youngkingdom/vllm
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[Core][Test] move local_rank to the last arg with default value(#3711)

Browse Source 
[Core][Test] move local_rank to the last arg with default value to keep api compatible (#3711)
This commit is contained in:
youkaichao
2024-03-28 21:19:45 -07:00
committed by GitHub
parent 395aa823ea
commit 756b30a5f3
7 changed files with 24 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
tests/distributed/test_comm_ops.py
View File

@ -24,7 +24,7 @@ def all_reduce_test_worker(tensor_parallel_size: int, rank: int,
del os.environ["CUDA_VISIBLE_DEVICES"]
device = torch.device(f"cuda:{rank}")
torch.cuda.set_device(device)
init_test_distributed_environment(1, tensor_parallel_size, rank, rank,
init_test_distributed_environment(1, tensor_parallel_size, rank,
distributed_init_port)
num_elements = 8
all_tensors = [
@ -46,7 +46,7 @@ def all_gather_test_worker(tensor_parallel_size: int, rank: int,
del os.environ["CUDA_VISIBLE_DEVICES"]
device = torch.device(f"cuda:{rank}")
torch.cuda.set_device(device)
init_test_distributed_environment(1, tensor_parallel_size, rank, rank,
init_test_distributed_environment(1, tensor_parallel_size, rank,
distributed_init_port)
num_dimensions = 3
tensor_size = list(range(2, num_dimensions + 2))
@ -74,7 +74,7 @@ def broadcast_tensor_dict_test_worker(tensor_parallel_size: int, rank: int,
del os.environ["CUDA_VISIBLE_DEVICES"]
device = torch.device(f"cuda:{rank}")
torch.cuda.set_device(device)
init_test_distributed_environment(1, tensor_parallel_size, rank, rank,
init_test_distributed_environment(1, tensor_parallel_size, rank,
distributed_init_port)
test_dict = {
"a": torch.arange(8, dtype=torch.float32, device="cuda"),

4
tests/distributed/test_custom_all_reduce.py
View File

@ -23,7 +23,7 @@ def graph_allreduce(world_size, rank, distributed_init_port):
del os.environ["CUDA_VISIBLE_DEVICES"]
device = torch.device(f"cuda:{rank}")
torch.cuda.set_device(device)
init_test_distributed_environment(1, world_size, rank, rank,
init_test_distributed_environment(1, world_size, rank,
distributed_init_port)
custom_ar.init_custom_ar()
@ -58,7 +58,7 @@ def eager_allreduce(world_size, rank, distributed_init_port):
del os.environ["CUDA_VISIBLE_DEVICES"]
device = torch.device(f"cuda:{rank}")
torch.cuda.set_device(device)
init_test_distributed_environment(1, world_size, rank, rank,
init_test_distributed_environment(1, world_size, rank,
distributed_init_port)
sz = 1024

2
tests/distributed/test_pynccl.py
View File

@ -14,7 +14,9 @@ def distributed_run(fn, world_size):
for i in range(number_of_processes):
env = os.environ.copy()
env['RANK'] = str(i)
env['LOCAL_RANK'] = str(i)
env['WORLD_SIZE'] = str(number_of_processes)
env['LOCAL_WORLD_SIZE'] = str(number_of_processes)
env['MASTER_ADDR'] = 'localhost'
env['MASTER_PORT'] = '12345'
p = multiprocessing.Process(target=fn, args=(env, ))