Google Git
Sign in
android / platform / external / pytorch / 35aeb6aa8533cf9bb88c0f8375b9ba8be20b2425 / . / test / distributed / test_c10d_nccl.py
blob: f82909a2222c9d632e59bc6c97c7800284b87bd6 [file] [log] [blame]
# Owner(s): ["oncall: distributed"]
import copy
import math
import os
import random
import re
import signal
import sys
import tempfile
import threading
from contextlib import contextmanager
from datetime import timedelta
from itertools import chain, product
from unittest import mock
import torch
import torch.distributed as c10d
if not c10d.is_available() or not c10d.is_nccl_available():
print("c10d NCCL not available, skipping tests", file=sys.stderr)
sys.exit(0)
import test_c10d_common
import torch.distributed as dist
import torch.distributed.algorithms.ddp_comm_hooks.default_hooks as default
import torch.distributed.algorithms.ddp_comm_hooks.powerSGD_hook as powerSGD
import torch.nn.functional as F
import torch.testing._internal.common_utils as common
from test_c10d_common import gpus_for_rank, DoubleGpuNet, ConvNet, ModuleForDdpCommHook
from torch import nn
from torch._C._distributed_c10d import OpType
from torch.nn.parallel import DistributedDataParallel
from torch.testing._internal.common_distributed import (
MultiProcessTestCase,
init_multigpu_helper,
requires_nccl,
requires_gloo,
requires_nccl_version,
skip_if_lt_x_gpu,
get_timeout,
skip_if_rocm,
with_dist_debug_levels,
with_nccl_blocking_wait,
)
from torch.testing._internal.common_utils import (
TestCase,
run_tests,
retry_on_connect_failures,
TEST_WITH_DEV_DBG_ASAN,
TEST_WITH_ROCM,
skip_but_pass_in_sandcastle,
skip_but_pass_in_sandcastle_if,
)
if TEST_WITH_DEV_DBG_ASAN:
print(
"Skip ASAN as torch + multiprocessing spawn have known issues", file=sys.stderr
)
sys.exit(0)
# bfloat16 is only supported by CUDA 11+
BFLOAT16_AVAILABLE = (
torch.cuda.is_available()
and
(
(torch.version.cuda is not None and int(torch.version.cuda.split('.')[0]) >= 11)
or torch.version.hip is not None
)
)
class RendezvousEnvTest(TestCase):
@retry_on_connect_failures
@requires_nccl()
@skip_but_pass_in_sandcastle_if(
torch.cuda.device_count() == 0, "No GPUs available, skipping test"
)
def test_common_errors(self):
vars = {
"WORLD_SIZE": "1",
"RANK": "0",
"MASTER_ADDR": "127.0.0.1",
"MASTER_PORT": str(common.find_free_port()),
}
class Env:
def __init__(self, vars):
self.env_patcher = mock.patch.dict(os.environ, vars, clear=True)
def __enter__(self):
self.env_patcher.start()
def __exit__(self, type, value, traceback):
self.env_patcher.stop()
def without(d, key):
d = d.copy()
d.pop(key)
return d
def withouts(d, keys):
d = d.copy()
for key in keys:
d.pop(key)
return d
with Env(without(vars, "WORLD_SIZE")):
self.assertEqual(None, os.environ.get("WORLD_SIZE"))
with self.assertRaisesRegex(ValueError, "WORLD_SIZE expected"):
gen = c10d.rendezvous("env://")
next(gen)
c10d.init_process_group(backend="nccl", world_size=1)
self.assertEqual(c10d.get_rank(), 0)
self.assertEqual(c10d.get_world_size(), 1)
c10d.destroy_process_group()
with Env(without(vars, "RANK")):
self.assertEqual(None, os.environ.get("RANK"))
with self.assertRaisesRegex(ValueError, "RANK expected"):
gen = c10d.rendezvous("env://")
next(gen)
c10d.init_process_group(backend="nccl", rank=0)
self.assertEqual(c10d.get_rank(), 0)
self.assertEqual(c10d.get_world_size(), 1)
c10d.destroy_process_group()
with Env(withouts(vars, ["RANK", "WORLD_SIZE"])):
self.assertEqual(None, os.environ.get("RANK"))
self.assertEqual(None, os.environ.get("WORLD_SIZE"))
c10d.init_process_group(backend="nccl", rank=0, world_size=1)
self.assertEqual(c10d.get_rank(), 0)
self.assertEqual(c10d.get_world_size(), 1)
c10d.destroy_process_group()
with Env(vars):
c10d.init_process_group(backend="nccl")
self.assertEqual(c10d.get_rank(), 0)
self.assertEqual(c10d.get_world_size(), 1)
c10d.destroy_process_group()
with Env(without(vars, "MASTER_ADDR")):
self.assertEqual(None, os.environ.get("MASTER_ADDR"))
with self.assertRaisesRegex(ValueError, "MASTER_ADDR expected"):
gen = c10d.rendezvous("env://")
next(gen)
with Env(without(vars, "MASTER_PORT")):
self.assertEqual(None, os.environ.get("MASTER_PORT"))
with self.assertRaisesRegex(ValueError, "MASTER_PORT expected"):
gen = c10d.rendezvous("env://")
next(gen)
with Env(without(vars, "WORLD_SIZE")):
self.assertEqual(None, os.environ.get("WORLD_SIZE"))
gen = c10d.rendezvous(f"env://?world_size={1}")
_, _, size = next(gen)
self.assertEqual(size, 1)
with Env(without(vars, "RANK")):
self.assertEqual(None, os.environ.get("RANK"))
gen = c10d.rendezvous(f"env://?rank={0}")
_, rank, _ = next(gen)
self.assertEqual(rank, 0)
with Env(withouts(vars, ["RANK", "WORLD_SIZE"])):
self.assertEqual(None, os.environ.get("RANK"))
self.assertEqual(None, os.environ.get("WORLD_SIZE"))
gen = c10d.rendezvous(f"env://?rank={0}&world_size={1}")
_, rank, size = next(gen)
self.assertEqual(rank, 0)
self.assertEqual(size, 1)
class TimeoutTest(test_c10d_common.AbstractTimeoutTest, TestCase):
@requires_nccl()
@retry_on_connect_failures
@skip_but_pass_in_sandcastle_if(
torch.cuda.device_count() == 0, "No GPUs available, skipping test"
)
def test_default_store_timeout_nccl(self):
self._test_default_store_timeout("nccl")
class ProcessGroupNCCLNoGPUTest(TestCase):
MAIN_PROCESS_RANK = 0
def setUp(self):
self.rank = self.MAIN_PROCESS_RANK
self.world_size = 1
self.file = tempfile.NamedTemporaryFile(delete=False)
def tearDown(self):
pass
@requires_nccl()
@skip_but_pass_in_sandcastle_if(
torch.cuda.device_count() > 0, "GPUs are available, skipping test"
)
def test_init_no_gpus(self):
store = c10d.FileStore(self.file.name, self.world_size)
with self.assertRaisesRegex(
RuntimeError, "ProcessGroupNCCL is only supported with GPUs, no GPUs found!"
