t5gemma_text_encoder.py:
- Added custom type LLM_ATTENTION_MASK as a return type
- Changed encode_text function to not use a chat template since T5Gemma doesn't use one.
- Applied attention_mask to inputs of encode_text
t5gemma_apply_llm_to_sdxl_adapter.py:
- Added separate apply_llm_to_sdxl_adapter node for t5gemma since it uses an attention mask.
- Also allows for comparison of old code vs new code.
__init__.py:
- Added mandatory setup code for new node.
- Added custom type "LLM_ATTENTION_MASK".
NeuroSenko