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

573
templates/adding_a_new_example_script/run_xxx.py
View File

@@ -24,8 +24,7 @@ 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:
@@ -35,19 +34,32 @@ except:
from tqdm import tqdm, trange
from transformers import (WEIGHTS_NAME, BertConfig,
BertForQuestionAnswering, BertTokenizer,
XLMConfig, XLMForQuestionAnswering,
XLMTokenizer, XLNetConfig,
XLNetForQuestionAnswering,
XLNetTokenizer,
DistilBertConfig, DistilBertForQuestionAnswering, DistilBertTokenizer)
from transformers import (
WEIGHTS_NAME,
BertConfig,
BertForQuestionAnswering,
BertTokenizer,
XLMConfig,
XLMForQuestionAnswering,
XLMTokenizer,
XLNetConfig,
XLNetForQuestionAnswering,
XLNetTokenizer,
DistilBertConfig,
DistilBertForQuestionAnswering,
DistilBertTokenizer,
)
from transformers import AdamW, get_linear_schedule_with_warmup
from utils_squad import (read_squad_examples, convert_examples_to_features,
RawResult, write_predictions,
RawResultExtended, write_predictions_extended)
from utils_squad import (
read_squad_examples,
convert_examples_to_features,
RawResult,
write_predictions,
RawResultExtended,
write_predictions_extended,
)
# The follwing import is the official SQuAD evaluation script (2.0).
# You can remove it from the dependencies if you are using this script outside of the library
@@ -56,16 +68,18 @@ from utils_squad_evaluate import EVAL_OPTS, main as evaluate_on_squad
logger = logging.getLogger(__name__)
ALL_MODELS = sum((tuple(conf.pretrained_config_archive_map.keys()) \
for conf in (BertConfig, XLNetConfig, XLMConfig)), ())
ALL_MODELS = sum(
(tuple(conf.pretrained_config_archive_map.keys()) for conf in (BertConfig, XLNetConfig, XLMConfig)), ()
)
MODEL_CLASSES = {
'bert': (BertConfig, BertForQuestionAnswering, BertTokenizer),
'xlnet': (XLNetConfig, XLNetForQuestionAnswering, XLNetTokenizer),
'xlm': (XLMConfig, XLMForQuestionAnswering, XLMTokenizer),
'distilbert': (DistilBertConfig, DistilBertForQuestionAnswering, DistilBertTokenizer)
"bert": (BertConfig, BertForQuestionAnswering, BertTokenizer),
"xlnet": (XLNetConfig, XLNetForQuestionAnswering, XLNetTokenizer),
"xlm": (XLMConfig, XLMForQuestionAnswering, XLMTokenizer),
"distilbert": (DistilBertConfig, DistilBertForQuestionAnswering, DistilBertTokenizer),
}
def set_seed(args):
random.seed(args.seed)
np.random.seed(args.seed)
@@ -73,9 +87,11 @@ def set_seed(args):
if args.n_gpu > 0:
torch.cuda.manual_seed_all(args.seed)
def to_list(tensor):
return tensor.detach().cpu().tolist()
def train(args, train_dataset, model, tokenizer):
""" Train the model """
if args.local_rank in [-1, 0]:
@@ -92,13 +108,18 @@ 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
@@ -112,17 +133,21 @@ 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)
@@ -136,20 +161,21 @@ 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],
'start_positions': batch[3],
'end_positions': batch[4]}
if args.model_type != 'distilbert':
inputs['token_type_ids'] = None if args.model_type == 'xlm' else batch[2]
if args.model_type in ['xlnet', 'xlm']:
inputs.update({'cls_index': batch[5],
'p_mask': batch[6]})
inputs = {
"input_ids": batch[0],
"attention_mask": batch[1],
"start_positions": batch[3],
"end_positions": batch[4],
}
if args.model_type != "distilbert":
inputs["token_type_ids"] = None if args.model_type == "xlm" else batch[2]
if args.model_type in ["xlnet", "xlm"]:
inputs.update({"cls_index": batch[5], "p_mask": batch[6]})
outputs = model(**inputs)
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
@@ -173,22 +199,26 @@ 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)
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:
@@ -224,32 +254,31 @@ 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]
}
if args.model_type != 'distilbert':
inputs['token_type_ids'] = None if args.model_type == 'xlm' else batch[2] # XLM don't use segment_ids
inputs = {"input_ids": batch[0], "attention_mask": batch[1]}
if args.model_type != "distilbert":
inputs["token_type_ids"] = None if args.model_type == "xlm" else batch[2] # XLM don't use segment_ids
example_indices = batch[3]
if args.model_type in ['xlnet', 'xlm']:
inputs.update({'cls_index': batch[4],
'p_mask': batch[5]})
if args.model_type in ["xlnet", "xlm"]:
inputs.update({"cls_index": batch[4], "p_mask": batch[5]})
outputs = model(**inputs)
for i, example_index in enumerate(example_indices):
eval_feature = features[example_index.item()]
unique_id = int(eval_feature.unique_id)
if args.model_type in ['xlnet', 'xlm']:
if args.model_type in ["xlnet", "xlm"]:
# XLNet uses a more complex post-processing procedure
result = RawResultExtended(unique_id = unique_id,
start_top_log_probs = to_list(outputs[0][i]),
start_top_index = to_list(outputs[1][i]),
end_top_log_probs = to_list(outputs[2][i]),
end_top_index = to_list(outputs[3][i]),
cls_logits = to_list(outputs[4][i]))
result = RawResultExtended(
unique_id=unique_id,
start_top_log_probs=to_list(outputs[0][i]),
start_top_index=to_list(outputs[1][i]),
end_top_log_probs=to_list(outputs[2][i]),
end_top_index=to_list(outputs[3][i]),
cls_logits=to_list(outputs[4][i]),
)
else:
result = RawResult(unique_id = unique_id,
start_logits = to_list(outputs[0][i]),
end_logits = to_list(outputs[1][i]))
result = RawResult(
unique_id=unique_id, start_logits=to_list(outputs[0][i]), end_logits=to_list(outputs[1][i])
)
all_results.append(result)
# Compute predictions
@@ -260,23 +289,44 @@ def evaluate(args, model, tokenizer, prefix=""):
else:
output_null_log_odds_file = None
if args.model_type in ['xlnet', 'xlm']:
if args.model_type in ["xlnet", "xlm"]:
# XLNet uses a more complex post-processing procedure
write_predictions_extended(examples, features, all_results, args.n_best_size,
args.max_answer_length, output_prediction_file,
output_nbest_file, output_null_log_odds_file, args.predict_file,
model.config.start_n_top, model.config.end_n_top,
args.version_2_with_negative, tokenizer, args.verbose_logging)
write_predictions_extended(
examples,
features,
all_results,
args.n_best_size,
args.max_answer_length,
output_prediction_file,
output_nbest_file,
output_null_log_odds_file,
args.predict_file,
model.config.start_n_top,
model.config.end_n_top,
args.version_2_with_negative,
tokenizer,
args.verbose_logging,
)
else:
write_predictions(examples, features, all_results, args.n_best_size,
args.max_answer_length, args.do_lower_case, output_prediction_file,
output_nbest_file, output_null_log_odds_file, args.verbose_logging,
args.version_2_with_negative, args.null_score_diff_threshold)
write_predictions(
examples,
features,
all_results,
args.n_best_size,
args.max_answer_length,
args.do_lower_case,
output_prediction_file,
output_nbest_file,
output_null_log_odds_file,
args.verbose_logging,
args.version_2_with_negative,
args.null_score_diff_threshold,
)
# Evaluate with the official SQuAD script
evaluate_options = EVAL_OPTS(data_file=args.predict_file,
pred_file=output_prediction_file,
na_prob_file=output_null_log_odds_file)
evaluate_options = EVAL_OPTS(
data_file=args.predict_file, pred_file=output_prediction_file, na_prob_file=output_null_log_odds_file
)
results = evaluate_on_squad(evaluate_options)
return results
@@ -287,24 +337,30 @@ def load_and_cache_examples(args, tokenizer, evaluate=False, output_examples=Fal
# 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_squad_examples(input_file=input_file,
is_training=not evaluate,
version_2_with_negative=args.version_2_with_negative)
features = convert_examples_to_features(examples=examples,
tokenizer=tokenizer,
max_seq_length=args.max_seq_length,
doc_stride=args.doc_stride,
max_query_length=args.max_query_length,
is_training=not evaluate)
examples = read_squad_examples(
input_file=input_file, is_training=not evaluate, version_2_with_negative=args.version_2_with_negative
)
features = convert_examples_to_features(
examples=examples,
tokenizer=tokenizer,
max_seq_length=args.max_seq_length,
doc_stride=args.doc_stride,
max_query_length=args.max_query_length,
is_training=not evaluate,
)
if args.local_rank in [-1, 0]:
logger.info("Saving features into cached file %s", cached_features_file)
torch.save(features, cached_features_file)
@@ -320,14 +376,21 @@ def load_and_cache_examples(args, tokenizer, evaluate=False, output_examples=Fal
all_p_mask = torch.tensor([f.p_mask for f in features], dtype=torch.float)
if evaluate:
all_example_index = torch.arange(all_input_ids.size(0), dtype=torch.long)
dataset = TensorDataset(all_input_ids, all_input_mask, all_segment_ids,
all_example_index, all_cls_index, all_p_mask)
dataset = TensorDataset(
all_input_ids, all_input_mask, all_segment_ids, all_example_index, all_cls_index, all_p_mask
)
else:
all_start_positions = torch.tensor([f.start_position for f in features], dtype=torch.long)
all_end_positions = torch.tensor([f.end_position for f in features], dtype=torch.long)
dataset = TensorDataset(all_input_ids, all_input_mask, all_segment_ids,
all_start_positions, all_end_positions,
all_cls_index, all_p_mask)
dataset = TensorDataset(
all_input_ids,
all_input_mask,
all_segment_ids,
all_start_positions,
all_end_positions,
all_cls_index,
all_p_mask,
)
if output_examples:
return dataset, examples, features
@@ -338,109 +401,190 @@ def main():
parser = argparse.ArgumentParser()
## Required parameters
parser.add_argument("--train_file", default=None, type=str, required=True,
help="SQuAD json for training. E.g., train-v1.1.json")
parser.add_argument("--predict_file", default=None, type=str, required=True,
help="SQuAD json for predictions. E.g., dev-v1.1.json or test-v1.1.json")
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="SQuAD json for training. E.g., train-v1.1.json"
)
parser.add_argument(
"--predict_file",
default=None,
type=str,
required=True,
help="SQuAD json for predictions. E.g., dev-v1.1.json or test-v1.1.json",
)
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("--cache_dir", default="", type=str,
help="Where do you want to store the pre-trained models downloaded from s3")
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(
"--cache_dir",
default="",
type=str,
help="Where do you want to store the pre-trained models downloaded from s3",
)
parser.add_argument('--version_2_with_negative', action='store_true',
help='If true, the SQuAD examples contain some that do not have an answer.')
parser.add_argument('--null_score_diff_threshold', type=float, default=0.0,
help="If null_score - best_non_null is greater than the threshold predict null.")
parser.add_argument(
"--version_2_with_negative",
action="store_true",
help="If true, the SQuAD examples contain some that do not have an answer.",
)
parser.add_argument(
"--null_score_diff_threshold",
type=float,
default=0.0,
help="If null_score - best_non_null is greater than the threshold predict null.",
)
parser.add_argument("--max_seq_length", default=384, type=int,
help="The maximum total input sequence length after WordPiece tokenization. Sequences "
"longer than this will be truncated, and sequences shorter than this will be padded.")
parser.add_argument("--doc_stride", default=128, type=int,
help="When splitting up a long document into chunks, how much stride to take between chunks.")
parser.add_argument("--max_query_length", default=64, type=int,
help="The maximum number of tokens for the question. Questions longer than this will "
"be truncated to this length.")
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(
"--max_seq_length",
default=384,
type=int,
help="The maximum total input sequence length after WordPiece tokenization. Sequences "
"longer than this will be truncated, and sequences shorter than this will be padded.",
)
parser.add_argument(
"--doc_stride",
default=128,
type=int,
help="When splitting up a long document into chunks, how much stride to take between chunks.",
)
parser.add_argument(
"--max_query_length",
default=64,
type=int,
help="The maximum number of tokens for the question. Questions longer than this will "
"be truncated to this length.",
)
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("--n_best_size", default=20, type=int,
help="The total number of n-best predictions to generate in the nbest_predictions.json output file.")
parser.add_argument("--max_answer_length", default=30, type=int,
help="The maximum length of an answer that can be generated. This is needed because the start "
"and end predictions are not conditioned on one another.")
parser.add_argument("--verbose_logging", action='store_true',
help="If true, all of the warnings related to data processing will be printed. "
"A number of warnings are expected for a normal SQuAD evaluation.")
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(
"--n_best_size",
default=20,
type=int,
help="The total number of n-best predictions to generate in the nbest_predictions.json output file.",
)
parser.add_argument(
"--max_answer_length",
default=30,
type=int,
help="The maximum length of an answer that can be generated. This is needed because the start "
"and end predictions are not conditioned on one another.",
)
parser.add_argument(
"--verbose_logging",
action="store_true",
help="If true, all of the warnings related to data processing will be printed. "
"A number of warnings are expected for a normal SQuAD evaluation.",
)
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()
@@ -452,16 +596,24 @@ 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)
@@ -472,15 +624,21 @@ 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,
cache_dir=args.cache_dir if args.cache_dir else None)
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,
cache_dir=args.cache_dir if args.cache_dir else None)
model = model_class.from_pretrained(args.model_name_or_path,
from_tf=bool('.ckpt' in args.model_name_or_path),
config=config,
cache_dir=args.cache_dir if args.cache_dir else None)
config = config_class.from_pretrained(
args.config_name if args.config_name else args.model_name_or_path,
cache_dir=args.cache_dir if args.cache_dir else None,
)
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,
cache_dir=args.cache_dir if args.cache_dir else None,
)
model = model_class.from_pretrained(
args.model_name_or_path,
from_tf=bool(".ckpt" in args.model_name_or_path),
config=config,
cache_dir=args.cache_dir if args.cache_dir else None,
)
if args.local_rank == 0:
torch.distributed.barrier() # Make sure only the first process in distributed training will download model & vocab
@@ -495,7 +653,8 @@ def main():
if args.fp16:
try:
import apex
apex.amp.register_half_function(torch, 'einsum')
apex.amp.register_half_function(torch, "einsum")
except ImportError:
raise ImportError("Please install apex from https://www.github.com/nvidia/apex to use fp16 training.")
@@ -505,7 +664,6 @@ 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.do_train and (args.local_rank == -1 or torch.distributed.get_rank() == 0):
# Create output directory if needed
@@ -515,39 +673,42 @@ 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, do_lower_case=args.do_lower_case)
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]:
checkpoints = [args.output_dir]
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)
model.to(args.device)
# 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))

