mirror of
https://github.com/hpcaitech/ColossalAI.git
synced 2025-05-05 23:18:22 +00:00
96780e6ee4
* add pipeline shared module wrapper and update load batch * added model parallel process group for amp and clip grad (#86) * added model parallel process group for amp and clip grad * update amp and clip with model parallel process group * remove pipeline_prev/next group (#88) * micro batch offload * optimize pipeline gpu memory usage * pipeline can receive tensor shape (#93) * optimize pipeline gpu memory usage * fix grad accumulation step counter * rename classes and functions Co-authored-by: Frank Lee <somerlee.9@gmail.com>
40 lines
1.5 KiB
Python
40 lines
1.5 KiB
Python
#!/usr/bin/env python
|
|
# -*- encoding: utf-8 -*-
|
|
|
|
from torch import distributed as dist
|
|
|
|
from colossalai.registry import DIST_GROUP_INITIALIZER
|
|
from .process_group_initializer import ProcessGroupInitializer
|
|
from ..parallel_mode import ParallelMode
|
|
|
|
|
|
@DIST_GROUP_INITIALIZER.register_module
|
|
class Initializer_Pipeline(ProcessGroupInitializer):
|
|
def __init__(self, *args, **kwargs):
|
|
super().__init__(*args, **kwargs)
|
|
self.data_group_size = self.world_size // self.data_parallel_size
|
|
self.pipeline_stage_size = self.data_group_size // self.pipeline_parallel_size
|
|
|
|
def init_dist_group(self):
|
|
dist_settings = list()
|
|
for i in range(self.data_parallel_size):
|
|
for j in range(self.pipeline_stage_size):
|
|
pipe_ranks = list(
|
|
range(i * self.data_group_size + j,
|
|
(i + 1) * self.data_group_size,
|
|
self.pipeline_stage_size))
|
|
pipe_group_size = len(pipe_ranks)
|
|
pipe_group = dist.new_group(pipe_ranks)
|
|
|
|
if self.rank in pipe_ranks:
|
|
local_rank = pipe_ranks.index(self.rank)
|
|
group_world_size = pipe_group_size
|
|
process_group = pipe_group
|
|
ranks_in_group = pipe_ranks
|
|
dist_settings.append(
|
|
tuple((local_rank, group_world_size,
|
|
process_group, ranks_in_group,
|
|
ParallelMode.PIPELINE)))
|
|
|
|
return dist_settings
|