Tensors of the same index must be on the same device and the same dtype except `step` tensors that can be CPU and float32 notwithstanding

[wurevvc]

When I was training, this error occurred

[NitzanHod]

Please reply if you managed to solve this issue!

The error is raised on the optimizer.step() when running the fine tune script as described in the README on a fresh new cloned repository, i suspect this is due to errors in newer pytorch versions, see pytorch/pytorch#127197
I get this error with torch==2.5.1.
I attach the stack below:

{'loss': 1.0453, 'grad_norm': 1.2838969230651855, 'learning_rate': 1.5557084630007206e-05, 'epoch': 1.0}
33%|███▎ | 406/1218 [24:06<48:52, 3.61s/it]Traceback (most recent call last):
File "/home/h/stanford_alpaca/train.py", line 222, in
Traceback (most recent call last):
File "/home/h/stanford_alpaca/train.py", line 222, in
Traceback (most recent call last):
File "/home/h/stanford_alpaca/train.py", line 222, in
train()
File "/home/h/stanford_alpaca/train.py", line 216, in train
train()
File "/home/h/stanford_alpaca/train.py", line 216, in train
trainer.train()
File "/home/h/miniconda3/lib/python3.9/site-packages/transformers/trainer.py", line 2123, in train
train()
File "/home/h/stanford_alpaca/train.py", line 216, in train
trainer.train()
File "/home/h/miniconda3/lib/python3.9/site-packages/transformers/trainer.py", line 2123, in train
trainer.train()
File "/home/h/miniconda3/lib/python3.9/site-packages/transformers/trainer.py", line 2123, in train
return inner_training_loop(
File "/home/h/miniconda3/lib/python3.9/site-packages/transformers/trainer.py", line 2534, in _inner_training_loop
return inner_training_loop(
File "/home/h/miniconda3/lib/python3.9/site-packages/transformers/trainer.py", line 2534, in _inner_training_loop
return inner_training_loop(
File "/home/h/miniconda3/lib/python3.9/site-packages/transformers/trainer.py", line 2534, in _inner_training_loop
self.optimizer.step()
File "/home/h/miniconda3/lib/python3.9/site-packages/accelerate/optimizer.py", line 149, in step
self.optimizer.step()
File "/home/h/miniconda3/lib/python3.9/site-packages/accelerate/optimizer.py", line 149, in step
self.optimizer.step()
File "/home/h/miniconda3/lib/python3.9/site-packages/accelerate/optimizer.py", line 149, in step
self.optimizer.step(closure)
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/lr_scheduler.py", line 75, in wrapper
self.optimizer.step(closure)self.optimizer.step(closure)

File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/lr_scheduler.py", line 75, in wrapper
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/lr_scheduler.py", line 75, in wrapper
return wrapped(*args, **kwargs)
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/optimizer.py", line 385, in wrapper
return wrapped(*args, **kwargs)return wrapped(*args, **kwargs)

File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/optimizer.py", line 385, in wrapper
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/optimizer.py", line 385, in wrapper
out = func(*args, **kwargs)
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/optimizer.py", line 76, in _use_grad
out = func(*args, **kwargs)out = func(*args, **kwargs)

File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/optimizer.py", line 76, in _use_grad
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/optimizer.py", line 76, in _use_grad
ret = func(self, *args, **kwargs)
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/adamw.py", line 187, in step
ret = func(self, *args, **kwargs)
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/adamw.py", line 187, in step
ret = func(self, *args, **kwargs)
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/adamw.py", line 187, in step
adamw(
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/adamw.py", line 339, in adamw
adamw(
adamw(
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/adamw.py", line 339, in adamw
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/adamw.py", line 339, in adamw
func(
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/adamw.py", line 516, in _multi_tensor_adamw
func(func(

File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/adamw.py", line 516, in _multi_tensor_adamw
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/adamw.py", line 516, in _multi_tensor_adamw
grouped_tensors = Optimizer._group_tensors_by_device_and_dtype([
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/optimizer.py", line 409, in _group_tensors_by_device_and_dtype
grouped_tensors = Optimizer._group_tensors_by_device_and_dtype([
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/optimizer.py", line 409, in _group_tensors_by_device_and_dtype
grouped_tensors = Optimizer._group_tensors_by_device_and_dtype([
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/optimizer.py", line 409, in _group_tensors_by_device_and_dtype
return _group_tensors_by_device_and_dtype(tensorlistlist, with_indices)
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return _group_tensors_by_device_and_dtype(tensorlistlist, with_indices)
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return _group_tensors_by_device_and_dtype(tensorlistlist, with_indices)
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/utils/_foreach_utils.py", line 38, in _group_tensors_by_device_and_dtype
return func(*args, **kwargs)return func(*args, **kwargs)

File "/home/h/miniconda3/lib/python3.9/site-packages/torch/utils/_foreach_utils.py", line 38, in _group_tensors_by_device_and_dtype
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/utils/_foreach_utils.py", line 38, in _group_tensors_by_device_and_dtype
torch._C._group_tensors_by_device_and_dtype(tensorlistlist, with_indices).items()
RuntimeError: Tensors of the same index must be on the same device and the same dtype except step tensors that can be CPU and float32 notwithstanding
torch._C._group_tensors_by_device_and_dtype(tensorlistlist, with_indices).items()torch._C._group_tensors_by_device_and_dtype(tensorlistlist, with_indices).items()

RuntimeError: Tensors of the same index must be on the same device and the same dtype except step tensors that can be CPU and float32 notwithstanding
RuntimeError: Traceback (most recent call last):
Tensors of the same index must be on the same device and the same dtype except step tensors that can be CPU and float32 notwithstanding
File "/home/h/stanford_alpaca/train.py", line 222, in
train()
File "/home/h/stanford_alpaca/train.py", line 216, in train
trainer.train()
File "/home/h/miniconda3/lib/python3.9/site-packages/transformers/trainer.py", line 2123, in train
return inner_training_loop(
File "/home/h/miniconda3/lib/python3.9/site-packages/transformers/trainer.py", line 2534, in _inner_training_loop
self.optimizer.step()
File "/home/h/miniconda3/lib/python3.9/site-packages/accelerate/optimizer.py", line 149, in step
self.optimizer.step(closure)
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/lr_scheduler.py", line 75, in wrapper
return wrapped(*args, **kwargs)
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/optimizer.py", line 385, in wrapper
out = func(*args, **kwargs)
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/optimizer.py", line 76, in _use_grad
ret = func(self, *args, **kwargs)
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/adamw.py", line 187, in step
adamw(
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/adamw.py", line 339, in adamw
func(
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/adamw.py", line 516, in _multi_tensor_adamw
grouped_tensors = Optimizer._group_tensors_by_device_and_dtype([
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/optim/optimizer.py", line 409, in _group_tensors_by_device_and_dtype
return _group_tensors_by_device_and_dtype(tensorlistlist, with_indices)
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "/home/h/miniconda3/lib/python3.9/site-packages/torch/utils/_foreach_utils.py", line 38, in _group_tensors_by_device_and_dtype
torch._C._group_tensors_by_device_and_dtype(tensorlistlist, with_indices).items()
RuntimeError: Tensors of the same index must be on the same device and the same dtype except step tensors that can be CPU and float32 notwithstanding

[NitzanHod]

For me the issue was solved by downgrading the transformers package to transformers==4.28.1 (and using torch==2.5.1)