From c25f83c85f10a14ac9443838ee355258d2eb1d3f Mon Sep 17 00:00:00 2001 From: Edenzzzz Date: Mon, 6 May 2024 18:17:26 +0800 Subject: [PATCH] fix missing pad token (#5690) Co-authored-by: Edenzzzz --- examples/language/gpt/hybridparallelism/finetune.py | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/examples/language/gpt/hybridparallelism/finetune.py b/examples/language/gpt/hybridparallelism/finetune.py index 32b2dfcc0..777d16cb9 100644 --- a/examples/language/gpt/hybridparallelism/finetune.py +++ b/examples/language/gpt/hybridparallelism/finetune.py @@ -243,7 +243,12 @@ def main(): # ==================================== # gpt2 pretrained model - cfg = AutoConfig.from_pretrained(model_name, num_labels=data_builder.num_labels) + cfg = AutoConfig.from_pretrained( + model_name, + num_labels=data_builder.num_labels, + pad_token=data_builder.tokenizer.pad_token, + pad_token_id=data_builder.tokenizer.pad_token_id, + ) if model_name == "gpt2": model = GPT2ForSequenceClassification.from_pretrained(model_name, config=cfg).cuda()