Do not warn about unexpected decoder weights when loading T5EncoderMo…

…del and LongT5EncoderModel (huggingface#26211)

Ignore decoder weights when using T5EncoderModel and LongT5EncoderModel

Both T5EncoderModel and LongT5EncoderModel do not have any decoder layers, so
loading a pretrained model checkpoint such as t5-small will give warnings about
keys found in the model checkpoint that are not in the model itself.

To prevent this log warning, r"decoder" has been added to _keys_to_ignore_on_load_unexpected for
both T5EncoderModel and LongT5EncoderModel

src/transformers/models/longt5/modeling_longt5.py

@@ -2152,6 +2152,7 @@ def _reorder_cache(self, past_key_values, beam_idx):
 )
 class LongT5EncoderModel(LongT5PreTrainedModel):
     _tied_weights_keys = ["encoder.embed_tokens.weight"]
+    _keys_to_ignore_on_load_unexpected = [r"decoder"]
 
     def __init__(self, config: LongT5Config):
         super().__init__(config)

src/transformers/models/t5/modeling_t5.py

@@ -1866,6 +1866,7 @@ def _reorder_cache(self, past_key_values, beam_idx):
 )
 class T5EncoderModel(T5PreTrainedModel):
     _tied_weights_keys = ["encoder.embed_tokens.weight"]
+    _keys_to_ignore_on_load_unexpected = [r"decoder"]
 
     def __init__(self, config: T5Config):
         super().__init__(config)