Error using Intel PyTorch extension for Arc GPU on Ubuntu

[SoldierWz]

🐛 Describe the bug

---

RuntimeError Traceback (most recent call last)
Cell In[6], line 107
105 outputs = model(features)
106 loss = criterion(outputs, labels)
--> 107 loss.backward()
108 optimizer.step()
109 epoch_train_loss += loss.item() * len(features) # 记录训练损失

File ~/mambaforge/envs/pytorch-arc/lib/python3.11/site-packages/torch/_tensor.py:492, in Tensor.backward(self, gradient, retain_graph, create_graph, inputs)
482 if has_torch_function_unary(self):
483 return handle_torch_function(
484 Tensor.backward,
485 (self,),
(...)
490 inputs=inputs,
491 )
--> 492 torch.autograd.backward(
493 self, gradient, retain_graph, create_graph, inputs=inputs
494 )

File ~/mambaforge/envs/pytorch-arc/lib/python3.11/site-packages/torch/autograd/init.py:251, in backward(tensors, grad_tensors, retain_graph, create_graph, grad_variables, inputs)
246 retain_graph = create_graph
248 # The reason we repeat the same comment below is that
249 # some Python versions print out the first line of a multi-line function
...
257 allow_unreachable=True,
258 accumulate_grad=True,
259 )

RuntimeError: 0 <= device.index() && device.index() < static_castc10::DeviceIndex(device_ready_queues_.size()) INTERNAL ASSERT FAILED at "/build/pytorch/torch/csrc/autograd/engine.cpp":1418, please report a bug to PyTorch.

Versions

wget https://raw.githubusercontent.com/pytorch/pytorch/main/torch/utils/collect_env.py

For security purposes, please check the contents of collect_env.py before running it.

python collect_env.py

cc @frank-wei @jgong5 @mingfeima @XiaobingSuper @sanchitintel @ashokei @jingxu10

[ZhaoqiongZ]

Hi @SoldierWz , Could you offer a reproducible script ?

[malfet]

Please file your issue to https://github.com/intel/intel-extension-for-pytorch
but cc: @jgong5

[SoldierWz]

Please file your issue to https://github.com/intel/intel-extension-for-pytorch but cc: @jgong5

I have already submitted at intel/intel-extension-for-pytorch#571