287
templates/adding_a_new_example_script/utils_xxx.py
View File

@@ -1,4 +1,3 @@
# coding=utf-8
# Copyright 2018 XXX. All rights reserved.
#
@@ -37,14 +36,16 @@ class SquadExample(object):
For examples without an answer, the start and end position are -1.
"""
def __init__(self,
qas_id,
question_text,
doc_tokens,
orig_answer_text=None,
start_position=None,
end_position=None,
is_impossible=None):
def __init__(
self,
qas_id,
question_text,
doc_tokens,
orig_answer_text=None,
start_position=None,
end_position=None,
is_impossible=None,
):
self.qas_id = qas_id
self.question_text = question_text
self.doc_tokens = doc_tokens
@@ -59,8 +60,7 @@ class SquadExample(object):
def __repr__(self):
s = ""
s += "qas_id: %s" % (self.qas_id)
s += ", question_text: %s" % (
self.question_text)
s += ", question_text: %s" % (self.question_text)
s += ", doc_tokens: [%s]" % (" ".join(self.doc_tokens))
if self.start_position:
s += ", start_position: %d" % (self.start_position)
@@ -74,22 +74,24 @@ class SquadExample(object):
class InputFeatures(object):
"""A single set of features of data."""
def __init__(self,
unique_id,
example_index,
doc_span_index,
tokens,
token_to_orig_map,
token_is_max_context,
input_ids,
input_mask,
segment_ids,
cls_index,
p_mask,
paragraph_len,
start_position=None,
end_position=None,
is_impossible=None):
def __init__(
self,
unique_id,
example_index,
doc_span_index,
tokens,
token_to_orig_map,
token_is_max_context,
input_ids,
input_mask,
segment_ids,
cls_index,
p_mask,
paragraph_len,
start_position=None,
end_position=None,
is_impossible=None,
):
self.unique_id = unique_id
self.example_index = example_index
self.doc_span_index = doc_span_index
@@ -109,7 +111,7 @@ class InputFeatures(object):
def read_squad_examples(input_file, is_training, version_2_with_negative):
"""Read a SQuAD json file into a list of SquadExample."""
with open(input_file, "r", encoding='utf-8') as reader:
with open(input_file, "r", encoding="utf-8") as reader:
input_data = json.load(reader)["data"]
def is_whitespace(c):
@@ -146,8 +148,7 @@ def read_squad_examples(input_file, is_training, version_2_with_negative):
if version_2_with_negative:
is_impossible = qa["is_impossible"]
if (len(qa["answers"]) != 1) and (not is_impossible):
raise ValueError(
"For training, each question should have exactly 1 answer.")
raise ValueError("For training, each question should have exactly 1 answer.")
if not is_impossible:
answer = qa["answers"][0]
orig_answer_text = answer["text"]
@@ -161,12 +162,10 @@ def read_squad_examples(input_file, is_training, version_2_with_negative):
#
# Note that this means for training mode, every example is NOT
# guaranteed to be preserved.
actual_text = " ".join(doc_tokens[start_position:(end_position + 1)])
cleaned_answer_text = " ".join(
whitespace_tokenize(orig_answer_text))
actual_text = " ".join(doc_tokens[start_position : (end_position + 1)])
cleaned_answer_text = " ".join(whitespace_tokenize(orig_answer_text))
if actual_text.find(cleaned_answer_text) == -1:
logger.warning("Could not find answer: '%s' vs. '%s'",
actual_text, cleaned_answer_text)
logger.warning("Could not find answer: '%s' vs. '%s'", actual_text, cleaned_answer_text)
continue
else:
start_position = -1
@@ -180,18 +179,29 @@ def read_squad_examples(input_file, is_training, version_2_with_negative):
orig_answer_text=orig_answer_text,
start_position=start_position,
end_position=end_position,
is_impossible=is_impossible)
is_impossible=is_impossible,
)
examples.append(example)
return examples
def convert_examples_to_features(examples, tokenizer, max_seq_length,
doc_stride, max_query_length, is_training,
cls_token_at_end=False,
cls_token='[CLS]', sep_token='[SEP]', pad_token=0,
sequence_a_segment_id=0, sequence_b_segment_id=1,
cls_token_segment_id=0, pad_token_segment_id=0,
mask_padding_with_zero=True):
def convert_examples_to_features(
examples,
tokenizer,
max_seq_length,
doc_stride,
max_query_length,
is_training,
cls_token_at_end=False,
cls_token="[CLS]",
sep_token="[SEP]",
pad_token=0,
sequence_a_segment_id=0,
sequence_b_segment_id=1,
cls_token_segment_id=0,
pad_token_segment_id=0,
mask_padding_with_zero=True,
):
"""Loads a data file into a list of `InputBatch`s."""
unique_id = 1000000000
@@ -232,8 +242,8 @@ def convert_examples_to_features(examples, tokenizer, max_seq_length,
else:
tok_end_position = len(all_doc_tokens) - 1
(tok_start_position, tok_end_position) = _improve_answer_span(
all_doc_tokens, tok_start_position, tok_end_position, tokenizer,
example.orig_answer_text)
all_doc_tokens, tok_start_position, tok_end_position, tokenizer, example.orig_answer_text
)
# The -3 accounts for [CLS], [SEP] and [SEP]
max_tokens_for_doc = max_seq_length - len(query_tokens) - 3
@@ -241,8 +251,7 @@ def convert_examples_to_features(examples, tokenizer, max_seq_length,
# We can have documents that are longer than the maximum sequence length.
# To deal with this we do a sliding window approach, where we take chunks
# of the up to our max length with a stride of `doc_stride`.
_DocSpan = collections.namedtuple( # pylint: disable=invalid-name
"DocSpan", ["start", "length"])
_DocSpan = collections.namedtuple("DocSpan", ["start", "length"]) # pylint: disable=invalid-name
doc_spans = []
start_offset = 0
while start_offset < len(all_doc_tokens):
@@ -287,8 +296,7 @@ def convert_examples_to_features(examples, tokenizer, max_seq_length,
split_token_index = doc_span.start + i
token_to_orig_map[len(tokens)] = tok_to_orig_index[split_token_index]
is_max_context = _check_is_max_context(doc_spans, doc_span_index,
split_token_index)
is_max_context = _check_is_max_context(doc_spans, doc_span_index, split_token_index)
token_is_max_context[len(tokens)] = is_max_context
tokens.append(all_doc_tokens[split_token_index])
segment_ids.append(sequence_b_segment_id)
@@ -333,8 +341,7 @@ def convert_examples_to_features(examples, tokenizer, max_seq_length,
doc_start = doc_span.start
doc_end = doc_span.start + doc_span.length - 1
out_of_span = False
if not (tok_start_position >= doc_start and
tok_end_position <= doc_end):
if not (tok_start_position >= doc_start and tok_end_position <= doc_end):
out_of_span = True
if out_of_span:
start_position = 0
@@ -355,24 +362,23 @@ def convert_examples_to_features(examples, tokenizer, max_seq_length,
logger.info("example_index: %s" % (example_index))
logger.info("doc_span_index: %s" % (doc_span_index))
logger.info("tokens: %s" % " ".join(tokens))
logger.info("token_to_orig_map: %s" % " ".join([
"%d:%d" % (x, y) for (x, y) in token_to_orig_map.items()]))
logger.info("token_is_max_context: %s" % " ".join([
"%d:%s" % (x, y) for (x, y) in token_is_max_context.items()
]))
logger.info(
"token_to_orig_map: %s" % " ".join(["%d:%d" % (x, y) for (x, y) in token_to_orig_map.items()])
)
logger.info(
"token_is_max_context: %s"
% " ".join(["%d:%s" % (x, y) for (x, y) in token_is_max_context.items()])
)
logger.info("input_ids: %s" % " ".join([str(x) for x in input_ids]))
logger.info(
"input_mask: %s" % " ".join([str(x) for x in input_mask]))
logger.info(
"segment_ids: %s" % " ".join([str(x) for x in segment_ids]))
logger.info("input_mask: %s" % " ".join([str(x) for x in input_mask]))
logger.info("segment_ids: %s" % " ".join([str(x) for x in segment_ids]))
if is_training and span_is_impossible:
logger.info("impossible example")
if is_training and not span_is_impossible:
answer_text = " ".join(tokens[start_position:(end_position + 1)])
answer_text = " ".join(tokens[start_position : (end_position + 1)])
logger.info("start_position: %d" % (start_position))
logger.info("end_position: %d" % (end_position))
logger.info(
"answer: %s" % (answer_text))
logger.info("answer: %s" % (answer_text))
features.append(
InputFeatures(
@@ -390,14 +396,15 @@ def convert_examples_to_features(examples, tokenizer, max_seq_length,
paragraph_len=paragraph_len,
start_position=start_position,
end_position=end_position,
is_impossible=span_is_impossible))
is_impossible=span_is_impossible,
)
)
unique_id += 1
return features
def _improve_answer_span(doc_tokens, input_start, input_end, tokenizer,
orig_answer_text):
def _improve_answer_span(doc_tokens, input_start, input_end, tokenizer, orig_answer_text):
"""Returns tokenized answer spans that better match the annotated answer."""
# The SQuAD annotations are character based. We first project them to
@@ -426,7 +433,7 @@ def _improve_answer_span(doc_tokens, input_start, input_end, tokenizer,
for new_start in range(input_start, input_end + 1):
for new_end in range(input_end, new_start - 1, -1):
text_span = " ".join(doc_tokens[new_start:(new_end + 1)])
text_span = " ".join(doc_tokens[new_start : (new_end + 1)])
if text_span == tok_answer_text:
return (new_start, new_end)
@@ -470,13 +477,23 @@ def _check_is_max_context(doc_spans, cur_span_index, position):
return cur_span_index == best_span_index
RawResult = collections.namedtuple("RawResult",
["unique_id", "start_logits", "end_logits"])
RawResult = collections.namedtuple("RawResult", ["unique_id", "start_logits", "end_logits"])
def write_predictions(all_examples, all_features, all_results, n_best_size,
max_answer_length, do_lower_case, output_prediction_file,
output_nbest_file, output_null_log_odds_file, verbose_logging,
version_2_with_negative, null_score_diff_threshold):
def write_predictions(
all_examples,
all_features,
all_results,
n_best_size,
max_answer_length,
do_lower_case,
output_prediction_file,
output_nbest_file,
output_null_log_odds_file,
verbose_logging,
version_2_with_negative,
null_score_diff_threshold,
):
"""Write final predictions to the json file and log-odds of null if needed."""
logger.info("Writing predictions to: %s" % (output_prediction_file))
logger.info("Writing nbest to: %s" % (output_nbest_file))
@@ -490,8 +507,8 @@ def write_predictions(all_examples, all_features, all_results, n_best_size,
unique_id_to_result[result.unique_id] = result
_PrelimPrediction = collections.namedtuple( # pylint: disable=invalid-name
