Traceback (most recent call last):
File "/mnt/petrelfs/wangzhao/HumanT2V/AnimateAnyone/train_svd.py", line 803, in <module>
main(config)
File "/mnt/petrelfs/wangzhao/HumanT2V/AnimateAnyone/train_svd.py", line 672, in main
accelerator.clip_grad_norm_(
File "/mnt/petrelfs/wangzhao/anaconda3/envs/consisti2v/lib/python3.10/site-packages/accelerate/accelerator.py", line 1925, in clip_grad_norm_
self.unscale_gradients()
File "/mnt/petrelfs/wangzhao/anaconda3/envs/consisti2v/lib/python3.10/site-packages/accelerate/accelerator.py", line 1888, in unscale_gradients
self.scaler.unscale_(opt)
File "/mnt/petrelfs/wangzhao/anaconda3/envs/consisti2v/lib/python3.10/site-packages/torch/cuda/amp/grad_scaler.py", line 284, in unscale_
optimizer_state["found_inf_per_device"] = self._unscale_grads_(optimizer, inv_scale, found_inf, False)
File "/mnt/petrelfs/wangzhao/anaconda3/envs/consisti2v/lib/python3.10/site-packages/torch/cuda/amp/grad_scaler.py", line 212, in _unscale_grads_
raise ValueError("Attempting to unscale FP16 gradients.")
ValueError: Attempting to unscale FP16 gradients.
Get an error during stage 2.