
all_label = torch.tensor([f.label for f in features], dtype=torch.long)
if evaluate:
dataset = TensorDataset(all_input_ids, all_input_mask, all_segment_ids,
all_label)
dataset = TensorDataset(all_input_ids, all_input_mask, all_segment_ids, all_label)
else:
dataset = TensorDataset(all_input_ids, all_input_mask, all_segment_ids,
all_label)
dataset = TensorDataset(all_input_ids, all_input_mask, all_segment_ids, all_label)
if output_examples:
return dataset, examples, features
return dataset
def train(args, train_dataset, model, tokenizer):
""" Train the model """
if args.local_rank in [-1, 0]:
@@ -316,13 +294,18 @@ def train(args, train_dataset, model, tokenizer):
t_total = len(train_dataloader) // args.gradient_accumulation_steps * args.num_train_epochs
# Prepare optimizer and schedule (linear warmup and decay)
no_decay = ['bias', 'LayerNorm.weight']
no_decay = ["bias", "LayerNorm.weight"]
optimizer_grouped_parameters = [
{'params': [p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay)], 'weight_decay': args.weight_decay},
{'params': [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay)], 'weight_decay': 0.0}
]
{
"params": [p for n, p in model.named_parameters() if not any(nd in n for nd in no_decay)],
"weight_decay": args.weight_decay,
},
{"params": [p for n, p in model.named_parameters() if any(nd in n for nd in no_decay)], "weight_decay": 0.0},
]
optimizer = AdamW(optimizer_grouped_parameters, lr=args.learning_rate, eps=args.adam_epsilon)
scheduler = get_linear_schedule_with_warmup(optimizer, num_warmup_steps=args.warmup_steps, num_training_steps=t_total)
scheduler = get_linear_schedule_with_warmup(
optimizer, num_warmup_steps=args.warmup_steps, num_training_steps=t_total
)
if args.fp16:
try:
from apex import amp
@@ -336,17 +319,21 @@ def train(args, train_dataset, model, tokenizer):
# Distributed training (should be after apex fp16 initialization)
if args.local_rank != -1:
model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.local_rank],
output_device=args.local_rank,
find_unused_parameters=True)
model = torch.nn.parallel.DistributedDataParallel(
model, device_ids=[args.local_rank], output_device=args.local_rank, find_unused_parameters=True
)
# Train!
logger.info("***** Running training *****")
logger.info(" Num examples = %d", len(train_dataset))
logger.info(" Num Epochs = %d", args.num_train_epochs)
logger.info(" Instantaneous batch size per GPU = %d", args.per_gpu_train_batch_size)
logger.info(" Total train batch size (w. parallel, distributed & accumulation) = %d",
args.train_batch_size * args.gradient_accumulation_steps * (torch.distributed.get_world_size() if args.local_rank != -1 else 1))
logger.info(
" Total train batch size (w. parallel, distributed & accumulation) = %d",
args.train_batch_size
* args.gradient_accumulation_steps
* (torch.distributed.get_world_size() if args.local_rank != -1 else 1),
)
logger.info(" Gradient Accumulation steps = %d", args.gradient_accumulation_steps)
logger.info(" Total optimization steps = %d", t_total)
@@ -360,11 +347,13 @@ def train(args, train_dataset, model, tokenizer):
for step, batch in enumerate(epoch_iterator):
model.train()
batch = tuple(t.to(args.device) for t in batch)
inputs = {'input_ids': batch[0],
'attention_mask': batch[1],
#'token_type_ids': None if args.model_type == 'xlm' else batch[2],
'token_type_ids': batch[2],
'labels': batch[3]}
inputs = {
"input_ids": batch[0],
"attention_mask": batch[1],
#'token_type_ids': None if args.model_type == 'xlm' else batch[2],
"token_type_ids": batch[2],
"labels": batch[3],
}
# if args.model_type in ['xlnet', 'xlm']:
# inputs.update({'cls_index': batch[5],
# 'p_mask': batch[6]})
@@ -372,7 +361,7 @@ def train(args, train_dataset, model, tokenizer):
loss = outputs[0] # model outputs are always tuple in transformers (see doc)
if args.n_gpu > 1:
loss = loss.mean() # mean() to average on multi-gpu parallel (not distributed) training
loss = loss.mean() # mean() to average on multi-gpu parallel (not distributed) training
if args.gradient_accumulation_steps > 1:
loss = loss / args.gradient_accumulation_steps
@@ -393,23 +382,27 @@ def train(args, train_dataset, model, tokenizer):
if args.local_rank in [-1, 0] and args.logging_steps > 0 and global_step % args.logging_steps == 0:
# Log metrics
if args.local_rank == -1 and args.evaluate_during_training: # Only evaluate when single GPU otherwise metrics may not average well
