[npu] change device to accelerator api (#5239)

* update accelerator

* fix timer

* fix amp

* update

* fix

* update bug

* add error raise

* fix autocast

* fix set device

* remove doc accelerator

* update doc

* update doc

* update doc

* use nullcontext

* update cpu

* update null context

* change time limit for example

* udpate

* update

* update

* update

* [npu] polish accelerator code

---------

Co-authored-by: Xuanlei Zhao <xuanlei.zhao@gmail.com>
Co-authored-by: zxl <43881818+oahzxl@users.noreply.github.com>

tests/test_legacy/test_trainer/test_pipeline/test_p2p.py

@@ -5,6 +5,7 @@ import pytest
 import torch
 import torch.distributed as dist
 
+from colossalai.accelerator import get_accelerator
 from colossalai.legacy.communication import (
     recv_backward,
     recv_forward,
@@ -18,7 +19,6 @@ from colossalai.legacy.core import global_context as gpc
 from colossalai.legacy.initialize import launch
 from colossalai.logging import get_dist_logger
 from colossalai.testing import rerun_if_address_is_in_use, spawn
-from colossalai.utils import get_current_device
 
 BATCH_SIZE = 4
 SEQ_LENGTH = 2
@@ -73,7 +73,7 @@ def check_forward_backward(output_tensor, output_grad, rank, logger):
 
 def check_comm(size, rank, prev_rank, next_rank, logger):
     dtype = torch.float32
-    device = get_current_device()
+    device = get_accelerator().get_current_device()
     tensor_shape = (BATCH_SIZE, SEQ_LENGTH, HIDDEN_SIZE)
     grad_shape = (BATCH_SIZE, SEQ_LENGTH, HIDDEN_SIZE)
     tensor = torch.randn(tensor_shape, dtype=dtype, device=device)