"PrelimPrediction",
["feature_index", "start_index", "end_index", "start_logit", "end_logit"])
"PrelimPrediction", ["feature_index", "start_index", "end_index", "start_logit", "end_logit"]
)
all_predictions = collections.OrderedDict()
all_nbest_json = collections.OrderedDict()
@@ -544,7 +561,9 @@ def write_predictions(all_examples, all_features, all_results, n_best_size,
start_index=start_index,
end_index=end_index,
start_logit=result.start_logits[start_index],
end_logit=result.end_logits[end_index]))
end_logit=result.end_logits[end_index],
)
)
if version_2_with_negative:
prelim_predictions.append(
_PrelimPrediction(
@@ -552,14 +571,14 @@ def write_predictions(all_examples, all_features, all_results, n_best_size,
start_index=0,
end_index=0,
start_logit=null_start_logit,
end_logit=null_end_logit))
prelim_predictions = sorted(
prelim_predictions,
key=lambda x: (x.start_logit + x.end_logit),
reverse=True)
end_logit=null_end_logit,
)
)
prelim_predictions = sorted(prelim_predictions, key=lambda x: (x.start_logit + x.end_logit), reverse=True)
_NbestPrediction = collections.namedtuple( # pylint: disable=invalid-name
"NbestPrediction", ["text", "start_logit", "end_logit"])
"NbestPrediction", ["text", "start_logit", "end_logit"]
)
seen_predictions = {}
nbest = []
@@ -568,10 +587,10 @@ def write_predictions(all_examples, all_features, all_results, n_best_size,
break
feature = features[pred.feature_index]
if pred.start_index > 0: # this is a non-null prediction
tok_tokens = feature.tokens[pred.start_index:(pred.end_index + 1)]
tok_tokens = feature.tokens[pred.start_index : (pred.end_index + 1)]
orig_doc_start = feature.token_to_orig_map[pred.start_index]
orig_doc_end = feature.token_to_orig_map[pred.end_index]
orig_tokens = example.doc_tokens[orig_doc_start:(orig_doc_end + 1)]
orig_tokens = example.doc_tokens[orig_doc_start : (orig_doc_end + 1)]
tok_text = " ".join(tok_tokens)
# De-tokenize WordPieces that have been split off.
@@ -592,31 +611,21 @@ def write_predictions(all_examples, all_features, all_results, n_best_size,
final_text = ""
seen_predictions[final_text] = True
nbest.append(
_NbestPrediction(
text=final_text,
start_logit=pred.start_logit,
end_logit=pred.end_logit))
nbest.append(_NbestPrediction(text=final_text, start_logit=pred.start_logit, end_logit=pred.end_logit))
# if we didn't include the empty option in the n-best, include it
if version_2_with_negative:
if "" not in seen_predictions:
nbest.append(
_NbestPrediction(
text="",
start_logit=null_start_logit,
end_logit=null_end_logit))
nbest.append(_NbestPrediction(text="", start_logit=null_start_logit, end_logit=null_end_logit))
# In very rare edge cases we could only have single null prediction.
# So we just create a nonce prediction in this case to avoid failure.
if len(nbest)==1:
nbest.insert(0,
_NbestPrediction(text="empty", start_logit=0.0, end_logit=0.0))
if len(nbest) == 1:
nbest.insert(0, _NbestPrediction(text="empty", start_logit=0.0, end_logit=0.0))
# In very rare edge cases we could have no valid predictions. So we
# just create a nonce prediction in this case to avoid failure.
if not nbest:
nbest.append(
_NbestPrediction(text="empty", start_logit=0.0, end_logit=0.0))
nbest.append(_NbestPrediction(text="empty", start_logit=0.0, end_logit=0.0))
assert len(nbest) >= 1
@@ -645,8 +654,7 @@ def write_predictions(all_examples, all_features, all_results, n_best_size,
all_predictions[example.qas_id] = nbest_json[0]["text"]
else:
# predict "" iff the null score - the score of best non-null > threshold
score_diff = score_null - best_non_null_entry.start_logit - (
best_non_null_entry.end_logit)
score_diff = score_null - best_non_null_entry.start_logit - (best_non_null_entry.end_logit)
scores_diff_json[example.qas_id] = score_diff
if score_diff > null_score_diff_threshold:
all_predictions[example.qas_id] = ""
@@ -668,29 +676,40 @@ def write_predictions(all_examples, all_features, all_results, n_best_size,
# For XLNet (and XLM which uses the same head)
RawResultExtended = collections.namedtuple("RawResultExtended",
["unique_id", "start_top_log_probs", "start_top_index",
"end_top_log_probs", "end_top_index", "cls_logits"])
RawResultExtended = collections.namedtuple(
"RawResultExtended",
["unique_id", "start_top_log_probs", "start_top_index", "end_top_log_probs", "end_top_index", "cls_logits"],
)
def write_predictions_extended(all_examples, all_features, all_results, n_best_size,
max_answer_length, output_prediction_file,
output_nbest_file,
output_null_log_odds_file, orig_data_file,
start_n_top, end_n_top, version_2_with_negative,
tokenizer, verbose_logging):
def write_predictions_extended(
all_examples,
all_features,
all_results,
n_best_size,
max_answer_length,
output_prediction_file,
output_nbest_file,
output_null_log_odds_file,
orig_data_file,
start_n_top,
end_n_top,
version_2_with_negative,
tokenizer,
verbose_logging,
):
""" XLNet write prediction logic (more complex than Bert's).
Write final predictions to the json file and log-odds of null if needed.
Requires utils_squad_evaluate.py
"""
_PrelimPrediction = collections.namedtuple( # pylint: disable=invalid-name
"PrelimPrediction",
["feature_index", "start_index", "end_index",
"start_log_prob", "end_log_prob"])
"PrelimPrediction", ["feature_index", "start_index", "end_index", "start_log_prob", "end_log_prob"]
)
_NbestPrediction = collections.namedtuple( # pylint: disable=invalid-name
"NbestPrediction", ["text", "start_log_prob", "end_log_prob"])
"NbestPrediction", ["text", "start_log_prob", "end_log_prob"]
)
logger.info("Writing predictions to: %s", output_prediction_file)
# logger.info("Writing nbest to: %s" % (output_nbest_file))
@@ -754,12 +773,13 @@ def write_predictions_extended(all_examples, all_features, all_results, n_best_s
start_index=start_index,
end_index=end_index,
start_log_prob=start_log_prob,
end_log_prob=end_log_prob))
end_log_prob=end_log_prob,
)
)
prelim_predictions = sorted(
prelim_predictions,
key=lambda x: (x.start_log_prob + x.end_log_prob),
reverse=True)
prelim_predictions, key=lambda x: (x.start_log_prob + x.end_log_prob), reverse=True
)
seen_predictions = {}
nbest = []
@@ -770,7 +790,7 @@ def write_predictions_extended(all_examples, all_features, all_results, n_best_s
# XLNet un-tokenizer
# Let's keep it simple for now and see if we need all this later.
#
#
# tok_start_to_orig_index = feature.tok_start_to_orig_index
# tok_end_to_orig_index = feature.tok_end_to_orig_index
# start_orig_pos = tok_start_to_orig_index[pred.start_index]
@@ -779,10 +799,10 @@ def write_predictions_extended(all_examples, all_features, all_results, n_best_s
# final_text = paragraph_text[start_orig_pos: end_orig_pos + 1].strip()
# Previously used Bert untokenizer
tok_tokens = feature.tokens[pred.start_index:(pred.end_index + 1)]
tok_tokens = feature.tokens[pred.start_index : (pred.end_index + 1)]
orig_doc_start = feature.token_to_orig_map[pred.start_index]
orig_doc_end = feature.token_to_orig_map[pred.end_index]
orig_tokens = example.doc_tokens[orig_doc_start:(orig_doc_end + 1)]
orig_tokens = example.doc_tokens[orig_doc_start : (orig_doc_end + 1)]
tok_text = tokenizer.convert_tokens_to_string(tok_tokens)
# Clean whitespace
@@ -790,8 +810,7 @@ def write_predictions_extended(all_examples, all_features, all_results, n_best_s
tok_text = " ".join(tok_text.split())
orig_text = " ".join(orig_tokens)
final_text = get_final_text(tok_text, orig_text, tokenizer.do_lower_case,
verbose_logging)
final_text = get_final_text(tok_text, orig_text, tokenizer.do_lower_case, verbose_logging)
if final_text in seen_predictions:
continue
@@ -799,17 +818,13 @@ def write_predictions_extended(all_examples, all_features, all_results, n_best_s
seen_predictions[final_text] = True
nbest.append(
_NbestPrediction(
text=final_text,
start_log_prob=pred.start_log_prob,
end_log_prob=pred.end_log_prob))
_NbestPrediction(text=final_text, start_log_prob=pred.start_log_prob, end_log_prob=pred.end_log_prob)
)
# In very rare edge cases we could have no valid predictions. So we
# just create a nonce prediction in this case to avoid failure.
if not nbest:
nbest.append(
_NbestPrediction(text="", start_log_prob=-1e6,
end_log_prob=-1e6))
nbest.append(_NbestPrediction(text="", start_log_prob=-1e6, end_log_prob=-1e6))
total_scores = []
best_non_null_entry = None
@@ -850,7 +865,7 @@ def write_predictions_extended(all_examples, all_features, all_results, n_best_s
with open(output_null_log_odds_file, "w") as writer:
writer.write(json.dumps(scores_diff_json, indent=4) + "\n")
with open(orig_data_file, "r", encoding='utf-8') as reader:
with open(orig_data_file, "r", encoding="utf-8") as reader:
orig_data = json.load(reader)["data"]
qid_to_has_ans = make_qid_to_has_ans(orig_data)
@@ -914,8 +929,7 @@ def get_final_text(pred_text, orig_text, do_lower_case, verbose_logging=False):
start_position = tok_text.find(pred_text)
if start_position == -1:
if verbose_logging:
logger.info(
"Unable to find text: '%s' in '%s'" % (pred_text, orig_text))
logger.info("Unable to find text: '%s' in '%s'" % (pred_text, orig_text))
return orig_text
end_position = start_position + len(pred_text) - 1
@@ -924,8 +938,7 @@ def get_final_text(pred_text, orig_text, do_lower_case, verbose_logging=False):
if len(orig_ns_text) != len(tok_ns_text):
if verbose_logging:
logger.info("Length not equal after stripping spaces: '%s' vs '%s'",
orig_ns_text, tok_ns_text)
logger.info("Length not equal after stripping spaces: '%s' vs '%s'", orig_ns_text, tok_ns_text)
return orig_text
# We then project the characters in `pred_text` back to `orig_text` using
@@ -956,7 +969,7 @@ def get_final_text(pred_text, orig_text, do_lower_case, verbose_logging=False):
logger.info("Couldn't map end position")
return orig_text
output_text = orig_text[orig_start_position:(orig_end_position + 1)]
output_text = orig_text[orig_start_position : (orig_end_position + 1)]
return output_text

