About fine-tuning on a new dataset

[bioanalysisSC]

Thank you for the excellent work!
I am very interested in applying your model to a new dataset. In order to further enhance performance, I am attempting to fine-tune the existing large model using the training methods mentioned in your documentation. However, I encountered the following error:
RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation: [torch.cuda.FloatTensor [8, 64, 266]] is at version 1; expected version 0 instead. Hint: the backtrace further above shows the operation that failed to compute its gradient. The variable in question was changed in there or anywhere later. Good luck!
ERROR conda.cli.main_run:execute(124): conda run python /vol/home/xxx/code/scFoundation/RNA_pretraining.py failed. (See above for error)
I have utilized torch.autograd.set_detect_anomaly(True) to help identify the issue, and the error was pinpointed to the operation data_dash = torch.einsum('...id,...jd->...ij', (data_normalizer * data), projection).
Could you please offer some guidance or suggest possible solutions to resolve this error? Any advice would be greatly appreciated.

Thank you for your time and assistance.