):
c10d.ProcessGroupNCCL(store, self.rank, self.world_size)
class ProcessGroupNCCLTest(MultiProcessTestCase):
def _create_process_group_nccl(self, store, opts):
# create nccl processgroup with opts
c10d.init_process_group(
"nccl",
world_size=self.world_size,
rank=self.rank,
store=store,
pg_options=opts)
pg = c10d.distributed_c10d._get_default_group()
return pg
def opts(self, high_priority_stream=False):
opts = c10d.ProcessGroupNCCL.Options()
opts.is_high_priority_stream = high_priority_stream
return opts
def setUp(self):
super().setUp()
# NCCL_BLOCKING_WAIT overrides NCCL_ASYNC_ERROR_HANDLING hence tests
# that use NCCL_BLOCKING_WAIT will test it as expected.
os.environ["NCCL_ASYNC_ERROR_HANDLING"] = "1"
# self.num_gpus = torch.cuda.device_count()
self._spawn_processes()
def tearDown(self):
super().tearDown()
try:
os.remove(self.file_name)
except OSError:
pass
@property
def world_size(self):
return 2
@property
def rank_to_GPU(self):
# return rank to GPU map
return init_multigpu_helper(self.world_size, "nccl")
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_empty_tensors(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
local_device_idx = self.rank_to_GPU[self.rank][0]
xs = [torch.FloatTensor([]).cuda(local_device_idx)]
pg.broadcast(xs).wait()
self.assertEqual(0, xs[0].numel())
pg.allreduce(xs).wait()
self.assertEqual(0, xs[0].numel())
pg.reduce(xs).wait()
self.assertEqual(0, xs[0].numel())
ys = [[torch.FloatTensor([]).cuda(local_device_idx) for _ in range(self.world_size)]]
pg.allgather(ys, xs).wait()
for y in ys[0]:
self.assertEqual(0, y.numel())
ys = [torch.FloatTensor([]).cuda(local_device_idx)]
xs = [[torch.FloatTensor([]).cuda(local_device_idx) for _ in range(self.world_size)]]
pg.reduce_scatter(ys, xs).wait()
self.assertEqual(0, ys[0].numel())
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_broadcast_ops(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
def broadcast(xs, rootRank, rootTensor):
opts = c10d.BroadcastOptions()
opts.rootRank = rootRank
opts.rootTensor = rootTensor
work = pg.broadcast(xs, opts)
work.wait()
return work.result()
# Every rank is root once
for i in range(self.world_size):
# Run with 1 input tensor
x = torch.tensor([self.rank]).cuda(self.rank_to_GPU[self.rank][0])
output = broadcast([x], i, 0)
self.assertEqual(torch.tensor([i]), output[0])
expected_tensor = torch.empty([i + 1, i + 1]).fill_(i + 1)
xs = [torch.empty([i + 1, i + 1]).fill_(-1).cuda(device=device_idx) for device_idx in self.rank_to_GPU[self.rank]]
# test with multiple input tensors (multiple gpu in one rank)
for j in range(len(xs)):
if self.rank == i:
xs[j] = expected_tensor.cuda(device=self.rank_to_GPU[self.rank][j])
broadcast(xs, i, j)
for tensor in xs:
self.assertEqual(tensor, expected_tensor)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_sparse_allreduce_ops(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
indices = torch.tensor([[0, 1]])
values = torch.tensor([[1, 2, 0], [4, 0, 6]])
sparse_tensor = torch.sparse_coo_tensor(indices, values, size=(2, 3)).to(self.rank)
# sparse allreduce call is wrapped in a try catch since the c10d API is only available in the nccl experimental branch
try:
work = pg.allreduce([sparse_tensor])
work.wait()
# work.result() returns a list of size 1, with the allreduce output as a dense tensor
a = torch.tensor([[2, 4, 0], [8, 0, 12]]).to(self.rank)
self.assertEqual(work.result()[0], a)
except RuntimeError as e:
if "allreduce_sparse is only available in the NCCL experimental branch." in str(e):
pass
else:
# Rethrow the exception if it's a different error
raise
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_allreduce_ops(self):
store = c10d.FileStore(self.file_name, self.world_size)
device_count = torch.cuda.device_count()
pg = self._create_process_group_nccl(store, self.opts())
local_device_id = self.rank_to_GPU[self.rank][0]
def allreduce(tensors, op):
opts = c10d.AllreduceOptions()
opts.reduceOp = op
work = pg.allreduce(tensors, opts)
work.wait()
# Sum
tensors = [torch.tensor([self.rank + 1]).cuda(local_device_id)]
allreduce(tensors, c10d.ReduceOp.SUM)
ndev = self.world_size
self.assertEqual(
torch.tensor([ndev * (ndev + 1) // 2]),
tensors[0],
)
# Avg (only available for NCCL 2.10+)
if torch.cuda.nccl.version() >= (2, 10, 0):
tensors = [torch.tensor([self.rank + 1.]).cuda(local_device_id)]
allreduce(tensors, c10d.ReduceOp.AVG)
ndev = self.world_size
self.assertEqual(
torch.tensor([ndev * (ndev + 1.) / (2. * ndev)]),
tensors[0],
)
# Premul Sum
if torch.cuda.nccl.version() >= (2, 11, 1):
for dtype in torch.half, torch.float, torch.double:
for factor in (3.0, torch.tensor([5.0], device=local_device_id, dtype=dtype)):
tensors = [torch.tensor([self.rank + 1]).cuda(local_device_id).to(dtype=dtype)]
allreduce(tensors, c10d._make_nccl_premul_sum(factor))
self.assertEqual(
factor * torch.tensor([self.world_size * (self.world_size + 1) / 2],
dtype=dtype, device=local_device_id),
tensors[0],
)
# Product
tensors = [torch.tensor([self.rank + 1]).cuda(local_device_id)]
allreduce(tensors, c10d.ReduceOp.PRODUCT)
self.assertEqual(
torch.tensor([math.factorial(self.world_size)]), tensors[0]
)
# Min
tensors = [torch.tensor([self.rank + 1]).cuda(local_device_id)]
allreduce(tensors, c10d.ReduceOp.MIN)
self.assertEqual(torch.tensor([1]), tensors[0])
# Max
tensors = [torch.tensor([self.rank + 1]).cuda(local_device_id)]
allreduce(tensors, c10d.ReduceOp.MAX)
self.assertEqual(torch.tensor([self.world_size]), tensors[0])
for op, err in zip((c10d.ReduceOp.BAND, c10d.ReduceOp.BOR, c10d.ReduceOp.BXOR),
("ReduceOp.BAND", "ReduceOp.BOR", "ReduceOp.BXOR")):
with self.assertRaisesRegex(
RuntimeError, "Cannot use " + err + " with NCCL"
):
allreduce(tensors, op)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_alltoall_ops_with_cudafree_race(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
opts = c10d.AllToAllOptions()
local_device = f"cuda:{self.rank_to_GPU[self.rank][0]}"
torch.cuda.set_device(local_device)
input = torch.rand(1000, 1000, device=local_device)
output = torch.rand(1000, 1000, device=local_device)
race_tensors = []
# create some tensors to race with alltoall collective
for _ in range(10):
tmp = []
for i in range(5):
tmp.append(torch.rand(10 ** (3 + i), device=local_device))
race_tensors.append(tmp)
for i in range(10):
race_tensors.pop()
work = pg.alltoall_base(output, input, [], [], opts)
# this triggers cudaFree
torch.cuda.empty_cache()
work.wait()
torch.cuda.synchronize(local_device)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_allreduce_in_cudagraph(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
local_device_idx = self.rank_to_GPU[self.rank][0]
with torch.cuda.device(local_device_idx):
xs = [torch.FloatTensor([1]).cuda(local_device_idx)]
# single warmup
pg.allreduce(xs).wait()
self.assertEqual(xs[0].item(), 2)
graph = torch.cuda.CUDAGraph()
with torch.cuda.graph(graph):
pg.allreduce(xs).wait()
self.assertEqual(xs[0].item(), 2)
graph.replay()
graph.replay()
self.assertEqual(xs[0].item(), 8)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_reduce_ops(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
local_device_id = self.rank_to_GPU[self.rank][0]
def reduce(xs, rootRank, rootTensor, op=None):
opts = c10d.ReduceOptions()
opts.rootRank = rootRank
opts.rootTensor = rootTensor
if op:
opts.reduceOp = op
work = pg.reduce(xs, opts)
work.wait()
# for every root tensor
for rt in range(self.world_size):
tensors = [torch.tensor([self.rank + 1]).cuda(local_device_id)]
reduce(tensors, rt, 0)
if self.rank == rt:
self.assertEqual(
torch.tensor([self.world_size * (self.world_size + 1) // 2]),
tensors[0],
)
else:
self.assertEqual(
torch.tensor([self.rank + 1]),
tensors[0],
)
for op, err in zip(
(c10d.ReduceOp.BAND, c10d.ReduceOp.BOR, c10d.ReduceOp.BXOR),
("ReduceOp.BAND", "ReduceOp.BOR", "ReduceOp.BXOR"),
):
with self.assertRaisesRegex(
RuntimeError, "Cannot use " + err + " with NCCL"
):
reduce(tensors, self.rank, rt, op)
# Premul sum
if torch.cuda.nccl.version() >= (2, 11, 1):
for factor in (3.0, torch.tensor([5.0], device=local_device_id)):
if isinstance(factor, torch.Tensor):
factor_ref = factor.cpu().item()
else:
factor_ref = factor
float_tensors = [
torch.tensor(
[self.rank + 1.0], device=f"cuda:{local_device_id}")
]
float_tensors_ref = [
torch.tensor(
[(self.rank + 1.0) * factor_ref], device=f"cuda:{local_device_id}")
]
reduce(float_tensors_ref, rt, 0)
reduce(float_tensors, rt, 0, c10d._make_nccl_premul_sum(factor))
if self.rank == rt:
self.assertEqual(float_tensors_ref[0], float_tensors[0])
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_allgather_ops(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
local_device_ids = self.rank_to_GPU[self.rank]
def allgather(output_ts, input_ts):
work = pg.allgather(output_ts, input_ts)
return work.wait()
tensors = [torch.empty(2, 2).fill_(2).cuda(device=i) for i in local_device_ids]
output_tensors = []
expected_output = []
output_per_gpu = ([torch.empty(2, 2).fill_(-1)] * len(local_device_ids) * self.world_size)
expected_per_gpu = ([torch.empty(2, 2).fill_(2)] * len(local_device_ids) * self.world_size)
for gpu in local_device_ids:
output_tensors.append([t.cuda(device=gpu) for t in output_per_gpu])
expected_output.append([t.cuda(device=gpu) for t in expected_per_gpu])
result = allgather(output_tensors, tensors)
# Verification
self.assertEqual(output_tensors, expected_output)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_allgather_base_ops(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
local_device_id = self.rank_to_GPU[self.rank][0]
def allgather_base(output_t, input_t):
work = pg._allgather_base(output_t, input_t)
work.wait()
