Absent of target_embed in run_inference()

[midannii]

Hi. I was so impressed with your work since it models both MPNN and BERT. When I ran train.py, I faced some issues when run_inference().

About running https://github.com/Elanmarkowitz/StATIK/blob/main/train.py#L342, there is an error since target_embeddings (for input of model() ) is None.

I tentatively think that https://github.com/Elanmarkowitz/StATIK/blob/main/train.py#L335 ~ https://github.com/Elanmarkowitz/StATIK/blob/main/train.py#L347 is for making target_embeds for https://github.com/Elanmarkowitz/StATIK/blob/main/train.py#L352 ~ https://github.com/Elanmarkowitz/StATIK/blob/main/train.py#L378.

In my opinion, it is rational that target_embeddings is None for https://github.com/Elanmarkowitz/StATIK/blob/main/train.py#L335 ~ https://github.com/Elanmarkowitz/StATIK/blob/main/train.py#L347

I don't know how to solve this. I might be very grateful if someone help me ;<

My settings are below

• torch : 1.12.1+cuda11.3
• python : 3.8
• transformers : 4.38.2
• remove all codes about dist since I am not able to run multi-GPU settings (especially for port and address).