Skip to content

Download models on first tokenizer call instead of module import #180

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
carlthome wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Download models on first tokenizer call instead of module import #180

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 10 additions & 6 deletions src/laion_clap/training/data.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -41,20 +41,22 @@
except ImportError:
torchaudio = None

bert_tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")
roberta_tokenizer = RobertaTokenizer.from_pretrained("roberta-base")
bart_tokenizer = BartTokenizer.from_pretrained("facebook/bart-base")
bert_tokenizer = None
roberta_tokenizer = None
bart_tokenizer = None

def tokenizer(text, tmodel="roberta", max_length=77):
"""tokenizer for different models
tmodel is default to roberta as it is the best model for our task
max_length is default to 77 from the OpenAI CLIP parameters
We assume text to be a single string, but it can also be a list of strings
"""
global bert_tokenizer, roberta_tokenizer, bart_tokenizer
if tmodel == "transformer":
return clip_tokenizer(text).squeeze(0)

elif tmodel == "bert":
if bert_tokenizer is None:
bert_tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")
result = bert_tokenizer(
text,
padding="max_length",
Expand All @@ -63,8 +65,9 @@ def tokenizer(text, tmodel="roberta", max_length=77):
return_tensors="pt",
)
return {k: v.squeeze(0) for k, v in result.items()}

elif tmodel == "roberta":
if roberta_tokenizer is None:
roberta_tokenizer = RobertaTokenizer.from_pretrained("roberta-base")
result = roberta_tokenizer(
text,
padding="max_length",
Expand All @@ -73,8 +76,9 @@ def tokenizer(text, tmodel="roberta", max_length=77):
return_tensors="pt",
)
return {k: v.squeeze(0) for k, v in result.items()}

elif tmodel == "bart":
if bart_tokenizer is None:
bart_tokenizer = BartTokenizer.from_pretrained("facebook/bart-base")
result = bart_tokenizer(
text,
padding="max_length",
Expand Down