# allgather_base is GPU number agnostic.
# Each rank contribute one tensor regardless of GPU counts
tensor = torch.tensor([self.rank]).cuda(local_device_id)
output_t = torch.empty((self.world_size), dtype=tensor.dtype).cuda(local_device_id)
allgather_base(output_t, tensor)
# Verification
self.assertEqual(torch.arange(self.world_size), output_t)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_allgather_base_basics(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
local_device_id = self.rank_to_GPU[self.rank][0]
def allgather_base(output_t, input_t):
work = pg._allgather_base(output_t, input_t)
work.wait()
# anticipate an error
with self.assertRaisesRegex(
RuntimeError,
"output tensor size must be equal to world_size times input tensor size",
):
tensor = torch.tensor([self.rank]).cuda(local_device_id)
output_t = torch.empty((self.world_size + 1), dtype=tensor.dtype).cuda(
local_device_id
)
# fails the check because output_t is not correctly sized
allgather_base(output_t, tensor)
# anticipate an error
with self.assertRaisesRegex(
RuntimeError, "output tensor must have the same type as input tensor"
):
tensor = torch.tensor([self.rank], dtype=torch.float).cuda(local_device_id)
output_t = torch.empty((self.world_size + 1), dtype=torch.long).cuda(
local_device_id
)
# fails the check because the dtype is different
allgather_base(output_t, tensor)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_gather_ops(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
local_device_ids = self.rank_to_GPU[self.rank]
num_gpus = len(local_device_ids)
def gather(output_t, input_t, rootRank):
opts = c10d.GatherOptions()
opts.rootRank = rootRank
if rootRank == self.rank:
work = pg.gather(output_t, input_t, opts)
else:
work = pg.gather([], input_t, opts)
work.wait()
# init input
tensors = []
for device_id in local_device_ids:
tensors.append(torch.tensor([self.rank]).cuda(device_id))
# init output
output_ts = []
for idx in range(num_gpus):
gpu_idx = local_device_ids[idx]
output_ts.append([])
for rank in range(self.world_size):
output_ts[idx].append(torch.tensor([-1]).cuda(gpu_idx))
expected = [[torch.tensor([rank]) for rank in range(self.world_size)]]
for rank in range(self.world_size):
gather(output_ts, tensors, rank)
if rank == self.rank:
self.assertEqual(expected, output_ts)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_gather_stress(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
local_device_ids = self.rank_to_GPU[self.rank]
num_gpus = len(local_device_ids)
def gather(output_t, input_t, rootRank):
opts = c10d.GatherOptions()
opts.rootRank = rootRank
if rootRank == self.rank:
work = pg.gather(output_t, input_t, opts)
else:
work = pg.gather([], input_t, opts)
work.wait()
stress_length = 1000
# init input
tensors = []
for i in range(stress_length):
tensors.append([])
for device_id in local_device_ids:
tensors[i].append(torch.tensor([self.rank]).cuda(device_id))
# init output
output_ts = []
for i in range(stress_length):
output_ts.append([[] for _ in range(num_gpus)])
for idx, ls in enumerate(output_ts[i]):
gpu_idx = local_device_ids[idx]
for _ in range(self.world_size):
ls.append(torch.tensor([-1]).cuda(gpu_idx))
expected = [[torch.tensor([rank]) for rank in range(self.world_size)]]
for i in range(stress_length):
for rank in range(self.world_size):
gather(output_ts[i], tensors[i], rank)
# Verification
if rank == self.rank:
self.assertEqual(output_ts[i], expected)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_gather_checks(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
local_device_ids = self.rank_to_GPU[self.rank]
num_gpus = len(local_device_ids)
# init input
tensors = []
for device_id in local_device_ids:
tensors.append(torch.tensor([self.rank]).cuda(device_id))
# init output
output_ts = []
for idx in range(num_gpus):
gpu_idx = local_device_ids[idx]
output_ts.append([])
for rank in range(self.world_size):
output_ts[idx].append(torch.tensor([-1]).cuda(gpu_idx))
with self.assertRaisesRegex(RuntimeError, "invalid root rank"):
opts = c10d.GatherOptions()
opts.rootRank = -1
pg.gather(output_ts, tensors, opts)
with self.assertRaisesRegex(TypeError, "incompatible function arguments"):
pg.gather(output_ts, tensors, 0)
with self.assertRaisesRegex(RuntimeError, "invalid root rank"):
opts = c10d.GatherOptions()
opts.rootRank = self.world_size
pg.gather(output_ts, tensors, opts)
with self.assertRaisesRegex(
# throws error message from dispatcher
RuntimeError, "There were no tensor arguments to this function"
):
opts = c10d.GatherOptions()
opts.rootRank = 0
pg.gather(output_ts, [], opts)
with self.assertRaisesRegex(
RuntimeError, "Tensors must be on distinct GPU devices"
):
# init input
tensors2 = []
for device_id in local_device_ids:
tensors2.append(torch.tensor([self.rank]).cuda(device_id))
tensors2.append(torch.tensor([self.rank]).cuda(device_id))
opts = c10d.GatherOptions()
opts.rootRank = 0
pg.gather(output_ts, tensors2, opts)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_scatter_ops(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
local_device_ids = self.rank_to_GPU[self.rank]
num_gpus = len(local_device_ids)
def scatter(output_t, input_t, rootRank):
opts = c10d.ScatterOptions()
opts.rootRank = rootRank
if rootRank == self.rank:
work = pg.scatter(output_t, input_t, opts)
else:
work = pg.scatter(output_t, [], opts)
work.wait()
# init output
tensors = []
for device_id in local_device_ids:
tensors.append(torch.tensor([-1]).cuda(device_id))
# init input
scatter_list = []
for idx in range(num_gpus):
gpu_idx = local_device_ids[idx]
scatter_list.append([])
for rank in range(self.world_size):
scatter_list[idx].append(torch.tensor([rank]).cuda(gpu_idx))
# test each rank to scatter
expected = [torch.tensor([self.rank])]
for rank in range(self.world_size):
scatter(tensors, scatter_list, rank)
self.assertEqual(expected, tensors)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_scatter_stress(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
local_device_ids = self.rank_to_GPU[self.rank]
num_gpus = len(local_device_ids)
def scatter(output_t, input_t, rootRank):
opts = c10d.ScatterOptions()
opts.rootRank = rootRank
if rootRank == self.rank:
work = pg.scatter(output_t, input_t, opts)
else:
work = pg.scatter(output_t, [], opts)
work.wait()
stress_length = 1000
# init output
tensors = []
for i in range(stress_length):
tensors.append([])
for device_id in local_device_ids:
tensors[i].append(torch.tensor([-1]).cuda(device_id))