42
templates/adding_a_new_model/configuration_xxx.py
View File

@@ -27,8 +27,8 @@ from .configuration_utils import PretrainedConfig
logger = logging.getLogger(__name__)
XXX_PRETRAINED_CONFIG_ARCHIVE_MAP = {
'xxx-base-uncased': "https://s3.amazonaws.com/models.huggingface.co/bert/xxx-base-uncased-config.json",
'xxx-large-uncased': "https://s3.amazonaws.com/models.huggingface.co/bert/xxx-large-uncased-config.json",
"xxx-base-uncased": "https://s3.amazonaws.com/models.huggingface.co/bert/xxx-base-uncased-config.json",
"xxx-large-uncased": "https://s3.amazonaws.com/models.huggingface.co/bert/xxx-large-uncased-config.json",
}
@@ -63,24 +63,26 @@ class XxxConfig(PretrainedConfig):
"""
pretrained_config_archive_map = XXX_PRETRAINED_CONFIG_ARCHIVE_MAP
def __init__(self,
vocab_size=50257,
n_positions=1024,
n_ctx=1024,
n_embd=768,
n_layer=12,
n_head=12,
resid_pdrop=0.1,
embd_pdrop=0.1,
attn_pdrop=0.1,
layer_norm_epsilon=1e-5,
initializer_range=0.02,
summary_type='cls_index',
summary_use_proj=True,
summary_activation=None,
summary_proj_to_labels=True,
summary_first_dropout=0.1,
**kwargs):
def __init__(
self,
vocab_size=50257,
n_positions=1024,
n_ctx=1024,
n_embd=768,
n_layer=12,
n_head=12,
resid_pdrop=0.1,
embd_pdrop=0.1,
attn_pdrop=0.1,
layer_norm_epsilon=1e-5,
initializer_range=0.02,
summary_type="cls_index",
summary_use_proj=True,
summary_activation=None,
summary_proj_to_labels=True,
summary_first_dropout=0.1,
**kwargs
):
super(XxxConfig, self).__init__(**kwargs)
self.vocab_size = vocab_size
self.n_ctx = n_ctx

36
templates/adding_a_new_model/convert_xxx_original_tf_checkpoint_to_pytorch.py
View File

@@ -24,8 +24,10 @@ import torch
from transformers import XxxConfig, XxxForPreTraining, load_tf_weights_in_xxx
import logging
logging.basicConfig(level=logging.INFO)
def convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, config_file, pytorch_dump_path):
# Initialise PyTorch model
config = XxxConfig.from_json_file(config_file)
@@ -43,23 +45,19 @@ def convert_tf_checkpoint_to_pytorch(tf_checkpoint_path, config_file, pytorch_du
if __name__ == "__main__":
parser = argparse.ArgumentParser()
## Required parameters
parser.add_argument("--tf_checkpoint_path",
default = None,
type = str,
required = True,
help = "Path to the TensorFlow checkpoint path.")
parser.add_argument("--config_file",
default = None,
type = str,
required = True,
help = "The config json file corresponding to the pre-trained model. \n"
"This specifies the model architecture.")
parser.add_argument("--pytorch_dump_path",
default = None,
type = str,
required = True,
help = "Path to the output PyTorch model.")
parser.add_argument(
"--tf_checkpoint_path", default=None, type=str, required=True, help="Path to the TensorFlow checkpoint path."
)
parser.add_argument(
"--config_file",
default=None,
type=str,
required=True,
help="The config json file corresponding to the pre-trained model. \n"
"This specifies the model architecture.",
)
parser.add_argument(
"--pytorch_dump_path", default=None, type=str, required=True, help="Path to the output PyTorch model."
)
args = parser.parse_args()
convert_tf_checkpoint_to_pytorch(args.tf_checkpoint_path,
args.config_file,
args.pytorch_dump_path)
convert_tf_checkpoint_to_pytorch(args.tf_checkpoint_path, args.config_file, args.pytorch_dump_path)

