diff --git a/intx_sft.py b/intx_sft.py index 3ea9448..8971c07 100755 --- a/intx_sft.py +++ b/intx_sft.py @@ -313,6 +313,7 @@ def main(): ctx_encoder_model_config = ctx_encoder_model_config.text_config ctx_tokenizer = get_tokenizer(ctx_name) else: + ctx_name = model.base_model.config.name_or_path ctx_encoder_model_config = model.config ctx_tokenizer = tokenizer @@ -519,7 +520,7 @@ def main(): _apply_liger_kernel_to_instance(model=model.base_model.base_model.model) if ctx_name is not None: logger.info("Applying liger-kernel to ctx_encoder_model") - _apply_liger_kernel_to_instance(model=model.ctx_encoder) + _apply_liger_kernel_to_instance(model=model.ctx_encoder.base_model) elif isinstance(model, PeftModel): logger.info("Applying liger-kernel to PeftModel") _apply_liger_kernel_to_instance(model=model.base_model.model)