# init input
scatter_list = []
for i in range(stress_length):
scatter_list.append([[] for _ in range(num_gpus)])
for idx, ls in enumerate(scatter_list[i]):
gpu_idx = local_device_ids[idx]
for rank in range(self.world_size):
ls.append(torch.tensor([rank]).cuda(gpu_idx))
# test each rank to scatter
expected = [torch.tensor([self.rank])]
for i in range(stress_length):
for rank in range(self.world_size):
scatter(tensors[i], scatter_list[i], rank)
# Verification
self.assertEqual(tensors[i], expected)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_scatter_checks(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
local_device_ids = self.rank_to_GPU[self.rank]
num_gpus = len(local_device_ids)
# init output
tensors = []
for device_id in local_device_ids:
tensors.append(torch.tensor([-1]).cuda(device_id))
# init input
scatter_list = []
for idx in range(num_gpus):
gpu_idx = local_device_ids[idx]
scatter_list.append([])
for rank in range(self.world_size):
scatter_list[idx].append(torch.tensor([rank]).cuda(gpu_idx))
with self.assertRaisesRegex(RuntimeError, "invalid root rank"):
opts = c10d.ScatterOptions()
opts.rootRank = -1
pg.scatter(tensors, scatter_list, opts)
with self.assertRaisesRegex(TypeError, "incompatible function arguments"):
pg.scatter(tensors, scatter_list, 0)
with self.assertRaisesRegex(RuntimeError, "invalid root rank"):
opts = c10d.ScatterOptions()
opts.rootRank = self.world_size
pg.scatter(tensors, scatter_list, opts)
with self.assertRaisesRegex(
# throws error message from dispatcher
RuntimeError, "There were no tensor arguments to this function"
):
opts = c10d.ScatterOptions()
opts.rootRank = 0
pg.scatter([], scatter_list, opts)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_reduce_scatter_base_basics(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
local_device_id = self.rank_to_GPU[self.rank][0]
def reduce_scatter_base(output_t, input_t):
work = pg._reduce_scatter_base(output_t, input_t)
work.wait()
# anticipate an error
with self.assertRaisesRegex(
RuntimeError,
"input tensor must be the same size as output size times world size",
):
input_t = torch.tensor([self.rank]).cuda(local_device_id)
output_t = torch.empty((self.world_size + 1), dtype=input_t.dtype).cuda(
local_device_id
)
# fails the check because output_t is not correctly sized
reduce_scatter_base(output_t, input_t)
# anticipate an error
with self.assertRaisesRegex(
RuntimeError, "input tensor must be the same type as the output tensor."
):
tensor = torch.tensor([self.rank], dtype=torch.float).cuda(local_device_id)
output_t = torch.empty((self.world_size + 1), dtype=torch.long).cuda(
local_device_id
)
# fails the check because the dtype is different
reduce_scatter_base(output_t, tensor)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_reduce_scatter_ops(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
local_device_ids = self.rank_to_GPU[self.rank]
num_gpus = len(local_device_ids)
def reduce_scatter(outputs, input_lists, op):
opts = c10d.ReduceScatterOptions()
opts.reduceOp = op
work = pg.reduce_scatter(outputs, input_lists, opts)
work.wait()
output = [torch.tensor([0]).cuda(i) for i in local_device_ids]
# GPU/rank
# 0 [1], [2], [3], [4]
# 1 [2], [3], [4], [5]
# 2 [3], [4], [5], [6]
# 3 [4], [5], [6], [7]
# Sum
tensor_lists = []
input_per_gpu = []
for i in range(self.world_size):
input_per_gpu.append(torch.tensor([self.rank + i + 1]))
for gpu in local_device_ids:
tensor_lists.append([t.cuda(device=gpu) for t in input_per_gpu])
reduce_scatter(output, tensor_lists, c10d.ReduceOp.SUM)
for i in range(num_gpus):
expected = torch.tensor(
[
(1 + self.world_size) * self.world_size // 2
+ self.world_size * self.rank
])
self.assertEqual(expected, output[i])
# Min
reduce_scatter(output, tensor_lists, c10d.ReduceOp.MIN)
for i in range(num_gpus):
expected = torch.tensor([self.rank + 1 + i])
self.assertEqual(expected, output[i])
# Max
reduce_scatter(output, tensor_lists, c10d.ReduceOp.MAX)
for i in range(num_gpus):
expected = torch.tensor(
[self.rank + self.world_size + i]
)
self.assertEqual(expected, output[i])
# Product
reduce_scatter(output, tensor_lists, c10d.ReduceOp.PRODUCT)
# math package don't have math.perm until python 3.8, so
# we implement a naive version here.
def perm(n, k):
prod_val = n
for val in range(n - k + 1, n):
prod_val *= val
return prod_val
for i in range(num_gpus):
prod_val = perm(self.rank + self.world_size, self.world_size)
expected = torch.tensor([prod_val])
self.assertEqual(expected, output[i])
# Test the input params overridden scenarios, aka, when the input is
# a list and output is just one tensor.
# Sum
output_tensor = torch.empty_like(input_per_gpu[0][0]).cuda(self.rank)
input_list = [tensor[0].cuda(self.rank) for tensor in input_per_gpu]
pg.reduce_scatter(output_tensor, input_list, c10d.ReduceOp.SUM).wait()
expected = torch.tensor(
(1 + self.world_size) * self.world_size // 2 + self.world_size * self.rank
)
self.assertEqual(expected, output_tensor)
# Min
pg.reduce_scatter(output_tensor, input_list, c10d.ReduceOp.MIN).wait()
expected = torch.tensor(self.rank + 1)
self.assertEqual(expected, output_tensor)
# Max
pg.reduce_scatter(output_tensor, input_list, c10d.ReduceOp.MAX).wait()
expected = torch.tensor(self.rank + self.world_size)
self.assertEqual(expected, output_tensor)
# Product
pg.reduce_scatter(output_tensor, input_list, c10d.ReduceOp.PRODUCT).wait()
prod_val = self.rank + 1
for k in range(1, self.world_size):
prod_val = prod_val * (self.rank + 1 + k)
expected = torch.tensor(prod_val)
self.assertEqual(expected, output_tensor)
if torch.cuda.nccl.version() >= (2, 11, 1):
for factor in (3.0, torch.tensor([5.0], device=self.rank)):
if isinstance(factor, torch.Tensor):
factor_ref = factor.cpu().item()
else:
factor_ref = factor
output = [t.float() for t in output]
tensor_lists = [[t.float() for t in tl] for tl in tensor_lists]
output_ref = [t.float() for t in output]
tensor_lists_ref = [[t.float() * factor_ref for t in tl] for tl in tensor_lists]
reduce_scatter(output, tensor_lists, c10d._make_nccl_premul_sum(factor))
reduce_scatter(output_ref, tensor_lists_ref, c10d.ReduceOp.SUM)
self.assertEqual(output_ref, output)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_reduce_scatter_base_ops(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
local_device_id = self.rank_to_GPU[self.rank][0]
def reduce_scatter_base(output_t, input_t):
work = pg._reduce_scatter_base(output_t, input_t)
work.wait()