100
templates/adding_a_new_model/modeling_tf_xxx.py
View File

@@ -44,8 +44,8 @@ logger = logging.getLogger(__name__)
# for the pretrained weights provided with the models
####################################################
TF_XXX_PRETRAINED_MODEL_ARCHIVE_MAP = {
'xxx-base-uncased': "https://s3.amazonaws.com/models.huggingface.co/bert/xxx-base-uncased-tf_model.h5",
'xxx-large-uncased': "https://s3.amazonaws.com/models.huggingface.co/bert/xxx-large-uncased-tf_model.h5",
"xxx-base-uncased": "https://s3.amazonaws.com/models.huggingface.co/bert/xxx-base-uncased-tf_model.h5",
"xxx-large-uncased": "https://s3.amazonaws.com/models.huggingface.co/bert/xxx-large-uncased-tf_model.h5",
}
####################################################
@@ -69,9 +69,9 @@ TF_XXX_PRETRAINED_MODEL_ARCHIVE_MAP = {
class TFXxxLayer(tf.keras.layers.Layer):
def __init__(self, config, **kwargs):
super(TFXxxLayer, self).__init__(**kwargs)
self.attention = TFXxxAttention(config, name='attention')
self.intermediate = TFXxxIntermediate(config, name='intermediate')
self.transformer_output = TFXxxOutput(config, name='output')
self.attention = TFXxxAttention(config, name="attention")
self.intermediate = TFXxxIntermediate(config, name="intermediate")
self.transformer_output = TFXxxOutput(config, name="output")
def call(self, inputs, training=False):
hidden_states, attention_mask, head_mask = inputs
@@ -98,7 +98,9 @@ class TFXxxMainLayer(tf.keras.layers.Layer):
def _prune_heads(self, heads_to_prune):
raise NotImplementedError # Not implemented yet in the library fr TF 2.0 models
def call(self, inputs, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, training=False):
def call(
self, inputs, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, training=False
):
# We allow three types of multi-inputs:
# - traditional keyword arguments in the call method
# - all the arguments provided as a dict in the first positional argument of call
@@ -113,11 +115,11 @@ class TFXxxMainLayer(tf.keras.layers.Layer):
head_mask = inputs[4] if len(inputs) > 4 else head_mask
assert len(inputs) <= 5, "Too many inputs."
elif isinstance(inputs, dict):
input_ids = inputs.get('input_ids')
attention_mask = inputs.get('attention_mask', attention_mask)
token_type_ids = inputs.get('token_type_ids', token_type_ids)
position_ids = inputs.get('position_ids', position_ids)
head_mask = inputs.get('head_mask', head_mask)
input_ids = inputs.get("input_ids")
attention_mask = inputs.get("attention_mask", attention_mask)
token_type_ids = inputs.get("token_type_ids", token_type_ids)
position_ids = inputs.get("position_ids", position_ids)
head_mask = inputs.get("head_mask", head_mask)
assert len(inputs) <= 5, "Too many inputs."
else:
input_ids = inputs
@@ -175,6 +177,7 @@ class TFXxxPreTrainedModel(TFPreTrainedModel):
""" An abstract class to handle weights initialization and
a simple interface for dowloading and loading pretrained models.
"""
config_class = XxxConfig
pretrained_model_archive_map = TF_XXX_PRETRAINED_MODEL_ARCHIVE_MAP
base_model_prefix = "transformer"
@@ -263,8 +266,12 @@ XXX_INPUTS_DOCSTRING = r"""
than the model's internal embedding lookup matrix.
"""
@add_start_docstrings("The bare Xxx Model transformer outputing raw hidden-states without any specific head on top.",
XXX_START_DOCSTRING, XXX_INPUTS_DOCSTRING)
@add_start_docstrings(
"The bare Xxx Model transformer outputing raw hidden-states without any specific head on top.",
XXX_START_DOCSTRING,
XXX_INPUTS_DOCSTRING,
)
class TFXxxModel(TFXxxPreTrainedModel):
r"""
Outputs: `Tuple` comprising various elements depending on the configuration (config) and inputs:
@@ -297,17 +304,19 @@ class TFXxxModel(TFXxxPreTrainedModel):
last_hidden_states = outputs[0] # The last hidden-state is the first element of the output tuple
"""
def __init__(self, config, *inputs, **kwargs):
super(TFXxxModel, self).__init__(config, *inputs, **kwargs)
self.transformer = TFXxxMainLayer(config, name='transformer')
self.transformer = TFXxxMainLayer(config, name="transformer")
def call(self, inputs, **kwargs):
outputs = self.transformer(inputs, **kwargs)
return outputs
@add_start_docstrings("""Xxx Model with a `language modeling` head on top. """,
XXX_START_DOCSTRING, XXX_INPUTS_DOCSTRING)
@add_start_docstrings(
"""Xxx Model with a `language modeling` head on top. """, XXX_START_DOCSTRING, XXX_INPUTS_DOCSTRING
)
class TFXxxForMaskedLM(TFXxxPreTrainedModel):
r"""
Outputs: `Tuple` comprising various elements depending on the configuration (config) and inputs:
@@ -333,26 +342,30 @@ class TFXxxForMaskedLM(TFXxxPreTrainedModel):
prediction_scores = outputs[0]
"""
def __init__(self, config, *inputs, **kwargs):
super(TFXxxForMaskedLM, self).__init__(config, *inputs, **kwargs)
self.transformer = TFXxxMainLayer(config, name='transformer')
self.mlm = TFXxxMLMHead(config, self.transformer.embeddings, name='mlm')
self.transformer = TFXxxMainLayer(config, name="transformer")
self.mlm = TFXxxMLMHead(config, self.transformer.embeddings, name="mlm")
def call(self, inputs, **kwargs):
outputs = self.transformer(inputs, **kwargs)
sequence_output = outputs[0]
prediction_scores = self.mlm(sequence_output, training=kwargs.get('training', False))
prediction_scores = self.mlm(sequence_output, training=kwargs.get("training", False))
outputs = (prediction_scores,) + outputs[2:] # Add hidden states and attention if they are here
return outputs # prediction_scores, (hidden_states), (attentions)
@add_start_docstrings("""Xxx Model transformer with a sequence classification/regression head on top (a linear layer on top of
@add_start_docstrings(
"""Xxx Model transformer with a sequence classification/regression head on top (a linear layer on top of
the pooled output) e.g. for GLUE tasks. """,
XXX_START_DOCSTRING, XXX_INPUTS_DOCSTRING)
XXX_START_DOCSTRING,
XXX_INPUTS_DOCSTRING,
)
class TFXxxForSequenceClassification(TFXxxPreTrainedModel):
r"""
Outputs: `Tuple` comprising various elements depending on the configuration (config) and inputs:
@@ -378,22 +391,23 @@ class TFXxxForSequenceClassification(TFXxxPreTrainedModel):
logits = outputs[0]
"""
def __init__(self, config, *inputs, **kwargs):
super(TFXxxForSequenceClassification, self).__init__(config, *inputs, **kwargs)
self.num_labels = config.num_labels
self.transformer = TFXxxMainLayer(config, name='transformer')
self.transformer = TFXxxMainLayer(config, name="transformer")
self.dropout = tf.keras.layers.Dropout(config.hidden_dropout_prob)
self.classifier = tf.keras.layers.Dense(config.num_labels,
kernel_initializer=get_initializer(config.initializer_range),
name='classifier')
self.classifier = tf.keras.layers.Dense(
config.num_labels, kernel_initializer=get_initializer(config.initializer_range), name="classifier"
)
def call(self, inputs, **kwargs):
outputs = self.transformer(inputs, **kwargs)
pooled_output = outputs[1]
pooled_output = self.dropout(pooled_output, training=kwargs.get('training', False))
pooled_output = self.dropout(pooled_output, training=kwargs.get("training", False))
logits = self.classifier(pooled_output)
outputs = (logits,) + outputs[2:] # add hidden states and attention if they are here
@@ -401,9 +415,12 @@ class TFXxxForSequenceClassification(TFXxxPreTrainedModel):
return outputs # logits, (hidden_states), (attentions)
@add_start_docstrings("""Xxx Model with a token classification head on top (a linear layer on top of
@add_start_docstrings(
"""Xxx Model with a token classification head on top (a linear layer on top of
the hidden-states output) e.g. for Named-Entity-Recognition (NER) tasks. """,
XXX_START_DOCSTRING, XXX_INPUTS_DOCSTRING)
XXX_START_DOCSTRING,
XXX_INPUTS_DOCSTRING,
)
class TFXxxForTokenClassification(TFXxxPreTrainedModel):
r"""
Outputs: `Tuple` comprising various elements depending on the configuration (config) and inputs:
@@ -429,22 +446,23 @@ class TFXxxForTokenClassification(TFXxxPreTrainedModel):
scores = outputs[0]
"""
def __init__(self, config, *inputs, **kwargs):
super(TFXxxForTokenClassification, self).__init__(config, *inputs, **kwargs)
self.num_labels = config.num_labels
self.transformer = TFXxxMainLayer(config, name='transformer')
self.transformer = TFXxxMainLayer(config, name="transformer")
self.dropout = tf.keras.layers.Dropout(config.hidden_dropout_prob)
self.classifier = tf.keras.layers.Dense(config.num_labels,
kernel_initializer=get_initializer(config.initializer_range),
name='classifier')
self.classifier = tf.keras.layers.Dense(
config.num_labels, kernel_initializer=get_initializer(config.initializer_range), name="classifier"
)
def call(self, inputs, **kwargs):
outputs = self.transformer(inputs, **kwargs)
sequence_output = outputs[0]
sequence_output = self.dropout(sequence_output, training=kwargs.get('training', False))
sequence_output = self.dropout(sequence_output, training=kwargs.get("training", False))
logits = self.classifier(sequence_output)
outputs = (logits,) + outputs[2:] # add hidden states and attention if they are here
@@ -452,9 +470,12 @@ class TFXxxForTokenClassification(TFXxxPreTrainedModel):
return outputs # scores, (hidden_states), (attentions)
@add_start_docstrings("""Xxx Model with a span classification head on top for extractive question-answering tasks like SQuAD (a linear layers on top of
@add_start_docstrings(
"""Xxx Model with a span classification head on top for extractive question-answering tasks like SQuAD (a linear layers on top of
the hidden-states output to compute `span start logits` and `span end logits`). """,
XXX_START_DOCSTRING, XXX_INPUTS_DOCSTRING)
XXX_START_DOCSTRING,
XXX_INPUTS_DOCSTRING,
)
class TFXxxForQuestionAnswering(TFXxxPreTrainedModel):
r"""
Outputs: `Tuple` comprising various elements depending on the configuration (config) and inputs:
@@ -482,14 +503,15 @@ class TFXxxForQuestionAnswering(TFXxxPreTrainedModel):
start_scores, end_scores = outputs[:2]
"""
def __init__(self, config, *inputs, **kwargs):
super(TFXxxForQuestionAnswering, self).__init__(config, *inputs, **kwargs)
self.num_labels = config.num_labels
self.transformer = TFXxxMainLayer(config, name='transformer')
self.qa_outputs = tf.keras.layers.Dense(config.num_labels,
kernel_initializer=get_initializer(config.initializer_range),
name='qa_outputs')
self.transformer = TFXxxMainLayer(config, name="transformer")
self.qa_outputs = tf.keras.layers.Dense(
config.num_labels, kernel_initializer=get_initializer(config.initializer_range), name="qa_outputs"
)
def call(self, inputs, **kwargs):
outputs = self.transformer(inputs, **kwargs)

