[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>

colossalai/initialize.py

@@ -6,12 +6,12 @@ import warnings
 from pathlib import Path
 from typing import Dict, Union
 
-import torch
 import torch.distributed as dist
 
+from colossalai.accelerator import get_accelerator
 from colossalai.context import Config
 from colossalai.logging import get_dist_logger
-from colossalai.utils import IS_NPU_AVAILABLE, set_device, set_seed
+from colossalai.utils import set_seed
 
 
 def launch(
@@ -47,17 +47,18 @@ def launch(
     if rank == 0:
         warnings.warn("`config` is deprecated and will be removed soon.")
 
-    if IS_NPU_AVAILABLE and backend == "nccl":
-        backend = "hccl"
+    cur_accelerator = get_accelerator()
+
+    backend = cur_accelerator.communication_backend
 
     # init default process group
     init_method = f"tcp://[{host}]:{port}"
     dist.init_process_group(rank=rank, world_size=world_size, backend=backend, init_method=init_method)
 
     # set cuda device
-    if torch.cuda.is_available() or IS_NPU_AVAILABLE:
-        # if local rank is not given, calculate automatically
-        set_device(local_rank)
+    # if local rank is not given, calculate automatically
+    if cur_accelerator.support_set_device:
+        cur_accelerator.set_device(local_rank)
 
     set_seed(seed)