# reduce_scatter_base is GPU number agnostic.
# Each rank contribute one tensor regardless of GPU counts
output_t = torch.empty([1]).cuda(local_device_id)
tensor = torch.arange(self.world_size, dtype=output_t.dtype).cuda(local_device_id)
reduce_scatter_base(output_t, tensor)
# Verification
self.assertEqual(output_t[0], self.rank * self.world_size)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_barrier(self):
store = c10d.FileStore(self.file_name, self.world_size)
pg = self._create_process_group_nccl(store, self.opts())
local_device_ids = self.rank_to_GPU[self.rank]
def allreduce(tensors):
opts = c10d.AllreduceOptions()
work = pg.allreduce(tensors, opts)
return work
# Making the collective to operate on
# 1, 2, 3, 4, .... len(local_device_ids) GPUs
tensors_list = [[] for _ in range(len(local_device_ids))]
for i in range(1, len(local_device_ids) + 1):
for j in range(i):
tensors_list[i - 1].append(torch.tensor([j + 1]).cuda(local_device_ids[j]))
works = []
for tensors in tensors_list:
work = allreduce(tensors)
works.append(work)
# Barrier will ensure that all previous work is completed
pg.barrier().wait()
for i in range(1, len(local_device_ids) + 1):
for j in range(i):
self.assertEqual(
torch.tensor([(j + 1) * self.world_size]), tensors_list[i - 1][j]
)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_send_recv(self):
store = c10d.FileStore(self.file_name, self.world_size)
self._create_process_group_nccl(store, self.opts())
device = self.rank_to_GPU[self.rank][0]
# Generate the same random tensor
torch.manual_seed(0)
send_tensor = torch.rand(10, 10, device=device)
if self.rank == 0:
dist.send(send_tensor, 1)
if self.rank == 1:
recv_tensor = torch.rand(10, 10, device=device)
dist.recv(recv_tensor, 0)
self.assertEqual(send_tensor, recv_tensor)
# Test with non-contiguous tensors.
send_tensor_view = send_tensor.t()
if self.rank == 0:
with self.assertRaisesRegex(RuntimeError, 'Tensors must be contiguous'):
dist.send(send_tensor_view, 1)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 1, "NCCL test requires 1 GPU")
@skip_if_lt_x_gpu(1)
def test_nccl_dist_backend_error(self):
store = c10d.FileStore(self.file_name, self.world_size)
self._create_process_group_nccl(store, self.opts())
# Both rank 0 and 1 will use the same CUDA device resulting in ncclInvalidUsage
with self.assertRaises(dist.DistBackendError) as cm:
dist.broadcast(torch.tensor([1, 2, 3]).cuda(), 0)
self.assertTrue(isinstance(cm.exception, dist.DistError))
self.assertIsInstance(cm.exception, RuntimeError)
@requires_nccl()
@skip_but_pass_in_sandcastle_if(torch.cuda.device_count() < 2, "NCCL test requires 2+ GPUs")
def test_abort_pg(self):
# Disable ASYNC_ERROR_HANDLING for this test to ensure we can programmatically
# abort the process group.
os.environ["NCCL_ASYNC_ERROR_HANDLING"] = "0"
store = c10d.FileStore(self.file_name, self.world_size)
self._create_process_group_nccl(store, self.opts())
device = self.rank_to_GPU[self.rank][0]
t = torch.rand(10, 10, device=device)
# First allreduce to initialize state.
dist.all_reduce(t)
def abortpg():
c10d.distributed_c10d._get_default_group()._get_backend(torch.device(device))._abort()
# Initialize DDP to ensure "destroy_process_group" will not call
# ProcessGroupNCCL destructor since DDP holds a reference to process group.
# Run a single iteration of DDP to initialize state.
model = DistributedDataParallel(
torch.nn.Linear(10, 10).to(device), device_ids=[device]
)
model(t).sum().backward()
# Now simulate collective getting stuck and abort gets us unstuck
if self.rank == 0:
dist.all_reduce(t)
# Schedule thread before we get stuck to abort pg.
thread = threading.Thread(target=abortpg)
thread.start()
# We would get stuck here due to d2h if we didn't abort.
t_cpu = t.cpu()
thread.join()
class DistributedDataParallelTest(
test_c10d_common.CommonDistributedDataParallelTest, MultiProcessTestCase
):
def setUp(self):
super().setUp()
# NCCL_BLOCKING_WAIT overrides NCCL_ASYNC_ERROR_HANDLING hence tests
# that use NCCL_BLOCKING_WAIT will test it as expected.
os.environ["NCCL_ASYNC_ERROR_HANDLING"] = "1"
self._spawn_processes()
def _get_process_group(self):
store = self._get_store()
c10d.init_process_group("nccl", store=store, rank=self.rank, world_size=self.world_size)
return c10d.distributed_c10d._get_default_group()
def _test_nccl_backend(
self, devices, device_ids, multi_device=False, gradient_as_bucket_view=False
):
process_group = self._get_process_group()
self._test_ddp_with_process_group(
process_group, devices, device_ids, multi_device, gradient_as_bucket_view
)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_nccl_propagate_error_reason(self):
# Need to use NCCL_BLOCKING_WAIT and not ASYNC_ERROR_HANDLING,
# otherwise process will be taken down and we can't check for errors.
os.environ["NCCL_ASYNC_ERROR_HANDLING"] = "0"
os.environ["NCCL_BLOCKING_WAIT"] = "1"
# TODO: smaller timeout can fail since PG NCCl does health check in
# constructor. Look into reducing this test's runtime.
store = c10d.FileStore(self.file_name, self.world_size)
# provide sufficient timeout to initialize NCCL comm.
pg = c10d.ProcessGroupNCCL(store, self.rank, self.world_size, timeout=timedelta(seconds=15))
pg_gloo = c10d.ProcessGroupGloo(store, self.rank, self.world_size)
pg.barrier().wait(timedelta(seconds=5))
# Simulate stuckness in rank 0.
if self.rank == 0:
pg_gloo.barrier().wait()
inp = torch.ones(1).cuda(self.rank)
if self.rank != 0:
# Time out due to rank 0 not calling into allreduce.
with self.assertRaises(RuntimeError):
pg.allreduce([inp]).wait(timedelta(seconds=5))
# Now when nonzero rank attempts to use communicator, original failure reason should be logged.j
try:
pg.allreduce([torch.ones(2).cuda(self.rank)]).wait()
except RuntimeError as e:
self.assertTrue("aborted" in str(e))
else:
self.fail("Expected error to be raised!")
# Unblock rank 0
pg_gloo.barrier().wait()
# TODO: We can also test that if rank 0 attempts to use the communicator,
# then we should error out with the info that it was aborted due to
# timeout on another rank. Although this would only be the case after
# the watchdog has run on the rank, and there is no reliable way
# to confirm it has run.
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_nccl_backend_multi_device_ids_not_allowed(self):
int_devices = list(range(torch.cuda.device_count()))
devices = [torch.device("cuda:" + str(i)) for i in int_devices]
with self.assertRaisesRegex(
ValueError, "device_ids can only be None or contain a single element."
):
self._test_nccl_backend(devices, int_devices)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_nccl_backend_single_device_module_device_ids_None(self):
self._test_nccl_backend(None, None)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_nccl_backend_single_device_module_empty_device_ids(self):
# This tests the backward compatibility of accepting an empty list as `device_ids`,
# although we no longer document this in favor of the default value of `None`,
# which is consistent with multi-device modules and CPU modules.
self._test_nccl_backend(None, [])
@requires_nccl()
@skip_if_lt_x_gpu(4)
def test_nccl_backend_multi_device_module_device_ids_None(self):
int_devices = gpus_for_rank(self.world_size)[self.rank][:2]
devices = [torch.device("cuda:" + str(i)) for i in int_devices]
self._test_nccl_backend(devices, None, multi_device=True)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_nccl_backend_1gpu_module_device_ids_integer_list(self):
int_devices = gpus_for_rank(self.world_size)[self.rank][:1]
devices = [torch.device("cuda:" + str(i)) for i in int_devices]
self._test_nccl_backend(devices, int_devices)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_nccl_backend_1gpu_module_device_ids_torch_device_list(self):
int_devices = gpus_for_rank(self.world_size)[self.rank][:1]
devices = [torch.device("cuda:" + str(i)) for i in int_devices]
self._test_nccl_backend(devices, devices)
@requires_nccl()
@skip_if_lt_x_gpu(4)
def test_nccl_backend_2gpu_module(self):
int_devices = gpus_for_rank(self.world_size)[self.rank][:2]
devices = [torch.device("cuda:" + str(i)) for i in int_devices]
self._test_nccl_backend(devices, None, multi_device=True)
@requires_nccl()
@skip_if_lt_x_gpu(8)
def test_nccl_backend_4gpu_module(self):
int_devices = gpus_for_rank(self.world_size)[self.rank][:4]
devices = [torch.device("cuda:" + str(i)) for i in int_devices]
self._test_nccl_backend(devices, None, multi_device=True)
@requires_nccl()
@skip_if_lt_x_gpu(4)
def test_ddp_multi_device_module_config(self):
gpus = gpus_for_rank(self.world_size)[self.rank]
self.assertTrue(len(gpus) >= 2, "expecting at least 2 gpus per process")
process_group = self._get_process_group()
gpus = gpus[:2]
model = DoubleGpuNet(gpus)
with self.assertRaisesRegex(
ValueError,
"DistributedDataParallel device_ids and output_device arguments only work with "
"single-device/multiple-device GPU modules or CPU modules",
):
ddp_model = DistributedDataParallel(
model, output_device=gpus[1], process_group=process_group
)
with self.assertRaisesRegex(
ValueError, "device_ids can only be None or contain a single element."
