Raising `assert param.grad is not None` when finetuning LoRA.

[HashimotoPatrickMu]

While fine-tuning LoRA using PyTorch Lightning, I consistently encounter an assert param.grad is not None error during backpropagation. Increasing the gradient accumulation steps delays the issue but doesn't resolve it. I suspect this might be a backend problem. Has anyone else experienced this issue?

hi,i have the same question.Did you solve it?