if (
args.local_rank == -1 and args.evaluate_during_training
): # Only evaluate when single GPU otherwise metrics may not average well
results = evaluate(args, model, tokenizer)
for key, value in results.items():
tb_writer.add_scalar('eval_{}'.format(key), value, global_step)
tb_writer.add_scalar('lr', scheduler.get_lr()[0], global_step)
tb_writer.add_scalar('loss', (tr_loss - logging_loss)/args.logging_steps, global_step)
tb_writer.add_scalar("eval_{}".format(key), value, global_step)
tb_writer.add_scalar("lr", scheduler.get_lr()[0], global_step)
tb_writer.add_scalar("loss", (tr_loss - logging_loss) / args.logging_steps, global_step)
logging_loss = tr_loss
if args.local_rank in [-1, 0] and args.save_steps > 0 and global_step % args.save_steps == 0:
# Save model checkpoint
output_dir = os.path.join(args.output_dir, 'checkpoint-{}'.format(global_step))
output_dir = os.path.join(args.output_dir, "checkpoint-{}".format(global_step))
if not os.path.exists(output_dir):
os.makedirs(output_dir)
model_to_save = model.module if hasattr(model, 'module') else model # Take care of distributed/parallel training
model_to_save = (
model.module if hasattr(model, "module") else model
) # Take care of distributed/parallel training
model_to_save.save_pretrained(output_dir)
tokenizer.save_vocabulary(output_dir)
torch.save(args, os.path.join(output_dir, 'training_args.bin'))
torch.save(args, os.path.join(output_dir, "training_args.bin"))
logger.info("Saving model checkpoint to %s", output_dir)
if args.max_steps > 0 and global_step > args.max_steps:
@@ -424,6 +417,7 @@ def train(args, train_dataset, model, tokenizer):
return global_step, tr_loss / global_step
def evaluate(args, model, tokenizer, prefix=""):
dataset, examples, features = load_and_cache_examples(args, tokenizer, evaluate=True, output_examples=True)
@@ -440,7 +434,6 @@ def evaluate(args, model, tokenizer, prefix=""):
logger.info(" Num examples = %d", len(dataset))
logger.info(" Batch size = %d", args.eval_batch_size)
eval_loss, eval_accuracy = 0, 0
nb_eval_steps, nb_eval_examples = 0, 0
@@ -448,11 +441,13 @@ def evaluate(args, model, tokenizer, prefix=""):
model.eval()
batch = tuple(t.to(args.device) for t in batch)
with torch.no_grad():
inputs = {'input_ids': batch[0],
'attention_mask': batch[1],
# 'token_type_ids': None if args.model_type == 'xlm' else batch[2] # XLM don't use segment_ids
'token_type_ids': batch[2],
'labels': batch[3]}
inputs = {
"input_ids": batch[0],
"attention_mask": batch[1],
# 'token_type_ids': None if args.model_type == 'xlm' else batch[2] # XLM don't use segment_ids
"token_type_ids": batch[2],
"labels": batch[3],
}
# if args.model_type in ['xlnet', 'xlm']:
# inputs.update({'cls_index': batch[4],
@@ -462,17 +457,16 @@ def evaluate(args, model, tokenizer, prefix=""):
eval_loss += tmp_eval_loss.mean().item()
logits = logits.detach().cpu().numpy()
label_ids = inputs['labels'].to('cpu').numpy()
label_ids = inputs["labels"].to("cpu").numpy()
tmp_eval_accuracy = accuracy(logits, label_ids)
eval_accuracy += tmp_eval_accuracy
nb_eval_steps += 1
nb_eval_examples += inputs['input_ids'].size(0)
nb_eval_examples += inputs["input_ids"].size(0)
eval_loss = eval_loss / nb_eval_steps
eval_accuracy = eval_accuracy / nb_eval_examples
result = {'eval_loss': eval_loss,
'eval_accuracy': eval_accuracy}
result = {"eval_loss": eval_loss, "eval_accuracy": eval_accuracy}
output_eval_file = os.path.join(args.output_dir, "eval_results.txt")
with open(output_eval_file, "w") as writer:
@@ -483,92 +477,144 @@ def evaluate(args, model, tokenizer, prefix=""):
return result
def main():
parser = argparse.ArgumentParser()
## Required parameters
parser.add_argument("--train_file", default=None, type=str, required=True,
help="SWAG csv for training. E.g., train.csv")
parser.add_argument("--predict_file", default=None, type=str, required=True,
help="SWAG csv for predictions. E.g., val.csv or test.csv")
parser.add_argument("--model_type", default=None, type=str, required=True,
help="Model type selected in the list: " + ", ".join(MODEL_CLASSES.keys()))
parser.add_argument("--model_name_or_path", default=None, type=str, required=True,
help="Path to pre-trained model or shortcut name selected in the list: " + ", ".join(ALL_MODELS))
parser.add_argument("--output_dir", default=None, type=str, required=True,
help="The output directory where the model checkpoints and predictions will be written.")
