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

Truncate max length if needed in all examples (#10034)

Browse Source
This commit is contained in:
Sylvain Gugger
2021-02-08 05:03:55 -05:00
committed by GitHub
parent 45aaf5f7ab
commit b01483faa0
6 changed files with 68 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
examples/language-modeling/run_mlm.py
View File

@@ -303,6 +303,22 @@ def main():
column_names = datasets["validation"].column_names
text_column_name = "text" if "text" in column_names else column_names[0]
if data_args.max_seq_length is None:
max_seq_length = tokenizer.model_max_length
if max_seq_length > 1024:
logger.warn(
f"The tokenizer picked seems to have a very large `model_max_length` ({tokenizer.model_max_length}). "
"Picking 1024 instead. You can change that default value by passing --max_seq_length xxx."
)
max_seq_length = 1024
else:
if data_args.max_seq_length > tokenizer.model_max_length:
logger.warn(
f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
)
max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)
if data_args.line_by_line:
# When using line_by_line, we just tokenize each nonempty line.
padding = "max_length" if data_args.pad_to_max_length else False
@@ -314,7 +330,7 @@ def main():
examples["text"],
padding=padding,
truncation=True,
max_length=data_args.max_seq_length,
max_length=max_seq_length,
# We use this option because DataCollatorForLanguageModeling (see below) is more efficient when it
# receives the `special_tokens_mask`.
return_special_tokens_mask=True,
@@ -342,22 +358,6 @@ def main():
load_from_cache_file=not data_args.overwrite_cache,
)
if data_args.max_seq_length is None:
max_seq_length = tokenizer.model_max_length
if max_seq_length > 1024:
logger.warn(
f"The tokenizer picked seems to have a very large `model_max_length` ({tokenizer.model_max_length}). "
"Picking 1024 instead. You can change that default value by passing --max_seq_length xxx."
)
max_seq_length = 1024
else:
if data_args.max_seq_length > tokenizer.model_max_length:
logger.warn(
f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
)
max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)
# Main data processing function that will concatenate all texts from our dataset and generate chunks of
# max_seq_length.
def group_texts(examples):

16
examples/language-modeling/run_plm.py
View File

@@ -300,6 +300,13 @@ def main():
column_names = datasets["validation"].column_names
text_column_name = "text" if "text" in column_names else column_names[0]
if data_args.max_seq_length > tokenizer.model_max_length:
logger.warn(
f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
)
max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)
if data_args.line_by_line:
# When using line_by_line, we just tokenize each nonempty line.
padding = "max_length" if data_args.pad_to_max_length else False
@@ -307,7 +314,7 @@ def main():
def tokenize_function(examples):
# Remove empty lines
examples["text"] = [line for line in examples["text"] if len(line) > 0 and not line.isspace()]
return tokenizer(examples["text"], padding=padding, truncation=True, max_length=data_args.max_seq_length)
return tokenizer(examples["text"], padding=padding, truncation=True, max_length=max_seq_length)
tokenized_datasets = datasets.map(
tokenize_function,
@@ -329,13 +336,6 @@ def main():
load_from_cache_file=not data_args.overwrite_cache,
)
if data_args.max_seq_length > tokenizer.model_max_length:
logger.warn(
f"The max_seq_length passed ({data_args.max_seq_length}) is larger than the maximum length for the"
f"model ({tokenizer.model_max_length}). Using max_seq_length={tokenizer.model_max_length}."
)
max_seq_length = min(data_args.max_seq_length, tokenizer.model_max_length)
# Main data processing function that will concatenate all texts from our dataset and generate chunks of
# max_seq_length.
def group_texts(examples):