):
ddp_model = DistributedDataParallel(
model, device_ids=gpus, process_group=process_group
)
with self.assertRaisesRegex(
ValueError, "input module must be on the same type of devices"
):
model.fc1 = model.fc1.cpu()
ddp_model = DistributedDataParallel(model, process_group=process_group)
model = model.cpu()
with self.assertRaisesRegex(
ValueError, "device_ids can only be None or contain a single element."
):
ddp_model = DistributedDataParallel(
model, device_ids=gpus, process_group=process_group
)
def _test_fp16(self, gradient_as_bucket_view=False):
process_group = self._get_process_group()
gpus = gpus_for_rank(self.world_size)[self.rank]
model = nn.Linear(1, 1, bias=False).cuda(gpus[0]).half()
nn.init.constant_(model.weight, 1)
ddp_model = DistributedDataParallel(
model,
device_ids=[gpus[0]],
process_group=process_group,
bucket_cap_mb=0.001,
gradient_as_bucket_view=gradient_as_bucket_view,
)
# Input 2**15, so that the gradients will overflow with a
# world_size of 2, unless we normalize the gradient by the
# world_size before the reduction
input = torch.tensor([[2 ** 15]]).cuda(gpus[0]).half()
# Step model
ddp_model.train()
output = ddp_model(input)
loss = output.sum()
loss.backward()
self.assertFalse(any(torch.isinf(p.grad).any() for p in ddp_model.parameters()))
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_fp16(self):
self._test_fp16()
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_fp16_grad_is_view(self):
self._test_fp16(gradient_as_bucket_view=True)
def _test_arbitrary_forward_return_value(self, gradient_as_bucket_view=False):
"""
Note: this test can be sped up by only running it on a CPU module
once DistributedDataParallel supports them.
"""
process_group = self._get_process_group()
class ForwardReturnValueModule(nn.Module):
def __init__(self):
super().__init__()
self.fc1 = nn.Linear(2, 10, bias=False)
self.fc2 = nn.Linear(10, 4, bias=False)
self.fc3 = nn.Linear(4, 4, bias=False)
self.relu = nn.ReLU()
def forward(self, x, fn):
x = self.relu(self.fc1(x))
x = self.relu(self.fc2(x))
# The first softmax does NOT include fc3 in its autograd graph
# whereas the second softmax DOES. If we pass only the first
# tensor we see in the output to the reducer, it marks the
# gradient for fc3 as ready (because it doesn't show up). If
# downstream uses of this return value choose to differentiate
# against the second output tensor, it would still receive a
# gradient and a callback for this tensor, resulting in a crash.
return fn(
F.softmax(x, dim=1),
F.softmax(self.fc3(x), dim=1),
)
device_id = gpus_for_rank(self.world_size)[self.rank][0]
model = DistributedDataParallel(
ForwardReturnValueModule().float().to(device_id),
device_ids=[device_id],
process_group=process_group,
gradient_as_bucket_view=gradient_as_bucket_view,
)
batch_size = 4
criterion = nn.CrossEntropyLoss()
input = torch.rand([batch_size, 2], dtype=torch.float)
target = torch.LongTensor([random.randrange(4) for _ in range(batch_size)]).to(
device_id
)
# Always run "backward" to ensure the reducer is called by autograd.
# If we don't correctly capture the output tensors from the return value,
# the reducer won't see a hook for the unused parameter, and throw an error.
# The correct capture is what we're testing in this function.
def test(box, unbox):
output = model(input, fn=box)
loss = criterion(unbox(output), target)
loss.backward()
# Test with identity return value
test(
box=lambda x, y: (x, y),
unbox=lambda obj: obj[1],
)
# Test with list return value
test(
box=lambda x, y: ["foo", x, "bar", y],
unbox=lambda obj: obj[3],
)
# Test with tuple return value
test(
box=lambda x, y: ("foo", x, "bar", y),
unbox=lambda obj: obj[3],
)
# Test with dict return value
test(
box=lambda x, y: {"foo": "bar", "a": x, "b": y},
unbox=lambda obj: obj["b"],
)
# Test with list with dict return value
test(
box=lambda x, y: ["foo", "bar", {"a": x, "b": y}],
unbox=lambda obj: obj[2]["b"],
)
# Test with dict with list return value
test(
box=lambda x, y: {"foo": "bar", "list": [0, x, 1, y]},
unbox=lambda obj: obj["list"][3],
)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_arbitrary_forward_return_value(self):
self._test_arbitrary_forward_return_value()
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_arbitrary_forward_return_value_grad_is_view(self):
self._test_arbitrary_forward_return_value(gradient_as_bucket_view=True)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_ddp_with_lazy_parameters(self):
process_group = self._get_process_group()
with self.assertRaisesRegex(
RuntimeError, "Modules with uninitialized parameters"
):
DistributedDataParallel(
torch.nn.LazyLinear(10), process_group=process_group
)
def _test_find_unused_parameters_kwarg(self, gradient_as_bucket_view=False):
"""
Note: this test can be sped up by only running it on a CPU module
once DistributedDataParallel supports them.
"""
torch.cuda.set_device(self.rank)
dist.init_process_group(
backend="nccl",
world_size=self.world_size,
rank=self.rank,
init_method=f"file://{self.file_name}",
)
process_group = c10d.distributed_c10d._get_default_group()
class FindUnusedParametersModule(nn.Module):
def __init__(self):
super().__init__()
self.fc1 = nn.Linear(2, 10, bias=False)
self.fc2 = nn.Linear(10, 4, bias=False)
self.fc3 = nn.Linear(4, 4, bias=False)
self.relu = nn.ReLU()
def forward(self, x):
x = self.relu(self.fc1(x))
x = self.relu(self.fc2(x))
# Return the fc3 module so that the caller can invoke it
# outside of the forward function. While this is bad practice,
# we can use it to trigger a reducer error.
return (F.softmax(x, dim=1), self.fc3)
device_id = gpus_for_rank(self.world_size)[self.rank][0]
batch_size = 4
criterion = nn.CrossEntropyLoss()
input = torch.rand([batch_size, 2], dtype=torch.float)
target = torch.LongTensor([random.randrange(4) for _ in range(batch_size)]).to(
device_id
)
ddp_model = None
def test_find_unused_parameters(
find_unused_parameters, test_default=False, gradient_as_bucket_view=False
):
if test_default:
model = DistributedDataParallel(
FindUnusedParametersModule().float().to(device_id),
device_ids=[device_id],
process_group=process_group,
gradient_as_bucket_view=gradient_as_bucket_view,
)
else:
model = DistributedDataParallel(
FindUnusedParametersModule().float().to(device_id),
device_ids=[device_id],
process_group=process_group,
find_unused_parameters=find_unused_parameters,
gradient_as_bucket_view=gradient_as_bucket_view,
)
nonlocal ddp_model
ddp_model = model
output, fc3 = model(input)
output = fc3(output)
loss = criterion(output, target)
loss.backward()
# First test that finding unused params under these conditions is to
# trigger an error when `backward` is called (because fc3 is an unused
# parameter and will therefore be marked ready twice).
try:
test_find_unused_parameters(
True, gradient_as_bucket_view=gradient_as_bucket_view
)
except Exception as ex:
self.assertTrue(
str(ex).startswith(
"Expected to mark a variable ready only once.",
)
)
unused_index = 2
unused_index_str = f"Parameter at index {unused_index}"
model = ddp_model.module
for module_name, module in model.named_modules():
if module == model.fc3:
for parameter_name, _ in module.named_parameters(recurse=False):
unused_fqn = f"{module_name}.{parameter_name}"
# Only one such parameter in model.fc3, since bias=False
break
if dist.get_debug_level() != dist.DebugLevel.OFF:
unused_index_str += f" with name {unused_fqn}"
self.assertTrue(unused_index_str in str(ex))
else:
self.fail("Expected exception")
dist.barrier(process_group)
# Then test that the default behavior can be overridden by setting
# `find_unused_parameters=False`.
try:
test_find_unused_parameters(
False, gradient_as_bucket_view=gradient_as_bucket_view
)
except Exception as ex:
self.fail("Unexpected exception: %s" % ex)
# Test find_unused_parameters defaults to False
try:
test_find_unused_parameters(
True, test_default=True, gradient_as_bucket_view=gradient_as_bucket_view
)
except Exception as ex:
self.fail("Unexpected exception: %s" % ex)