parser.add_argument(
"--train_file", default=None, type=str, required=True, help="SWAG csv for training. E.g., train.csv"
)
parser.add_argument(
"--predict_file",
default=None,
type=str,
required=True,
help="SWAG csv for predictions. E.g., val.csv or test.csv",
)
parser.add_argument(
"--model_type",
default=None,
type=str,
required=True,
help="Model type selected in the list: " + ", ".join(MODEL_CLASSES.keys()),
)
parser.add_argument(
"--model_name_or_path",
default=None,
type=str,
required=True,
help="Path to pre-trained model or shortcut name selected in the list: " + ", ".join(ALL_MODELS),
)
parser.add_argument(
"--output_dir",
default=None,
type=str,
required=True,
help="The output directory where the model checkpoints and predictions will be written.",
)
## Other parameters
parser.add_argument("--config_name", default="", type=str,
help="Pretrained config name or path if not the same as model_name")
parser.add_argument("--tokenizer_name", default="", type=str,
help="Pretrained tokenizer name or path if not the same as model_name")
parser.add_argument("--max_seq_length", default=384, type=int,
help="The maximum total input sequence length after tokenization. Sequences "
"longer than this will be truncated, and sequences shorter than this will be padded.")
parser.add_argument("--do_train", action='store_true',
help="Whether to run training.")
parser.add_argument("--do_eval", action='store_true',
help="Whether to run eval on the dev set.")
parser.add_argument("--evaluate_during_training", action='store_true',
help="Rul evaluation during training at each logging step.")
parser.add_argument("--do_lower_case", action='store_true',
help="Set this flag if you are using an uncased model.")
parser.add_argument(
"--config_name", default="", type=str, help="Pretrained config name or path if not the same as model_name"
)
parser.add_argument(
"--tokenizer_name",
default="",
type=str,
help="Pretrained tokenizer name or path if not the same as model_name",
)
parser.add_argument(
"--max_seq_length",
default=384,
type=int,
help="The maximum total input sequence length after tokenization. Sequences "
"longer than this will be truncated, and sequences shorter than this will be padded.",
)
parser.add_argument("--do_train", action="store_true", help="Whether to run training.")
parser.add_argument("--do_eval", action="store_true", help="Whether to run eval on the dev set.")
parser.add_argument(
"--evaluate_during_training", action="store_true", help="Rul evaluation during training at each logging step."
)
parser.add_argument(
"--do_lower_case", action="store_true", help="Set this flag if you are using an uncased model."
)
parser.add_argument("--per_gpu_train_batch_size", default=8, type=int,
help="Batch size per GPU/CPU for training.")
parser.add_argument("--per_gpu_eval_batch_size", default=8, type=int,
help="Batch size per GPU/CPU for evaluation.")
parser.add_argument("--learning_rate", default=5e-5, type=float,
help="The initial learning rate for Adam.")
parser.add_argument('--gradient_accumulation_steps', type=int, default=1,
help="Number of updates steps to accumulate before performing a backward/update pass.")
parser.add_argument("--weight_decay", default=0.0, type=float,
help="Weight deay if we apply some.")
