[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>
This commit is contained in:
Hongxin Liu
2024-01-09 10:20:05 +08:00
committed by GitHub
parent dd2c28a323
commit d202cc28c0
128 changed files with 1773 additions and 868 deletions

View File

@@ -8,7 +8,8 @@ import pytest
import torch
from torch import Tensor
from colossalai.utils import get_current_device, multi_tensor_applier
from colossalai.accelerator import get_accelerator
from colossalai.utils import multi_tensor_applier
_FUSED_ALLOWED_P_G_TYPES = [
(torch.float, torch.half),
@@ -155,7 +156,9 @@ def test_fused_adam_kernel(adamw, weight_decay, p_dtype, g_dtype):
rtol, atol = 1e-3, 1e-3
if p_dtype is torch.bfloat16 or g_dtype is torch.bfloat16:
rtol, atol = 4e-3, 4e-3
check_adam_kernel(FusedAdamKernel, adamw, weight_decay, p_dtype, g_dtype, get_current_device(), 3, rtol, atol)
check_adam_kernel(
FusedAdamKernel, adamw, weight_decay, p_dtype, g_dtype, get_accelerator().get_current_device(), 3, rtol, atol
)
@pytest.mark.parametrize("adamw", [False, True])