# TODO: Combine the following tests once https://github.com/pytorch/pytorch/issues/55967
# is resolved.
@requires_nccl()
@skip_if_lt_x_gpu(2)
@with_dist_debug_levels(levels=["DETAIL"])
def test_find_unused_parameters_kwarg_debug_detail(self):
self._test_find_unused_parameters_kwarg()
@requires_nccl()
@skip_if_lt_x_gpu(2)
@with_dist_debug_levels(levels=["INFO"])
def test_find_unused_parameters_kwarg_debug_info(self):
self._test_find_unused_parameters_kwarg()
@requires_nccl()
@skip_if_lt_x_gpu(2)
@with_dist_debug_levels(levels=["OFF"])
def test_find_unused_parameters_kwarg_debug_off(self):
self._test_find_unused_parameters_kwarg()
@requires_nccl()
@skip_if_lt_x_gpu(2)
@with_dist_debug_levels(levels=["DETAIL"])
def test_find_unused_parameters_kwarg_grad_is_view_debug_detail(self):
self._test_find_unused_parameters_kwarg(gradient_as_bucket_view=True)
@requires_nccl()
@skip_if_lt_x_gpu(2)
@with_dist_debug_levels(levels=["INFO"])
def test_find_unused_parameters_kwarg_grad_is_view_debug_info(self):
self._test_find_unused_parameters_kwarg(gradient_as_bucket_view=True)
@requires_nccl()
@skip_if_lt_x_gpu(2)
@with_dist_debug_levels(levels=["OFF"])
def test_find_unused_parameters_kwarg_grad_is_view_debug_off(self):
self._test_find_unused_parameters_kwarg(gradient_as_bucket_view=True)
def _test_multiple_outputs_multiple_backward(self, gradient_as_bucket_view=False):
"""
Note: this test can be sped up by only running it on a CPU module
once DistributedDataParallel supports them.
"""
process_group = self._get_process_group()
class MultipleOutputModule(nn.Module):
def __init__(self):
super().__init__()
def define_module():
return nn.Sequential(
nn.Linear(2, 10, bias=False),
nn.ReLU(),
nn.Linear(10, 4, bias=False),
nn.ReLU(),
)
self.module0 = define_module()
self.module1 = define_module()
def forward(self, x):
return (
F.softmax(self.module0(x), dim=1),
F.softmax(self.module1(x), dim=1),
)
device_id = gpus_for_rank(self.world_size)[self.rank][0]
model = DistributedDataParallel(
MultipleOutputModule().float().to(device_id),
device_ids=[device_id],
process_group=process_group,
gradient_as_bucket_view=gradient_as_bucket_view,
)
batch_size = 4
criterion = nn.CrossEntropyLoss()
input = torch.rand([batch_size, 2], dtype=torch.float)
target = torch.LongTensor([random.randrange(4) for _ in range(batch_size)]).to(
device_id
)
# Compute loss and gradients for both outputs
output1, output2 = model(input)
loss1 = criterion(output1, target)
loss1.backward()
loss2 = criterion(output2, target)
loss2.backward()
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_multiple_outputs_multiple_backward(self):
self._test_multiple_outputs_multiple_backward()
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_multiple_outputs_multiple_backward_grad_is_view(self):
self._test_multiple_outputs_multiple_backward(gradient_as_bucket_view=True)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_no_grad(self):
"""
Note: this test can be sped up by only running it on a CPU module
once DistributedDataParallel supports them.
"""
process_group = self._get_process_group()
class NoGradModule(nn.Module):
def __init__(self):
super().__init__()
self.fc1 = nn.Linear(2, 10, bias=False)
self.fc2 = nn.Linear(10, 4, bias=False)
self.relu = nn.ReLU()
def forward(self, x):
x = self.relu(self.fc1(x))
x = self.relu(self.fc2(x))
return F.softmax(x, dim=1)
device_id = gpus_for_rank(self.world_size)[self.rank][0]
model = DistributedDataParallel(
NoGradModule().float().to(device_id),
device_ids=[device_id],
process_group=process_group,
)
batch_size = 4
input = torch.rand([batch_size, 2], dtype=torch.float)
def check_no_grads():
for p in model.parameters():
self.assertTrue(p.requires_grad)
self.assertIsNone(p.grad)
# After initialization, no parameter has their gradient set.
check_no_grads()
# Run `forward` function with torch.no_grad()
with torch.no_grad():
output = model(input)
self.assertTrue(isinstance(output, torch.Tensor))
# No parameter should have their gradient set.
check_no_grads()
def _test_accumulate_gradients_module(self, gradient_as_bucket_view=False):
# This is NOT the recommended way to implement accumulating grads, but
# we would like to make sure DDP does not mess up with the underlying
# module.
int_devices = gpus_for_rank(self.world_size)[self.rank][:1]
devices = [torch.device("cuda:" + str(i)) for i in int_devices]
process_group = self._get_process_group()
global_batch_size = self.world_size
model, ddp_model, input, target = self._prepare_single_device_module(
process_group, devices, devices, global_batch_size, gradient_as_bucket_view
)
def step_model(model, input, target):
model.train()
output = model(input)
loss = F.mse_loss(output, target.to(output.device))
loss.backward()
# ensure accumulate grads works with no_grad
with torch.no_grad():
ddp_model.train()
ddp_model.module(input)
# Check two model parameters over 4 iterations.
# Use 4 iterations because we alternate between reducing and
# not reducing and want to make sure we switch both ways.
for iteration in range(4):
step_model(model, input, target)
if iteration % 2 == 0:
# Skip gradients sync without calling prepare_for_backward
step_model(
ddp_model.module,
input[self.rank : (self.rank + 1)],
target[self.rank : (self.rank + 1)],
)
for i, j in zip(model.parameters(), ddp_model.parameters()):
self.assertNotEqual(i.grad, j.grad)
else:
step_model(
ddp_model,
input[self.rank : (self.rank + 1)],
target[self.rank : (self.rank + 1)],
)
for i, j in zip(model.parameters(), ddp_model.parameters()):
self.assertEqual(i.grad, j.grad, rtol=1.3e-06, atol=5e-5)
# Shuffle the input so that DDP input is different
torch.manual_seed(1337 + iteration)
input = input[torch.randperm(global_batch_size)]
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_accumulate_gradients_module(self):
self._test_accumulate_gradients_module()
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_accumulate_gradients_module_with_grad_is_view(self):
self._test_accumulate_gradients_module(gradient_as_bucket_view=True)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_failure_recovery(self):
process_group = self._get_process_group()
# need to create a separate file for the recovered FileStore, because
# the original one will be deleted when destructing the first FileStore.
recovery_filename = self.file_name + "_recovery"
if self.rank == 0:
# the file will be deleted by the recovered FileStore
open(recovery_filename, "w").close()
# not necessary to run barrier here, as DDP will synchronize
class TestModel(nn.Module):
def __init__(self):
super().__init__()
self.fc1 = nn.Linear(2, 10, bias=False)
self.fc2 = nn.Linear(10, 4, bias=False)
self.relu = nn.ReLU()
def forward(self, x):
x = self.relu(self.fc1(x))
x = self.relu(self.fc2(x))
return F.softmax(x, dim=1)
device_id = gpus_for_rank(self.world_size)[self.rank][0]
model = TestModel().float().to(device_id)
ddp = DistributedDataParallel(
model,
device_ids=[device_id],
process_group=process_group,
)
batch_size = 4
criterion = nn.CrossEntropyLoss()
input = torch.rand([batch_size, 2], dtype=torch.float)
target = torch.LongTensor([random.randrange(4) for _ in range(batch_size)]).to(
device_id
)
for _ in range(6):
output = ddp(input)
loss = criterion(output, target)
loss.backward()
del ddp
c10d.destroy_process_group(process_group)
store = c10d.FileStore(recovery_filename, self.world_size)
c10d.init_process_group("nccl", store=store, rank=self.rank, world_size=self.world_size)
process_group = c10d.distributed_c10d._get_default_group()
ddp = DistributedDataParallel(
model,
device_ids=[device_id],
process_group=process_group,
)
input = torch.rand([batch_size, 2], dtype=torch.float)
target = torch.LongTensor([random.randrange(4) for _ in range(batch_size)]).to(
device_id
)
for _ in range(6):
output = ddp(input)
loss = criterion(output, target)
loss.backward()
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_pass_default_pg(self):
dist.init_process_group(
"nccl",
init_method=f"file://{self.file_name}",
world_size=self.world_size,
rank=self.rank,
)
default_pg = c10d.distributed_c10d._get_default_group()
dist.destroy_process_group(default_pg)
self.assertFalse(dist.is_initialized())
def _test_grad_layout(self, replica_devices, layer_devs, local_batch_size):
process_group = self._get_process_group()
