[Hoxfix] Fix CUDA_DEVICE_MAX_CONNECTIONS for comm overlap

Co-authored-by: Edenzzzz <wtan45@wisc.edu>
pull/5894/head
Edenzzzz 5 months ago committed by GitHub
parent 3420921101
commit 8ec24b6a4d
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194

@ -3,6 +3,12 @@
import os import os
# set CUDA_DEVICE_MAX_CONNECTIONS=1 to ensure that when overlapping communication and computation,
# the order of of kernel launches on GPUs are the same as on the CPU so that comm is launched first.
# see https://github.com/NVIDIA/Megatron-LM/issues/533
# https://forums.developer.nvidia.com/t/how-many-streams-maximum-number-of-streams/6571/16
os.environ["CUDA_DEVICE_MAX_CONNECTIONS"] = "1"
import torch.distributed as dist import torch.distributed as dist
from colossalai.accelerator import get_accelerator from colossalai.accelerator import get_accelerator

@ -81,7 +81,6 @@ class LinearWithAsyncCommunication(torch.autograd.Function):
handle = dist.all_reduce(grad_input, group=gpc.get_group(ctx.parallel_mode), async_op=True) handle = dist.all_reduce(grad_input, group=gpc.get_group(ctx.parallel_mode), async_op=True)
# Delay the start of weight gradient computation shortly (3us) to have # Delay the start of weight gradient computation shortly (3us) to have
# all-reduce scheduled first and have GPU resources allocated # all-reduce scheduled first and have GPU resources allocated
_ = torch.empty(1, device=grad_output.device) + 1
grad_weight = grad_output.t().matmul(total_input) grad_weight = grad_output.t().matmul(total_input)
grad_bias = grad_output.sum(dim=0) if use_bias else None grad_bias = grad_output.sum(dim=0) if use_bias else None

@ -1,4 +1,3 @@
import os
from typing import Dict, List, Tuple from typing import Dict, List, Tuple
import torch.distributed as dist import torch.distributed as dist
@ -11,9 +10,6 @@ from ..policies.base_policy import Policy
from .shard_config import ShardConfig from .shard_config import ShardConfig
from .sharder import ModelSharder from .sharder import ModelSharder
# set CUDA_DEVICE_MAX_CONNECTIONS=1 to ensure that when communication and computation overlap, the order of core scheduling is correct
os.environ["CUDA_DEVICE_MAX_CONNECTIONS"] = "1"
class ShardFormer: class ShardFormer:
""" """

@ -292,7 +292,7 @@ def main():
with get_profile_context( with get_profile_context(
args.profile, args.profile,
args.ignore_steps, args.ignore_steps,
len(dataloader) - 1, 1, # avoid creating massive log files
save_dir=f"profile/{time.strftime('%H:%M', time.localtime())}-{args.plugin}-llama-{args.config}", save_dir=f"profile/{time.strftime('%H:%M', time.localtime())}-{args.plugin}-llama-{args.config}",
) as prof: ) as prof:
if isinstance(plugin, HybridParallelPlugin) and args.pp > 1: if isinstance(plugin, HybridParallelPlugin) and args.pp > 1:

Loading…
Cancel
Save