[legacy] move communication and nn to legacy and refactor logger (#4671)

* [legacy] move communication to legacy (#4640)

* [legacy] refactor logger and clean up legacy codes (#4654)

* [legacy] make logger independent to gpc

* [legacy] make optim independent to registry

* [legacy] move test engine to legacy

* [legacy] move nn to legacy (#4656)

* [legacy] move nn to legacy

* [checkpointio] fix save hf config

* [test] remove useledd rpc pp test

* [legacy] fix nn init

* [example] skip tutorial hybriad parallel example

* [devops] test doc check

* [devops] test doc check
This commit is contained in:
Hongxin Liu
2023-09-11 16:24:28 +08:00
committed by GitHub
parent 536397cc95
commit 554aa9592e
170 changed files with 781 additions and 758 deletions

View File

@@ -0,0 +1,57 @@
import pytest
import torch
from checks_2p5d.check_layer_2p5d import *
from checks_2p5d.check_operation_2p5d import check_AB, check_ABT, check_ATB
from colossalai.core import global_context as gpc
from colossalai.initialize import launch
from colossalai.logging import disable_existing_loggers
from colossalai.testing import rerun_if_address_is_in_use, spawn
CONFIG = dict(parallel=dict(
pipeline=dict(size=1),
tensor=dict(size=4, mode='2.5d', depth=1),
),)
def check_operations():
check_AB()
check_ABT()
check_ATB()
def check_layer():
check_linear()
check_layernorm()
check_embed()
check_patch_embed()
check_vocab_parallel_embed()
check_classifier_no_given_weight()
check_vocab_parallel_classifier_no_given_weight()
check_classifier_given_embed_weight()
check_vocab_parallel_classifier_given_embed_weight()
check_loss()
check_vocab_parallel_loss()
def check_layer_and_operation(rank, world_size, port):
disable_existing_loggers()
launch(config=CONFIG, rank=rank, world_size=world_size, host='localhost', port=port, backend='nccl')
torch.backends.cuda.matmul.allow_tf32 = False
torch.backends.cudnn.allow_tf32 = False
torch.backends.cudnn.deterministic = True
check_operations()
check_layer()
gpc.destroy()
torch.cuda.empty_cache()
@pytest.mark.dist
@rerun_if_address_is_in_use()
def test_2p5d():
spawn(check_layer_and_operation, 4)
if __name__ == '__main__':
test_2p5d()