208
templates/adding_a_new_model/modeling_xxx.py
View File

@@ -44,8 +44,8 @@ logger = logging.getLogger(__name__)
# for the pretrained weights provided with the models
####################################################
XXX_PRETRAINED_MODEL_ARCHIVE_MAP = {
'xxx-base-uncased': "https://s3.amazonaws.com/models.huggingface.co/bert/xxx-base-uncased-pytorch_model.bin",
'xxx-large-uncased': "https://s3.amazonaws.com/models.huggingface.co/bert/xxx-large-uncased-pytorch_model.bin",
"xxx-base-uncased": "https://s3.amazonaws.com/models.huggingface.co/bert/xxx-base-uncased-pytorch_model.bin",
"xxx-large-uncased": "https://s3.amazonaws.com/models.huggingface.co/bert/xxx-large-uncased-pytorch_model.bin",
}
####################################################
@@ -60,8 +60,10 @@ def load_tf_weights_in_xxx(model, config, tf_checkpoint_path):
import numpy as np
import tensorflow as tf
except ImportError:
logger.error("Loading a TensorFlow model in PyTorch, requires TensorFlow to be installed. Please see "
"https://www.tensorflow.org/install/ for installation instructions.")
logger.error(
"Loading a TensorFlow model in PyTorch, requires TensorFlow to be installed. Please see "
"https://www.tensorflow.org/install/ for installation instructions."
)
raise
tf_path = os.path.abspath(tf_checkpoint_path)
logger.info("Converting TensorFlow checkpoint from {}".format(tf_path))
@@ -76,7 +78,7 @@ def load_tf_weights_in_xxx(model, config, tf_checkpoint_path):
arrays.append(array)
for name, array in zip(names, arrays):
name = name.split('/')
name = name.split("/")
# adam_v and adam_m are variables used in AdamWeightDecayOptimizer to calculated m and v
# which are not required for using pretrained model
if any(n in ["adam_v", "adam_m", "global_step"] for n in name):
@@ -84,18 +86,18 @@ def load_tf_weights_in_xxx(model, config, tf_checkpoint_path):
continue
pointer = model
for m_name in name:
if re.fullmatch(r'[A-Za-z]+_\d+', m_name):
l = re.split(r'_(\d+)', m_name)
if re.fullmatch(r"[A-Za-z]+_\d+", m_name):
l = re.split(r"_(\d+)", m_name)
else:
l = [m_name]
if l[0] == 'kernel' or l[0] == 'gamma':
pointer = getattr(pointer, 'weight')
elif l[0] == 'output_bias' or l[0] == 'beta':
pointer = getattr(pointer, 'bias')
elif l[0] == 'output_weights':
pointer = getattr(pointer, 'weight')
elif l[0] == 'squad':
pointer = getattr(pointer, 'classifier')
if l[0] == "kernel" or l[0] == "gamma":
pointer = getattr(pointer, "weight")
elif l[0] == "output_bias" or l[0] == "beta":
pointer = getattr(pointer, "bias")
elif l[0] == "output_weights":
pointer = getattr(pointer, "weight")
elif l[0] == "squad":
pointer = getattr(pointer, "classifier")
else:
try:
pointer = getattr(pointer, l[0])
@@ -105,9 +107,9 @@ def load_tf_weights_in_xxx(model, config, tf_checkpoint_path):
if len(l) >= 2:
num = int(l[1])
pointer = pointer[num]
if m_name[-11:] == '_embeddings':
pointer = getattr(pointer, 'weight')
elif m_name == 'kernel':
if m_name[-11:] == "_embeddings":
pointer = getattr(pointer, "weight")
elif m_name == "kernel":
array = np.transpose(array)
try:
assert pointer.shape == array.shape
@@ -147,7 +149,6 @@ class XxxLayer(nn.Module):
return outputs
####################################################
# PreTrainedModel is a sub-class of torch.nn.Module
# which take care of loading and saving pretrained weights
@@ -161,6 +162,7 @@ class XxxPreTrainedModel(PreTrainedModel):
""" An abstract class to handle weights initialization and
a simple interface for dowloading and loading pretrained models.
"""
config_class = XxxConfig
pretrained_model_archive_map = XXX_PRETRAINED_MODEL_ARCHIVE_MAP
load_tf_weights = load_tf_weights_in_xxx
@@ -246,8 +248,12 @@ XXX_INPUTS_DOCSTRING = r"""
than the model's internal embedding lookup matrix.
"""
@add_start_docstrings("The bare Xxx Model transformer outputting raw hidden-states without any specific head on top.",
XXX_START_DOCSTRING, XXX_INPUTS_DOCSTRING)
@add_start_docstrings(
"The bare Xxx Model transformer outputting raw hidden-states without any specific head on top.",
XXX_START_DOCSTRING,
XXX_INPUTS_DOCSTRING,
)
class XxxModel(XxxPreTrainedModel):
r"""
Outputs: `Tuple` comprising various elements depending on the configuration (config) and inputs:
@@ -277,6 +283,7 @@ class XxxModel(XxxPreTrainedModel):
last_hidden_states = outputs[0] # The last hidden-state is the first element of the output tuple
"""
def __init__(self, config):
super(XxxModel, self).__init__(config)
@@ -300,7 +307,15 @@ class XxxModel(XxxPreTrainedModel):
for layer, heads in heads_to_prune.items():
self.encoder.layer[layer].attention.prune_heads(heads)
def forward(self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None):
def forward(
self,
input_ids=None,
attention_mask=None,
token_type_ids=None,
position_ids=None,
head_mask=None,
inputs_embeds=None,
):
if input_ids is not None and inputs_embeds is not None:
raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
elif input_ids is not None:
@@ -329,7 +344,7 @@ class XxxModel(XxxPreTrainedModel):
# positions we want to attend and -10000.0 for masked positions.
# Since we are adding it to the raw scores before the softmax, this is
# effectively the same as removing these entirely.
extended_attention_mask = extended_attention_mask.to(dtype=next(self.parameters()).dtype) # fp16 compatibility
extended_attention_mask = extended_attention_mask.to(dtype=next(self.parameters()).dtype) # fp16 compatibility
extended_attention_mask = (1.0 - extended_attention_mask) * -10000.0
# Prepare head mask if needed
@@ -342,14 +357,20 @@ class XxxModel(XxxPreTrainedModel):
head_mask = head_mask.unsqueeze(0).unsqueeze(0).unsqueeze(-1).unsqueeze(-1)
head_mask = head_mask.expand(self.config.num_hidden_layers, -1, -1, -1, -1)
elif head_mask.dim() == 2:
head_mask = head_mask.unsqueeze(1).unsqueeze(-1).unsqueeze(-1) # We can specify head_mask for each layer
head_mask = head_mask.to(dtype=next(self.parameters()).dtype) # switch to fload if need + fp16 compatibility
head_mask = (
head_mask.unsqueeze(1).unsqueeze(-1).unsqueeze(-1)
) # We can specify head_mask for each layer
head_mask = head_mask.to(
dtype=next(self.parameters()).dtype
) # switch to fload if need + fp16 compatibility
else:
head_mask = [None] * self.config.num_hidden_layers
##################################
# Replace this with your model code
embedding_output = self.embeddings(input_ids=input_ids, position_ids=position_ids, token_type_ids=token_type_ids, inputs_embeds=inputs_embeds)
embedding_output = self.embeddings(
input_ids=input_ids, position_ids=position_ids, token_type_ids=token_type_ids, inputs_embeds=inputs_embeds
)
encoder_outputs = self.encoder(embedding_output, extended_attention_mask, head_mask=head_mask)
sequence_output = encoder_outputs[0]
outputs = (sequence_output,) + encoder_outputs[1:] # add hidden_states and attentions if they are here
@@ -357,8 +378,9 @@ class XxxModel(XxxPreTrainedModel):
return outputs # sequence_output, (hidden_states), (attentions)
@add_start_docstrings("""Xxx Model with a `language modeling` head on top. """,
XXX_START_DOCSTRING, XXX_INPUTS_DOCSTRING)
@add_start_docstrings(
"""Xxx Model with a `language modeling` head on top. """, XXX_START_DOCSTRING, XXX_INPUTS_DOCSTRING
)
class XxxForMaskedLM(XxxPreTrainedModel):
r"""
**masked_lm_labels**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``:
@@ -389,6 +411,7 @@ class XxxForMaskedLM(XxxPreTrainedModel):
loss, prediction_scores = outputs[:2]
"""
def __init__(self, config):
super(XxxForMaskedLM, self).__init__(config)
@@ -400,15 +423,25 @@ class XxxForMaskedLM(XxxPreTrainedModel):
def get_output_embeddings(self):
return self.lm_head
def forward(self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None,
masked_lm_labels=None):
def forward(
self,
input_ids=None,
attention_mask=None,
token_type_ids=None,
position_ids=None,
head_mask=None,
inputs_embeds=None,
masked_lm_labels=None,
):
outputs = self.transformer(input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
position_ids=position_ids,
head_mask=head_mask,
inputs_embeds=inputs_embeds)
outputs = self.transformer(
input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
position_ids=position_ids,
head_mask=head_mask,
inputs_embeds=inputs_embeds,
)
sequence_output = outputs[0]
prediction_scores = self.cls(sequence_output)
@@ -422,9 +455,12 @@ class XxxForMaskedLM(XxxPreTrainedModel):
return outputs # (masked_lm_loss), prediction_scores, (hidden_states), (attentions)
@add_start_docstrings("""Xxx Model transformer with a sequence classification/regression head on top (a linear layer on top of
@add_start_docstrings(
"""Xxx Model transformer with a sequence classification/regression head on top (a linear layer on top of
the pooled output) e.g. for GLUE tasks. """,
XXX_START_DOCSTRING, XXX_INPUTS_DOCSTRING)
XXX_START_DOCSTRING,
XXX_INPUTS_DOCSTRING,
)
class XxxForSequenceClassification(XxxPreTrainedModel):
r"""
**labels**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size,)``:
@@ -456,6 +492,7 @@ class XxxForSequenceClassification(XxxPreTrainedModel):
loss, logits = outputs[:2]
"""
def __init__(self, config):
super(XxxForSequenceClassification, self).__init__(config)
self.num_labels = config.num_labels
@@ -466,15 +503,25 @@ class XxxForSequenceClassification(XxxPreTrainedModel):
self.init_weights()
def forward(self, input_ids=None, attention_mask=None, token_type_ids=None,
position_ids=None, head_mask=None, inputs_embeds=None, labels=None):
def forward(
self,
input_ids=None,
attention_mask=None,
token_type_ids=None,
position_ids=None,
head_mask=None,
inputs_embeds=None,
labels=None,
):
outputs = self.transformer(input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
position_ids=position_ids,
head_mask=head_mask,
inputs_embeds=inputs_embeds)
outputs = self.transformer(
input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
position_ids=position_ids,
head_mask=head_mask,
inputs_embeds=inputs_embeds,
)
pooled_output = outputs[1]
@@ -496,9 +543,12 @@ class XxxForSequenceClassification(XxxPreTrainedModel):
return outputs # (loss), logits, (hidden_states), (attentions)
@add_start_docstrings("""Xxx Model with a token classification head on top (a linear layer on top of
@add_start_docstrings(
"""Xxx Model with a token classification head on top (a linear layer on top of
the hidden-states output) e.g. for Named-Entity-Recognition (NER) tasks. """,
XXX_START_DOCSTRING, XXX_INPUTS_DOCSTRING)
XXX_START_DOCSTRING,
XXX_INPUTS_DOCSTRING,
)
class XxxForTokenClassification(XxxPreTrainedModel):
r"""
**labels**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size, sequence_length)``:
@@ -528,6 +578,7 @@ class XxxForTokenClassification(XxxPreTrainedModel):
loss, scores = outputs[:2]
"""
def __init__(self, config):
super(XxxForTokenClassification, self).__init__(config)
self.num_labels = config.num_labels
@@ -538,15 +589,25 @@ class XxxForTokenClassification(XxxPreTrainedModel):
self.init_weights()
def forward(self, input_ids=None, attention_mask=None, token_type_ids=None,
position_ids=None, head_mask=None, inputs_embeds=None, labels=None):
def forward(
self,
input_ids=None,
attention_mask=None,
token_type_ids=None,
position_ids=None,
head_mask=None,
inputs_embeds=None,
labels=None,
):
outputs = self.transformer(input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
position_ids=position_ids,
head_mask=head_mask,
inputs_embeds=inputs_embeds)
outputs = self.transformer(
input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
position_ids=position_ids,
head_mask=head_mask,
inputs_embeds=inputs_embeds,
)
sequence_output = outputs[0]
@@ -569,9 +630,12 @@ class XxxForTokenClassification(XxxPreTrainedModel):
return outputs # (loss), scores, (hidden_states), (attentions)
@add_start_docstrings("""Xxx Model with a span classification head on top for extractive question-answering tasks like SQuAD (a linear layers on top of
@add_start_docstrings(
"""Xxx Model with a span classification head on top for extractive question-answering tasks like SQuAD (a linear layers on top of
the hidden-states output to compute `span start logits` and `span end logits`). """,
XXX_START_DOCSTRING, XXX_INPUTS_DOCSTRING)
XXX_START_DOCSTRING,
XXX_INPUTS_DOCSTRING,
)
class XxxForQuestionAnswering(XxxPreTrainedModel):
r"""
**start_positions**: (`optional`) ``torch.LongTensor`` of shape ``(batch_size,)``:
@@ -613,6 +677,7 @@ class XxxForQuestionAnswering(XxxPreTrainedModel):
"""
def __init__(self, config):
super(XxxForQuestionAnswering, self).__init__(config)
self.num_labels = config.num_labels
@@ -622,15 +687,26 @@ class XxxForQuestionAnswering(XxxPreTrainedModel):
self.init_weights()
def forward(self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None,
start_positions=None, end_positions=None):
def forward(
self,
input_ids=None,
attention_mask=None,
token_type_ids=None,
position_ids=None,
head_mask=None,
inputs_embeds=None,
start_positions=None,
end_positions=None,
):
outputs = self.transformer(input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
position_ids=position_ids,
head_mask=head_mask,
inputs_embeds=inputs_embeds)
outputs = self.transformer(
input_ids,
attention_mask=attention_mask,
token_type_ids=token_type_ids,
position_ids=position_ids,
head_mask=head_mask,
inputs_embeds=inputs_embeds,
)
sequence_output = outputs[0]