parser.add_argument("--adam_epsilon", default=1e-8, type=float,
help="Epsilon for Adam optimizer.")
parser.add_argument("--max_grad_norm", default=1.0, type=float,
help="Max gradient norm.")
parser.add_argument("--num_train_epochs", default=3.0, type=float,
help="Total number of training epochs to perform.")
parser.add_argument("--max_steps", default=-1, type=int,
help="If > 0: set total number of training steps to perform. Override num_train_epochs.")
parser.add_argument("--warmup_steps", default=0, type=int,
help="Linear warmup over warmup_steps.")
parser.add_argument("--per_gpu_train_batch_size", default=8, type=int, help="Batch size per GPU/CPU for training.")
parser.add_argument(
"--per_gpu_eval_batch_size", default=8, type=int, help="Batch size per GPU/CPU for evaluation."
)
parser.add_argument("--learning_rate", default=5e-5, type=float, help="The initial learning rate for Adam.")
parser.add_argument(
"--gradient_accumulation_steps",
type=int,
default=1,
help="Number of updates steps to accumulate before performing a backward/update pass.",
)
parser.add_argument("--weight_decay", default=0.0, type=float, help="Weight deay if we apply some.")
parser.add_argument("--adam_epsilon", default=1e-8, type=float, help="Epsilon for Adam optimizer.")
parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
parser.add_argument(
"--num_train_epochs", default=3.0, type=float, help="Total number of training epochs to perform."
)
parser.add_argument(
"--max_steps",
default=-1,
type=int,
help="If > 0: set total number of training steps to perform. Override num_train_epochs.",
)
parser.add_argument("--warmup_steps", default=0, type=int, help="Linear warmup over warmup_steps.")
parser.add_argument('--logging_steps', type=int, default=50,
help="Log every X updates steps.")
parser.add_argument('--save_steps', type=int, default=50,
help="Save checkpoint every X updates steps.")
parser.add_argument("--eval_all_checkpoints", action='store_true',
help="Evaluate all checkpoints starting with the same prefix as model_name ending and ending with step number")
parser.add_argument("--no_cuda", action='store_true',
help="Whether not to use CUDA when available")
parser.add_argument('--overwrite_output_dir', action='store_true',
help="Overwrite the content of the output directory")
parser.add_argument('--overwrite_cache', action='store_true',
help="Overwrite the cached training and evaluation sets")
parser.add_argument('--seed', type=int, default=42,
help="random seed for initialization")
parser.add_argument("--logging_steps", type=int, default=50, help="Log every X updates steps.")
parser.add_argument("--save_steps", type=int, default=50, help="Save checkpoint every X updates steps.")
parser.add_argument(
"--eval_all_checkpoints",
action="store_true",
help="Evaluate all checkpoints starting with the same prefix as model_name ending and ending with step number",
)
parser.add_argument("--no_cuda", action="store_true", help="Whether not to use CUDA when available")
parser.add_argument(
"--overwrite_output_dir", action="store_true", help="Overwrite the content of the output directory"
)
parser.add_argument(
"--overwrite_cache", action="store_true", help="Overwrite the cached training and evaluation sets"
)
parser.add_argument("--seed", type=int, default=42, help="random seed for initialization")
parser.add_argument("--local_rank", type=int, default=-1,
help="local_rank for distributed training on gpus")
parser.add_argument('--fp16', action='store_true',
help="Whether to use 16-bit (mixed) precision (through NVIDIA apex) instead of 32-bit")
parser.add_argument('--fp16_opt_level', type=str, default='O1',
help="For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
"See details at https://nvidia.github.io/apex/amp.html")
parser.add_argument('--server_ip', type=str, default='', help="Can be used for distant debugging.")
parser.add_argument('--server_port', type=str, default='', help="Can be used for distant debugging.")
parser.add_argument("--local_rank", type=int, default=-1, help="local_rank for distributed training on gpus")
parser.add_argument(
"--fp16",
action="store_true",
help="Whether to use 16-bit (mixed) precision (through NVIDIA apex) instead of 32-bit",
)
parser.add_argument(
"--fp16_opt_level",
type=str,
default="O1",
help="For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']."
