Update gpt2.py

remove vocab size slicing for korean gpt2 model

Update gpt2.py
remove vocab size slicing for korean gpt2 model
e265bd7c · Jiwung Hyun · GitHub · 3670b32e · e265bd7c
Unverified Commit e265bd7c authored Sep 04, 2022 by Jiwung Hyun Committed by GitHub Sep 04, 2022
Hide whitespace changes
Inline Side-by-side

Showing with 1 addition and 1 deletion

lm_eval/models/gpt2.py lm_eval/models/gpt2.py +1 -1

No files found.
--- a/lm_eval/models/gpt2.py
+++ b/lm_eval/models/gpt2.py
@@ -88,7 +88,7 @@ class HFLM(BaseLM):
        logits returned from the model
        """
        with torch.no_grad():
-            return self.gpt2(inps)[0][:, :, :50257]
+            return self.gpt2(inps)[0]

    def _model_generate(self, context, max_length, eos_token_id):
        return self.gpt2.generate(