According to this answer for similar issue with tensorflow, it could occur because the VRAM memory limit was hit (which is rather non-intuitive from the error message).
For my case with PyTorch model training, decreasing batch size helped. You could try this or maybe decrease your model size to consume less VRAM.