"See details at https://nvidia.github.io/apex/amp.html",
)
parser.add_argument("--server_ip", type=str, default="", help="Can be used for distant debugging.")
parser.add_argument("--server_port", type=str, default="", help="Can be used for distant debugging.")
args = parser.parse_args()
if os.path.exists(args.output_dir) and os.listdir(args.output_dir) and args.do_train and not args.overwrite_output_dir:
raise ValueError("Output directory ({}) already exists and is not empty. Use --overwrite_output_dir to overcome.".format(args.output_dir))
if (
os.path.exists(args.output_dir)
and os.listdir(args.output_dir)
and args.do_train
and not args.overwrite_output_dir
):
raise ValueError(
"Output directory ({}) already exists and is not empty. Use --overwrite_output_dir to overcome.".format(
args.output_dir
)
)
# Setup distant debugging if needed
if args.server_ip and args.server_port:
# Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
import ptvsd
print("Waiting for debugger attach")
ptvsd.enable_attach(address=(args.server_ip, args.server_port), redirect_output=True)
ptvsd.wait_for_attach()
@@ -580,16 +626,24 @@ def main():
else: # Initializes the distributed backend which will take care of sychronizing nodes/GPUs
torch.cuda.set_device(args.local_rank)
device = torch.device("cuda", args.local_rank)
torch.distributed.init_process_group(backend='nccl')
torch.distributed.init_process_group(backend="nccl")
args.n_gpu = 1
args.device = device
# Setup logging
logging.basicConfig(format = '%(asctime)s - %(levelname)s - %(name)s - %(message)s',
datefmt = '%m/%d/%Y %H:%M:%S',
level = logging.INFO if args.local_rank in [-1, 0] else logging.WARN)
logger.warning("Process rank: %s, device: %s, n_gpu: %s, distributed training: %s, 16-bits training: %s",
args.local_rank, device, args.n_gpu, bool(args.local_rank != -1), args.fp16)
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
datefmt="%m/%d/%Y %H:%M:%S",
level=logging.INFO if args.local_rank in [-1, 0] else logging.WARN,
)
logger.warning(
"Process rank: %s, device: %s, n_gpu: %s, distributed training: %s, 16-bits training: %s",
args.local_rank,
device,
args.n_gpu,
bool(args.local_rank != -1),
args.fp16,
)
# Set seed
set_seed(args)
@@ -601,8 +655,12 @@ def main():
args.model_type = args.model_type.lower()
config_class, model_class, tokenizer_class = MODEL_CLASSES[args.model_type]
config = config_class.from_pretrained(args.config_name if args.config_name else args.model_name_or_path)
tokenizer = tokenizer_class.from_pretrained(args.tokenizer_name if args.tokenizer_name else args.model_name_or_path, do_lower_case=args.do_lower_case)
model = model_class.from_pretrained(args.model_name_or_path, from_tf=bool('.ckpt' in args.model_name_or_path), config=config)
tokenizer = tokenizer_class.from_pretrained(
args.tokenizer_name if args.tokenizer_name else args.model_name_or_path, do_lower_case=args.do_lower_case
)
model = model_class.from_pretrained(
args.model_name_or_path, from_tf=bool(".ckpt" in args.model_name_or_path), config=config
)
if args.local_rank == 0:
torch.distributed.barrier() # Make sure only the first process in distributed training will download model & vocab
@@ -617,7 +675,6 @@ def main():
global_step, tr_loss = train(args, train_dataset, model, tokenizer)
logger.info(" global_step = %s, average loss = %s", global_step, tr_loss)
# Save the trained model and the tokenizer
if args.local_rank == -1 or torch.distributed.get_rank() == 0:
# Create output directory if needed
@@ -627,19 +684,20 @@ def main():
logger.info("Saving model checkpoint to %s", args.output_dir)