173
templates/adding_a_new_model/tests/modeling_tf_xxx_test.py
View File

@@ -19,7 +19,7 @@ from __future__ import print_function
import unittest
import sys
from .modeling_tf_common_test import (TFCommonTestCases, ids_tensor)
from .modeling_tf_common_test import TFCommonTestCases, ids_tensor
from .configuration_common_test import ConfigTester
from .utils import CACHE_DIR, require_tf, slow
@@ -27,46 +27,57 @@ from transformers import XxxConfig, is_tf_available
if is_tf_available():
import tensorflow as tf
from transformers.modeling_tf_xxx import (TFXxxModel, TFXxxForMaskedLM,
TFXxxForSequenceClassification,
TFXxxForTokenClassification,
TFXxxForQuestionAnswering,
TF_XXX_PRETRAINED_MODEL_ARCHIVE_MAP)
from transformers.modeling_tf_xxx import (
TFXxxModel,
TFXxxForMaskedLM,
TFXxxForSequenceClassification,
TFXxxForTokenClassification,
TFXxxForQuestionAnswering,
TF_XXX_PRETRAINED_MODEL_ARCHIVE_MAP,
)
@require_tf
class TFXxxModelTest(TFCommonTestCases.TFCommonModelTester):
all_model_classes = (TFXxxModel, TFXxxForMaskedLM, TFXxxForQuestionAnswering,
TFXxxForSequenceClassification,
TFXxxForTokenClassification) if is_tf_available() else ()
all_model_classes = (
(
TFXxxModel,
TFXxxForMaskedLM,
TFXxxForQuestionAnswering,
TFXxxForSequenceClassification,
TFXxxForTokenClassification,
)
if is_tf_available()
else ()
)
class TFXxxModelTester(object):
def __init__(self,
parent,
batch_size=13,
seq_length=7,
is_training=True,
use_input_mask=True,
use_token_type_ids=True,
use_labels=True,
vocab_size=99,
hidden_size=32,
num_hidden_layers=5,
num_attention_heads=4,
intermediate_size=37,
hidden_act="gelu",
hidden_dropout_prob=0.1,
attention_probs_dropout_prob=0.1,
max_position_embeddings=512,
type_vocab_size=16,
type_sequence_label_size=2,
initializer_range=0.02,
num_labels=3,
num_choices=4,
scope=None,
):
def __init__(
self,
parent,
batch_size=13,
seq_length=7,
is_training=True,
use_input_mask=True,
use_token_type_ids=True,
use_labels=True,
vocab_size=99,
hidden_size=32,
num_hidden_layers=5,
num_attention_heads=4,
intermediate_size=37,
hidden_act="gelu",
hidden_dropout_prob=0.1,
attention_probs_dropout_prob=0.1,
max_position_embeddings=512,
type_vocab_size=16,
type_sequence_label_size=2,
initializer_range=0.02,
num_labels=3,
num_choices=4,
scope=None,
):
self.parent = parent
self.batch_size = batch_size
self.seq_length = seq_length
@@ -120,15 +131,16 @@ class TFXxxModelTest(TFCommonTestCases.TFCommonModelTester):
attention_probs_dropout_prob=self.attention_probs_dropout_prob,
max_position_embeddings=self.max_position_embeddings,
type_vocab_size=self.type_vocab_size,
initializer_range=self.initializer_range)
initializer_range=self.initializer_range,
)
return config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
def create_and_check_xxx_model(self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels):
def create_and_check_xxx_model(
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
):
model = TFXxxModel(config=config)
inputs = {'input_ids': input_ids,
'attention_mask': input_mask,
'token_type_ids': token_type_ids}
inputs = {"input_ids": input_ids, "attention_mask": input_mask, "token_type_ids": token_type_ids}
sequence_output, pooled_output = model(inputs)
inputs = [input_ids, input_mask]
@@ -141,78 +153,74 @@ class TFXxxModelTest(TFCommonTestCases.TFCommonModelTester):
"pooled_output": pooled_output.numpy(),
}
self.parent.assertListEqual(
list(result["sequence_output"].shape),
[self.batch_size, self.seq_length, self.hidden_size])
list(result["sequence_output"].shape), [self.batch_size, self.seq_length, self.hidden_size]
)
self.parent.assertListEqual(list(result["pooled_output"].shape), [self.batch_size, self.hidden_size])
def create_and_check_xxx_for_masked_lm(self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels):
def create_and_check_xxx_for_masked_lm(
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
):
model = TFXxxForMaskedLM(config=config)
inputs = {'input_ids': input_ids,
'attention_mask': input_mask,
'token_type_ids': token_type_ids}
prediction_scores, = model(inputs)
inputs = {"input_ids": input_ids, "attention_mask": input_mask, "token_type_ids": token_type_ids}
(prediction_scores,) = model(inputs)
result = {
"prediction_scores": prediction_scores.numpy(),
}
self.parent.assertListEqual(
list(result["prediction_scores"].shape),
[self.batch_size, self.seq_length, self.vocab_size])
list(result["prediction_scores"].shape), [self.batch_size, self.seq_length, self.vocab_size]
)
def create_and_check_xxx_for_sequence_classification(self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels):
def create_and_check_xxx_for_sequence_classification(
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
):
config.num_labels = self.num_labels
model = TFXxxForSequenceClassification(config=config)
inputs = {'input_ids': input_ids,
'attention_mask': input_mask,
'token_type_ids': token_type_ids}
logits, = model(inputs)
inputs = {"input_ids": input_ids, "attention_mask": input_mask, "token_type_ids": token_type_ids}
(logits,) = model(inputs)
result = {
"logits": logits.numpy(),
}
self.parent.assertListEqual(
list(result["logits"].shape),
[self.batch_size, self.num_labels])
self.parent.assertListEqual(list(result["logits"].shape), [self.batch_size, self.num_labels])
def create_and_check_xxx_for_token_classification(self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels):
def create_and_check_xxx_for_token_classification(
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
):
config.num_labels = self.num_labels
model = TFXxxForTokenClassification(config=config)
inputs = {'input_ids': input_ids,
'attention_mask': input_mask,
'token_type_ids': token_type_ids}
logits, = model(inputs)
inputs = {"input_ids": input_ids, "attention_mask": input_mask, "token_type_ids": token_type_ids}
(logits,) = model(inputs)
result = {
"logits": logits.numpy(),
}
self.parent.assertListEqual(
list(result["logits"].shape),
[self.batch_size, self.seq_length, self.num_labels])
list(result["logits"].shape), [self.batch_size, self.seq_length, self.num_labels]
)
def create_and_check_xxx_for_question_answering(self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels):
def create_and_check_xxx_for_question_answering(
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
):
model = TFXxxForQuestionAnswering(config=config)
inputs = {'input_ids': input_ids,
'attention_mask': input_mask,
'token_type_ids': token_type_ids}
inputs = {"input_ids": input_ids, "attention_mask": input_mask, "token_type_ids": token_type_ids}
start_logits, end_logits = model(inputs)
result = {
"start_logits": start_logits.numpy(),
"end_logits": end_logits.numpy(),
}
self.parent.assertListEqual(
list(result["start_logits"].shape),
[self.batch_size, self.seq_length])
self.parent.assertListEqual(
list(result["end_logits"].shape),
[self.batch_size, self.seq_length])
self.parent.assertListEqual(list(result["start_logits"].shape), [self.batch_size, self.seq_length])
self.parent.assertListEqual(list(result["end_logits"].shape), [self.batch_size, self.seq_length])
def prepare_config_and_inputs_for_common(self):
config_and_inputs = self.prepare_config_and_inputs()
(config, input_ids, token_type_ids, input_mask,
sequence_labels, token_labels, choice_labels) = config_and_inputs
inputs_dict = {'input_ids': input_ids, 'token_type_ids': token_type_ids, 'attention_mask': input_mask}
(
config,
input_ids,
token_type_ids,
input_mask,
sequence_labels,
token_labels,
choice_labels,
) = config_and_inputs
inputs_dict = {"input_ids": input_ids, "token_type_ids": token_type_ids, "attention_mask": input_mask}
return config, inputs_dict
def setUp(self):
@@ -244,9 +252,10 @@ class TFXxxModelTest(TFCommonTestCases.TFCommonModelTester):
@slow
def test_model_from_pretrained(self):
for model_name in ['xxx-base-uncased']:
for model_name in ["xxx-base-uncased"]:
model = TFXxxModel.from_pretrained(model_name, cache_dir=CACHE_DIR)
self.assertIsNotNone(model)
if __name__ == "__main__":
unittest.main()

166
templates/adding_a_new_model/tests/modeling_xxx_test.py
View File

@@ -20,51 +20,60 @@ import unittest
from transformers import is_torch_available
from .modeling_common_test import (CommonTestCases, ids_tensor)
from .modeling_common_test import CommonTestCases, ids_tensor
from .configuration_common_test import ConfigTester
from .utils import CACHE_DIR, require_torch, slow, torch_device
if is_torch_available():
from transformers import (XxxConfig, XxxModel, XxxForMaskedLM,
XxxForNextSentencePrediction, XxxForPreTraining,
XxxForQuestionAnswering, XxxForSequenceClassification,
XxxForTokenClassification, XxxForMultipleChoice)
from transformers import (
XxxConfig,
XxxModel,
XxxForMaskedLM,
XxxForNextSentencePrediction,
XxxForPreTraining,
XxxForQuestionAnswering,
XxxForSequenceClassification,
XxxForTokenClassification,
XxxForMultipleChoice,
)
from transformers.modeling_xxx import XXX_PRETRAINED_MODEL_ARCHIVE_MAP
@require_torch
class XxxModelTest(CommonTestCases.CommonModelTester):
all_model_classes = (XxxModel, XxxForMaskedLM, XxxForQuestionAnswering,
XxxForSequenceClassification,
XxxForTokenClassification) if is_torch_available() else ()
all_model_classes = (
(XxxModel, XxxForMaskedLM, XxxForQuestionAnswering, XxxForSequenceClassification, XxxForTokenClassification)
if is_torch_available()
else ()
)
class XxxModelTester(object):
def __init__(self,
parent,
batch_size=13,
seq_length=7,
is_training=True,
use_input_mask=True,
use_token_type_ids=True,
use_labels=True,
vocab_size=99,
hidden_size=32,
num_hidden_layers=5,
num_attention_heads=4,
intermediate_size=37,
hidden_act="gelu",
hidden_dropout_prob=0.1,
attention_probs_dropout_prob=0.1,
max_position_embeddings=512,
type_vocab_size=16,
type_sequence_label_size=2,
initializer_range=0.02,
num_labels=3,
num_choices=4,
scope=None,
):
def __init__(
self,
parent,
batch_size=13,
seq_length=7,
is_training=True,
use_input_mask=True,
use_token_type_ids=True,
use_labels=True,
vocab_size=99,
hidden_size=32,
num_hidden_layers=5,
num_attention_heads=4,
intermediate_size=37,
hidden_act="gelu",
hidden_dropout_prob=0.1,
attention_probs_dropout_prob=0.1,
max_position_embeddings=512,
type_vocab_size=16,
type_sequence_label_size=2,
initializer_range=0.02,
num_labels=3,
num_choices=4,
scope=None,
):
self.parent = parent
self.batch_size = batch_size
self.seq_length = seq_length
@@ -118,16 +127,17 @@ class XxxModelTest(CommonTestCases.CommonModelTester):
attention_probs_dropout_prob=self.attention_probs_dropout_prob,
max_position_embeddings=self.max_position_embeddings,
type_vocab_size=self.type_vocab_size,
initializer_range=self.initializer_range)
initializer_range=self.initializer_range,
)
return config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
def check_loss_output(self, result):
self.parent.assertListEqual(
list(result["loss"].size()),
[])
self.parent.assertListEqual(list(result["loss"].size()), [])
def create_and_check_xxx_model(self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels):
def create_and_check_xxx_model(
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
):
model = XxxModel(config=config)
model.to(torch_device)
model.eval()
@@ -140,83 +150,98 @@ class XxxModelTest(CommonTestCases.CommonModelTester):
"pooled_output": pooled_output,
}
self.parent.assertListEqual(
list(result["sequence_output"].size()),
[self.batch_size, self.seq_length, self.hidden_size])
list(result["sequence_output"].size()), [self.batch_size, self.seq_length, self.hidden_size]
)
self.parent.assertListEqual(list(result["pooled_output"].size()), [self.batch_size, self.hidden_size])
def create_and_check_xxx_for_masked_lm(self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels):
def create_and_check_xxx_for_masked_lm(
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
):
model = XxxForMaskedLM(config=config)
model.to(torch_device)
model.eval()
loss, prediction_scores = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, masked_lm_labels=token_labels)
loss, prediction_scores = model(
input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, masked_lm_labels=token_labels
)
result = {
"loss": loss,
"prediction_scores": prediction_scores,
}
self.parent.assertListEqual(
list(result["prediction_scores"].size()),
[self.batch_size, self.seq_length, self.vocab_size])
list(result["prediction_scores"].size()), [self.batch_size, self.seq_length, self.vocab_size]
)
self.check_loss_output(result)
def create_and_check_xxx_for_question_answering(self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels):
def create_and_check_xxx_for_question_answering(
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
):
model = XxxForQuestionAnswering(config=config)
model.to(torch_device)
model.eval()
loss, start_logits, end_logits = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids,
start_positions=sequence_labels, end_positions=sequence_labels)
loss, start_logits, end_logits = model(
input_ids,
attention_mask=input_mask,
token_type_ids=token_type_ids,
start_positions=sequence_labels,
end_positions=sequence_labels,
)
result = {
"loss": loss,
"start_logits": start_logits,
"end_logits": end_logits,
}
self.parent.assertListEqual(
list(result["start_logits"].size()),
[self.batch_size, self.seq_length])
self.parent.assertListEqual(
list(result["end_logits"].size()),
[self.batch_size, self.seq_length])
self.parent.assertListEqual(list(result["start_logits"].size()), [self.batch_size, self.seq_length])
self.parent.assertListEqual(list(result["end_logits"].size()), [self.batch_size, self.seq_length])
self.check_loss_output(result)
def create_and_check_xxx_for_sequence_classification(self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels):
def create_and_check_xxx_for_sequence_classification(
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
):
config.num_labels = self.num_labels
model = XxxForSequenceClassification(config)
model.to(torch_device)
model.eval()
loss, logits = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=sequence_labels)
loss, logits = model(
input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=sequence_labels
)
result = {
"loss": loss,
"logits": logits,
}
self.parent.assertListEqual(
list(result["logits"].size()),
[self.batch_size, self.num_labels])
self.parent.assertListEqual(list(result["logits"].size()), [self.batch_size, self.num_labels])
self.check_loss_output(result)
def create_and_check_xxx_for_token_classification(self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels):
def create_and_check_xxx_for_token_classification(
self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
):
config.num_labels = self.num_labels
model = XxxForTokenClassification(config=config)
model.to(torch_device)
model.eval()
loss, logits = model(input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=token_labels)
loss, logits = model(
input_ids, attention_mask=input_mask, token_type_ids=token_type_ids, labels=token_labels
)
result = {
"loss": loss,
"logits": logits,
}
self.parent.assertListEqual(
list(result["logits"].size()),
[self.batch_size, self.seq_length, self.num_labels])
list(result["logits"].size()), [self.batch_size, self.seq_length, self.num_labels]
)
self.check_loss_output(result)
def prepare_config_and_inputs_for_common(self):
config_and_inputs = self.prepare_config_and_inputs()
(config, input_ids, token_type_ids, input_mask,
sequence_labels, token_labels, choice_labels) = config_and_inputs
inputs_dict = {'input_ids': input_ids, 'token_type_ids': token_type_ids, 'attention_mask': input_mask}
(
config,
input_ids,
token_type_ids,
input_mask,
sequence_labels,
token_labels,
choice_labels,
) = config_and_inputs
inputs_dict = {"input_ids": input_ids, "token_type_ids": token_type_ids, "attention_mask": input_mask}
return config, inputs_dict
def setUp(self):
@@ -252,5 +277,6 @@ class XxxModelTest(CommonTestCases.CommonModelTester):
model = XxxModel.from_pretrained(model_name, cache_dir=CACHE_DIR)
self.assertIsNotNone(model)
if __name__ == "__main__":
unittest.main()

