github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-09-28 04:55:25 +00:00
Code Issues Packages Projects Releases Wiki Activity

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

Browse Source 
* 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
Download Patch File Download Diff File Expand all files Collapse all files

10
colossalai/legacy/zero/shard_utils/tensor_shard_strategy.py
View File

@@ -3,11 +3,11 @@ from typing import List, Optional
import torch
import torch.distributed as dist
from colossalai.accelerator import get_accelerator
from colossalai.legacy.zero.gemini.tensor_utils import colo_model_data_tensor_move_inline
from colossalai.legacy.zero.shard_utils import BaseShardStrategy
from colossalai.legacy.zero.shard_utils.commons import get_shard
from colossalai.legacy.zero.sharded_param.sharded_tensor import ShardedTensor
from colossalai.utils import get_current_device
class TensorShardStrategy(BaseShardStrategy):
@@ -34,9 +34,9 @@ class TensorShardStrategy(BaseShardStrategy):
if t.is_sharded:
return
if t.payload.device.type == "cuda":
assert t.payload.device == get_current_device(), (
assert t.payload.device == get_accelerator().get_current_device(), (
f"shard tensor on cuda device index {t.payload.device.index},"
f" but current cuda device is {get_current_device()}"
f" but current cuda device is {get_accelerator().get_current_device()}"
)
sharded_payload, _ = get_shard(t.payload, dist.get_rank(process_group), dist.get_world_size(process_group))
t.payload_reset(sharded_payload)
@@ -50,7 +50,9 @@ class TensorShardStrategy(BaseShardStrategy):
world_size = dist.get_world_size(process_group)
rank = dist.get_rank(process_group)
buffer = torch.empty(payload_numel * world_size, dtype=t.payload.dtype, device=get_current_device())
buffer = torch.empty(
payload_numel * world_size, dtype=t.payload.dtype, device=get_accelerator().get_current_device()
)
buffer_list = list(torch.chunk(buffer, chunks=world_size, dim=0))
buffer_list[rank].copy_(t.payload)