# Save a trained model, configuration and tokenizer using `save_pretrained()`.
# They can then be reloaded using `from_pretrained()`
model_to_save = model.module if hasattr(model, 'module') else model # Take care of distributed/parallel training
model_to_save = (
model.module if hasattr(model, "module") else model
) # Take care of distributed/parallel training
model_to_save.save_pretrained(args.output_dir)
tokenizer.save_pretrained(args.output_dir)
# Good practice: save your training arguments together with the trained model
torch.save(args, os.path.join(args.output_dir, 'training_args.bin'))
torch.save(args, os.path.join(args.output_dir, "training_args.bin"))
# Load a trained model and vocabulary that you have fine-tuned
model = model_class.from_pretrained(args.output_dir)
tokenizer = tokenizer_class.from_pretrained(args.output_dir)
model.to(args.device)
# Evaluation - we can ask to evaluate all the checkpoints (sub-directories) in a directory
results = {}
if args.do_eval and args.local_rank in [-1, 0]:
@@ -650,14 +708,16 @@ def main():
checkpoints = [args.model_name_or_path]
if args.eval_all_checkpoints:
checkpoints = list(os.path.dirname(c) for c in sorted(glob.glob(args.output_dir + '/**/' + WEIGHTS_NAME, recursive=True)))
checkpoints = list(
os.path.dirname(c) for c in sorted(glob.glob(args.output_dir + "/**/" + WEIGHTS_NAME, recursive=True))
)
logging.getLogger("transformers.modeling_utils").setLevel(logging.WARN) # Reduce model loading logs
logger.info("Evaluate the following checkpoints: %s", checkpoints)
for checkpoint in checkpoints:
# Reload the model
global_step = checkpoint.split('-')[-1] if len(checkpoints) > 1 else ""
global_step = checkpoint.split("-")[-1] if len(checkpoints) > 1 else ""
model = model_class.from_pretrained(checkpoint)
tokenizer = tokenizer_class.from_pretrained(checkpoint)
model.to(args.device)
@@ -665,7 +725,7 @@ def main():
# Evaluate
result = evaluate(args, model, tokenizer, prefix=global_step)
result = dict((k + ('_{}'.format(global_step) if global_step else ''), v) for k, v in result.items())
result = dict((k + ("_{}".format(global_step) if global_step else ""), v) for k, v in result.items())
results.update(result)
logger.info("Results: {}".format(results))

92
examples/contrib/run_transfo_xl.py
View File

@@ -30,44 +30,36 @@ import torch
from transformers import TransfoXLLMHeadModel, TransfoXLCorpus, TransfoXLTokenizer
logging.basicConfig(format = '%(asctime)s - %(levelname)s - %(name)s - %(message)s',
datefmt = '%m/%d/%Y %H:%M:%S',
level = logging.INFO)
logging.basicConfig(
format="%(asctime)s - %(levelname)s - %(name)s - %(message)s", datefmt="%m/%d/%Y %H:%M:%S", level=logging.INFO
)
logger = logging.getLogger(__name__)
def main():
parser = argparse.ArgumentParser(description='PyTorch Transformer Language Model')
parser.add_argument('--model_name', type=str, default='transfo-xl-wt103',
help='pretrained model name')
parser.add_argument('--split', type=str, default='test',
choices=['all', 'valid', 'test'],
help='which split to evaluate')
parser.add_argument('--batch_size', type=int, default=10,
help='batch size')
parser.add_argument('--tgt_len', type=int, default=128,
help='number of tokens to predict')
parser.add_argument('--ext_len', type=int, default=0,
help='length of the extended context')
parser.add_argument('--mem_len', type=int, default=1600,
help='length of the retained previous heads')
parser.add_argument('--clamp_len', type=int, default=1000,
help='max positional embedding index')
parser.add_argument('--no_cuda', action='store_true',
help='Do not use CUDA even though CUA is available')
parser.add_argument('--work_dir', type=str, required=True,
help='path to the work_dir')
parser.add_argument('--no_log', action='store_true',
help='do not log the eval result')
parser.add_argument('--same_length', action='store_true',
help='set same length attention with masking')
parser.add_argument('--server_ip', type=str, default='', help="Can be used for distant debugging.")
parser.add_argument('--server_port', type=str, default='', help="Can be used for distant debugging.")
parser = argparse.ArgumentParser(description="PyTorch Transformer Language Model")
parser.add_argument("--model_name", type=str, default="transfo-xl-wt103", help="pretrained model name")
parser.add_argument(
"--split", type=str, default="test", choices=["all", "valid", "test"], help="which split to evaluate"
)
parser.add_argument("--batch_size", type=int, default=10, help="batch size")
parser.add_argument("--tgt_len", type=int, default=128, help="number of tokens to predict")
parser.add_argument("--ext_len", type=int, default=0, help="length of the extended context")
parser.add_argument("--mem_len", type=int, default=1600, help="length of the retained previous heads")
parser.add_argument("--clamp_len", type=int, default=1000, help="max positional embedding index")
parser.add_argument("--no_cuda", action="store_true", help="Do not use CUDA even though CUA is available")
parser.add_argument("--work_dir", type=str, required=True, help="path to the work_dir")
parser.add_argument("--no_log", action="store_true", help="do not log the eval result")
parser.add_argument("--same_length", action="store_true", help="set same length attention with masking")
parser.add_argument("--server_ip", type=str, default="", help="Can be used for distant debugging.")
