[misc] refactor launch API and tensor constructor (#5666)

* [misc] remove config arg from initialize

* [misc] remove old tensor contrusctor

* [plugin] add npu support for ddp

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* [devops] fix doc test ci

* [test] fix test launch

* [doc] update launch doc

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>

colossalai/auto_parallel/offload/amp_optimizer.py

@@ -126,7 +126,7 @@ class AMPOptimizer(OptimizerWrapper):
         return self.grad_scaler.scale.item()
 
     def zero_grad(self, *args, **kwargs):
-        self.module.overflow_counter = torch.cuda.IntTensor([0])
+        self.module.overflow_counter = torch.tensor([0], dtype=torch.int, device=get_accelerator().get_current_device())
         return self.optim.zero_grad(set_to_none=True)
 
     def step(self, *args, **kwargs):

colossalai/auto_parallel/offload/base_offload_module.py

@@ -4,7 +4,7 @@ from typing import Optional, Set
 import torch
 import torch.nn as nn
 
-from colossalai.utils import _cast_float
+from colossalai.utils import _cast_float, get_current_device
 from colossalai.utils.common import free_storage
 
 from .region_manager import RegionManager
@@ -25,7 +25,7 @@ class BaseOffloadModule:
         self.model = model
         self.region_manager = region_manager
         self.grad_hook_list = []
-        self.overflow_counter = torch.cuda.IntTensor([0])
+        self.overflow_counter = torch.tensor([0], dtype=torch.int, device=get_current_device())
 
         self.grad_offload_stream = torch.cuda.current_stream() if is_sync else GlobalRuntimeInfo.d2h_stream