alternative more elegant way of setting pad token for gpt2 generate

lm_eval/models/gpt2.py

@@ -112,7 +112,8 @@ class HFLM(BaseLM):
         generation_kwargs = {'do_sample': False, 'max_length': max_length}
         if eos_token_id is not None:
             generation_kwargs['eos_token_id'] = eos_token_id
-        return self.gpt2.generate(context, pad_token_id=eos_token_id, **generation_kwargs)
+            generation_kwargs['pad_token_id'] = eos_token_id
+        return self.gpt2.generate(context, **generation_kwargs)
 
 # for backwards compatibility
 GPT2LM = HFLM