parser.add_argument("--server_port", type=str, default="", help="Can be used for distant debugging.")
args = parser.parse_args()
assert args.ext_len >= 0, 'extended context length must be non-negative'
assert args.ext_len >= 0, "extended context length must be non-negative"
if args.server_ip and args.server_port:
# Distant debugging - see https://code.visualstudio.com/docs/python/debugging#_attach-to-a-local-script
import ptvsd
print("Waiting for debugger attach")
ptvsd.enable_attach(address=(args.server_ip, args.server_port), redirect_output=True)
ptvsd.wait_for_attach()
@@ -84,17 +76,18 @@ def main():
corpus = TransfoXLCorpus.from_pretrained(args.model_name)
ntokens = len(corpus.vocab)
va_iter = corpus.get_iterator('valid', args.batch_size, args.tgt_len,
device=device, ext_len=args.ext_len)
te_iter = corpus.get_iterator('test', args.batch_size, args.tgt_len,
device=device, ext_len=args.ext_len)
va_iter = corpus.get_iterator("valid", args.batch_size, args.tgt_len, device=device, ext_len=args.ext_len)
te_iter = corpus.get_iterator("test", args.batch_size, args.tgt_len, device=device, ext_len=args.ext_len)
# Load a pre-trained model
model = TransfoXLLMHeadModel.from_pretrained(args.model_name)
model = model.to(device)
logger.info('Evaluating with bsz {} tgt_len {} ext_len {} mem_len {} clamp_len {}'.format(
args.batch_size, args.tgt_len, args.ext_len, args.mem_len, args.clamp_len))
logger.info(
"Evaluating with bsz {} tgt_len {} ext_len {} mem_len {} clamp_len {}".format(
args.batch_size, args.tgt_len, args.ext_len, args.mem_len, args.clamp_len
)
)
model.reset_length(args.tgt_len, args.ext_len, args.mem_len)
if args.clamp_len > 0:
@@ -108,7 +101,7 @@ def main():
def evaluate(eval_iter):
# Turn on evaluation mode which disables dropout.
model.eval()
total_len, total_loss = 0, 0.
total_len, total_loss = 0, 0.0
start_time = time.time()
with torch.no_grad():
mems = None
@@ -119,35 +112,34 @@ def main():
total_loss += seq_len * loss.item()
total_len += seq_len
total_time = time.time() - start_time
logger.info('Time : {:.2f}s, {:.2f}ms/segment'.format(
total_time, 1000 * total_time / (idx+1)))
logger.info("Time : {:.2f}s, {:.2f}ms/segment".format(total_time, 1000 * total_time / (idx + 1)))
return total_loss / total_len
# Run on test data.
if args.split == 'all':
if args.split == "all":
test_loss = evaluate(te_iter)
valid_loss = evaluate(va_iter)
elif args.split == 'valid':
elif args.split == "valid":
valid_loss = evaluate(va_iter)
test_loss = None
elif args.split == 'test':
elif args.split == "test":
test_loss = evaluate(te_iter)
valid_loss = None
def format_log(loss, split):
log_str = '| {0} loss {1:5.2f} | {0} ppl {2:9.3f} '.format(
split, loss, math.exp(loss))
log_str = "| {0} loss {1:5.2f} | {0} ppl {2:9.3f} ".format(split, loss, math.exp(loss))
return log_str
log_str = ''
log_str = ""
if valid_loss is not None:
log_str += format_log(valid_loss, 'valid')
log_str += format_log(valid_loss, "valid")
if test_loss is not None:
log_str += format_log(test_loss, 'test')
log_str += format_log(test_loss, "test")
logger.info('=' * 100)
logger.info("=" * 100)
logger.info(log_str)
logger.info('=' * 100)
logger.info("=" * 100)
if __name__ == '__main__':
if __name__ == "__main__":
main()