Hotfix/Colossalai layers (#92)

* optimized 1d layer apis; reorganized nn.layer modules; fixed tests

* fixed 2.5d runtime issue

* reworked split batch, now called in trainer.schedule.load_batch

Co-authored-by: BoxiangW <45734921+BoxiangW@users.noreply.github.com>
This commit is contained in:
アマデウス
2021-12-29 23:32:10 +08:00
committed by GitHub
parent 0fedef4f3c
commit 01a80cd86d
71 changed files with 1033 additions and 773 deletions

View File

@@ -1,15 +1,16 @@
#!/usr/bin/env python
# -*- encoding: utf-8 -*-
from functools import partial
from pathlib import Path
import pytest
import torch
import torch.multiprocessing as mp
from colossalai import launch
from colossalai.context.parallel_mode import ParallelMode
from colossalai.core import global_context as gpc
from functools import partial
from pathlib import Path
from colossalai.utils import free_port
CONFIG_PATH = Path(__file__).parent.joinpath('configs/parallel_2d_init.py').absolute()
@@ -87,7 +88,7 @@ def test_2d_init():
test_fn = partial(init_2d,
world_size=world_size,
backend='gloo',
port='29900',
port=free_port(),
host='localhost'
)
mp.spawn(test_fn, nprocs=world_size)

View File

@@ -7,10 +7,10 @@ from pathlib import Path
import pytest
import torch
import torch.multiprocessing as mp
from colossalai.context.parallel_mode import ParallelMode
from colossalai.core import global_context as gpc
from colossalai.initialize import launch
from colossalai.utils import free_port
CONFIG_PATH = Path(__file__).parent.joinpath('configs/parallel_2p5d_init.py').absolute()
@@ -111,7 +111,7 @@ def test_2halfd_init():
test_fn = partial(init_2halfd,
world_size=world_size,
backend='gloo',
port='29901',
port=free_port(),
host='localhost'
)
mp.spawn(test_fn, nprocs=world_size)

View File

@@ -7,11 +7,10 @@ from pathlib import Path
import pytest
import torch
import torch.multiprocessing as mp
from colossalai.context.parallel_mode import ParallelMode
from colossalai.core import global_context as gpc
from colossalai.initialize import launch
from colossalai.utils import free_port
CONFIG_PATH = Path(__file__).parent.joinpath('configs/parallel_3d_init.py').absolute()
@@ -104,7 +103,7 @@ def test_3d_init():
test_fn = partial(init_3d,
world_size=world_size,
backend='gloo',
port='29902',
port=free_port(),
host='localhost'
)
mp.spawn(test_fn, nprocs=world_size)