30
templates/adding_a_new_model/tests/tokenization_xxx_test.py
View File

@@ -18,10 +18,11 @@ import os
import unittest
from io import open
from transformers.tokenization_bert import (XxxTokenizer, VOCAB_FILES_NAMES)
from transformers.tokenization_bert import XxxTokenizer, VOCAB_FILES_NAMES
from .tokenization_tests_commons import CommonTestCases
class XxxTokenizationTest(CommonTestCases.CommonTokenizerTester):
tokenizer_class = XxxTokenizer
@@ -30,28 +31,39 @@ class XxxTokenizationTest(CommonTestCases.CommonTokenizerTester):
super(XxxTokenizationTest, self).setUp()
vocab_tokens = [
"[UNK]", "[CLS]", "[SEP]", "want", "##want", "##ed", "wa", "un", "runn",
"##ing", ",", "low", "lowest",
"[UNK]",
"[CLS]",
"[SEP]",
"want",
"##want",
"##ed",
"wa",
"un",
"runn",
"##ing",
",",
"low",
"lowest",
]
self.vocab_file = os.path.join(self.tmpdirname, VOCAB_FILES_NAMES['vocab_file'])
with open(self.vocab_file, "w", encoding='utf-8') as vocab_writer:
self.vocab_file = os.path.join(self.tmpdirname, VOCAB_FILES_NAMES["vocab_file"])
with open(self.vocab_file, "w", encoding="utf-8") as vocab_writer:
vocab_writer.write("".join([x + "\n" for x in vocab_tokens]))
def get_tokenizer(self, **kwargs):
return XxxTokenizer.from_pretrained(self.tmpdirname, **kwargs)
def get_input_output_texts(self):
input_text = u"UNwant\u00E9d,running"
output_text = u"unwanted, running"
input_text = "UNwant\u00E9d,running"
output_text = "unwanted, running"
return input_text, output_text
def test_full_tokenizer(self):
tokenizer = self.tokenizer_class(self.vocab_file)
tokens = tokenizer.tokenize(u"UNwant\u00E9d,running")
tokens = tokenizer.tokenize("UNwant\u00E9d,running")
self.assertListEqual(tokens, ["un", "##want", "##ed", ",", "runn", "##ing"])
self.assertListEqual(tokenizer.convert_tokens_to_ids(tokens), [7, 4, 5, 10, 8, 9])
if __name__ == '__main__':
if __name__ == "__main__":
unittest.main()

65
templates/adding_a_new_model/tokenization_xxx.py
View File

@@ -34,17 +34,16 @@ logger = logging.getLogger(__name__)
# Mapping from the keyword arguments names of Tokenizer `__init__`
# to file names for serializing Tokenizer instances
####################################################
VOCAB_FILES_NAMES = {'vocab_file': 'vocab.txt'}
VOCAB_FILES_NAMES = {"vocab_file": "vocab.txt"}
####################################################
# Mapping from the keyword arguments names of Tokenizer `__init__`
# to pretrained vocabulary URL for all the model shortcut names.
####################################################
PRETRAINED_VOCAB_FILES_MAP = {
'vocab_file':
{
'xxx-base-uncased': "https://s3.amazonaws.com/models.huggingface.co/bert/xxx-base-uncased-vocab.txt",
'xxx-large-uncased': "https://s3.amazonaws.com/models.huggingface.co/bert/xxx-large-uncased-vocab.txt",
"vocab_file": {
"xxx-base-uncased": "https://s3.amazonaws.com/models.huggingface.co/bert/xxx-base-uncased-vocab.txt",
"xxx-large-uncased": "https://s3.amazonaws.com/models.huggingface.co/bert/xxx-large-uncased-vocab.txt",
}
}
@@ -52,8 +51,8 @@ PRETRAINED_VOCAB_FILES_MAP = {
# Mapping from model shortcut names to max length of inputs
####################################################
PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES = {
'xxx-base-uncased': 512,
'xxx-large-uncased': 512,
"xxx-base-uncased": 512,
"xxx-large-uncased": 512,
}
####################################################
@@ -62,8 +61,8 @@ PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES = {
# To be used for checkpoint specific configurations.
####################################################
PRETRAINED_INIT_CONFIGURATION = {
'xxx-base-uncased': {'do_lower_case': True},
'xxx-large-uncased': {'do_lower_case': True},
"xxx-base-uncased": {"do_lower_case": True},
"xxx-large-uncased": {"do_lower_case": True},
}
@@ -73,7 +72,7 @@ def load_vocab(vocab_file):
with open(vocab_file, "r", encoding="utf-8") as reader:
tokens = reader.readlines()
for index, token in enumerate(tokens):
token = token.rstrip('\n')
token = token.rstrip("\n")
vocab[token] = index
return vocab
@@ -93,9 +92,17 @@ class XxxTokenizer(PreTrainedTokenizer):
pretrained_init_configuration = PRETRAINED_INIT_CONFIGURATION
max_model_input_sizes = PRETRAINED_POSITIONAL_EMBEDDINGS_SIZES
def __init__(self, vocab_file, do_lower_case=True,
unk_token="[UNK]", sep_token="[SEP]", pad_token="[PAD]", cls_token="[CLS]",
mask_token="[MASK]", **kwargs):
def __init__(
self,
vocab_file,
do_lower_case=True,
unk_token="[UNK]",
sep_token="[SEP]",
pad_token="[PAD]",
cls_token="[CLS]",
mask_token="[MASK]",
**kwargs
):
"""Constructs a XxxTokenizer.
Args:
@@ -104,16 +111,22 @@ class XxxTokenizer(PreTrainedTokenizer):
Whether to lower case the input
Only has an effect when do_basic_tokenize=True
"""
super(XxxTokenizer, self).__init__(unk_token=unk_token, sep_token=sep_token,
pad_token=pad_token, cls_token=cls_token,
mask_token=mask_token, **kwargs)
super(XxxTokenizer, self).__init__(
unk_token=unk_token,
sep_token=sep_token,
pad_token=pad_token,
cls_token=cls_token,
mask_token=mask_token,
**kwargs
)
self.max_len_single_sentence = self.max_len - 2 # take into account special tokens
self.max_len_sentences_pair = self.max_len - 3 # take into account special tokens
if not os.path.isfile(vocab_file):
raise ValueError(
"Can't find a vocabulary file at path '{}'. To load the vocabulary from a Google pretrained "
"model use `tokenizer = XxxTokenizer.from_pretrained(PRETRAINED_MODEL_NAME)`".format(vocab_file))
"model use `tokenizer = XxxTokenizer.from_pretrained(PRETRAINED_MODEL_NAME)`".format(vocab_file)
)
self.vocab = load_vocab(vocab_file)
@property
@@ -142,7 +155,7 @@ class XxxTokenizer(PreTrainedTokenizer):
def convert_tokens_to_string(self, tokens):
""" Converts a sequence of tokens (string) in a single string. """
out_string = ' '.join(tokens).replace(' ##', '').strip()
out_string = " ".join(tokens).replace(" ##", "").strip()
return out_string
def build_inputs_with_special_tokens(self, token_ids_0, token_ids_1=None):
@@ -177,8 +190,10 @@ class XxxTokenizer(PreTrainedTokenizer):
if already_has_special_tokens:
if token_ids_1 is not None:
raise ValueError("You should not supply a second sequence if the provided sequence of "
"ids is already formated with special tokens for the model.")
raise ValueError(
"You should not supply a second sequence if the provided sequence of "
"ids is already formated with special tokens for the model."
)
return list(map(lambda x: 1 if x in [self.sep_token_id, self.cls_token_id] else 0, token_ids_0))
if token_ids_1 is not None:
@@ -204,15 +219,17 @@ class XxxTokenizer(PreTrainedTokenizer):
"""Save the tokenizer vocabulary to a directory or file."""
index = 0
if os.path.isdir(vocab_path):
vocab_file = os.path.join(vocab_path, VOCAB_FILES_NAMES['vocab_file'])
vocab_file = os.path.join(vocab_path, VOCAB_FILES_NAMES["vocab_file"])
else:
vocab_file = vocab_path
with open(vocab_file, "w", encoding="utf-8") as writer:
for token, token_index in sorted(self.vocab.items(), key=lambda kv: kv[1]):
if index != token_index:
logger.warning("Saving vocabulary to {}: vocabulary indices are not consecutive."
" Please check that the vocabulary is not corrupted!".format(vocab_file))
logger.warning(
"Saving vocabulary to {}: vocabulary indices are not consecutive."
" Please check that the vocabulary is not corrupted!".format(vocab_file)
)
index = token_index
writer.write(token + u'\n')
writer.write(token + "\n")
index += 1
return (vocab_file,)