global_batch_size = local_batch_size * self.world_size
# Carry out some trials with small buckets and some with big buckets.
bucketsizes = (0.000001, 25)
# Tuples of lists. Each list describes per-layer characteristics for one trial.
layer_formats = (
[torch.contiguous_format] * 4,
[torch.channels_last] * 2 + [torch.contiguous_format] * 2,
[torch.channels_last] * 4,
)
layer_dtypes = (
[torch.float] * 4,
[torch.float] * 2 + [torch.half] * 2,
[torch.half] * 4,
)
input_dev = layer_devs[0] if isinstance(layer_devs, list) else layer_devs
target_dev = layer_devs[-1] if isinstance(layer_devs, list) else layer_devs
input = torch.randn(
(global_batch_size, 8, 8, 8), device=input_dev, dtype=torch.float
)
target = torch.randn(
(global_batch_size, 8, 4, 4), device=target_dev, dtype=torch.float
)
local_batch_start = self.rank * local_batch_size
local_batch_end = (self.rank + 1) * local_batch_size
# Reducer.cpp sneakily creates one "initial bucket" that ignores the "bucket_cap_mb"
# argument. The following makes sure the initial bucket also complies.
@contextmanager
def first_bucket_size(ddp_bucket_mb):
old_DEFAULT_FIRST_BUCKET_BYTES = dist._DEFAULT_FIRST_BUCKET_BYTES
dist._DEFAULT_FIRST_BUCKET_BYTES = int(ddp_bucket_mb * 1.0e6)
try:
yield
finally:
dist._DEFAULT_FIRST_BUCKET_BYTES = old_DEFAULT_FIRST_BUCKET_BYTES
with torch.backends.cudnn.flags(
enabled=True, deterministic=True, benchmark=False
):
for formats, dtypes, bucketsize in product(
layer_formats, layer_dtypes, bucketsizes
):
with first_bucket_size(bucketsize):
model_msg = (
f"rank = {self.rank} formats = {formats} dtypes = {dtypes} bucketsize = {bucketsize} "
)
try:
m = ConvNet(layer_devs, formats, dtypes)
m_ddp = DistributedDataParallel(
copy.deepcopy(m),
device_ids=replica_devices,
process_group=process_group,
bucket_cap_mb=bucketsize,
)
opt = torch.optim.SGD(m.parameters(), lr=0.1)
opt_ddp = torch.optim.SGD(m_ddp.parameters(), lr=0.1)
has_half = any(p.dtype is torch.half for p in m.parameters())
tol = 1.0e-3 if has_half else 1.0e-5
except BaseException:
# Prints case-specific debugging info to narrow down failing case.
print(
"Caught exception during model creation for " + model_msg,
flush=True,
)
raise
# 3 iters: First iter creates grads, second iter retests after rebucketing,
# third iter tries zeroed grads.
for it in range(3):
iter_msg = f"iter = {it} " + model_msg
named_msg = iter_msg
try:
F.mse_loss(m(input).float(), target).backward()
F.mse_loss(
m_ddp(input[local_batch_start:local_batch_end]).float(),
target[local_batch_start:local_batch_end],
).backward()
for i, ((layer_name, m_child), m_ddp_child) in enumerate(
zip(m.named_children(), m_ddp.module.children())
):
named_msg = layer_name + ".weight" + " " + iter_msg
self.assertTrue(
m_child.weight.grad.is_contiguous(
memory_format=formats[i]
),
named_msg,
)
self.assertTrue(
m_ddp_child.weight.grad.is_contiguous(
memory_format=formats[i]
),
named_msg,
)
for j, ((param_name, p), p_ddp) in enumerate(
zip(
m_child.named_parameters(),
m_ddp_child.parameters(),
)
):
named_msg = (
layer_name + "." + param_name + " " + iter_msg
)
self.assertEqual(
p.grad, p_ddp.grad, rtol=tol, atol=tol
)
opt.step()
opt_ddp.step()
if it == 0:
for p, p_ddp in zip(m.parameters(), m_ddp.parameters()):
p.grad = None
p_ddp.grad = None
else:
m.zero_grad()
m_ddp.zero_grad()
except BaseException:
# Makes sure we still get info if an error occurred somewhere other than the asserts.
print(
"Caught exception during iterations at " + named_msg,
flush=True,
)
raise
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_grad_layout_1devicemodule_1replicaperprocess(self):
dev0 = torch.device("cuda:" + str(gpus_for_rank(self.world_size)[self.rank][0]))
# Tells DDP to use just one device.
replica_devices = [dev0]
# Tells _test_grad_layout to construct ConvNet with all layers on this process's first assigned device.
layer_devs = dev0
local_batch_size = 8
self._test_grad_layout(replica_devices, layer_devs, local_batch_size)
@requires_nccl()
@skip_if_lt_x_gpu(4)
@skip_if_rocm
def test_grad_layout_2devicemodule(self):
int_devices = gpus_for_rank(self.world_size)[self.rank][:2]
dev0 = torch.device("cuda:" + str(int_devices[0]))
dev1 = torch.device("cuda:" + str(int_devices[1]))
# DDP's default behavior for a multi-device module is "don't replicate."
replica_devices = None
# Tells _test_grad_layout to constructs this process's ConvNet on 2 devices, with 2 layers on each device.
layer_devs = [dev0] * 2 + [dev1] * 2
local_batch_size = 8
self._test_grad_layout(replica_devices, layer_devs, local_batch_size)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_param_layout_mismatch_error(self):
process_group = self._get_process_group()
dev0 = torch.device("cuda:" + str(gpus_for_rank(self.world_size)[self.rank][0]))
layer_devs = dev0
layer_formats = (
[torch.contiguous_format] * 4
if self.rank == 0
else [torch.channels_last] * 4
)
layer_dtypes = [torch.float] * 4
m = ConvNet(layer_devs, layer_formats, layer_dtypes)
if self.rank == 0:
m_ddp = DistributedDataParallel(
m, device_ids=[dev0], process_group=process_group
)
else:
with self.assertRaisesRegex(
RuntimeError,
".* appears not to match strides of the same param in process 0",
):
m_ddp = DistributedDataParallel(
m, device_ids=[dev0], process_group=process_group
)
def _gpu_model_with_ddp_comm_hook(
self,
process_group,
hook=None,
gradient_as_bucket_view=False,
state=None,
static_graph=False,
):
device_id = gpus_for_rank(self.world_size)[self.rank][0]
gpu_model = DistributedDataParallel(
ModuleForDdpCommHook().to(device_id),
device_ids=[device_id],
process_group=process_group,
gradient_as_bucket_view=gradient_as_bucket_view,
static_graph=static_graph,
)
# Register a DDP communication hook if any.
if hook is not None:
gpu_model.register_comm_hook(state, hook)
return gpu_model
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_ddp_comm_hook_future_passing_gpu_nccl(self):
"""
This unit test verifies whether the Future object is passed properly using nccl backend.
The hook callback function creates a Future object and sets a value to it.
"""
process_group = self._get_process_group()
# Get GPU model with simple_hook registered.
gpu_model = self._gpu_model_with_ddp_comm_hook(process_group, self._simple_hook)
# check whether the grads are equal to what simple_hook's then callback returns.
# without the comm_hook, result would be 0.25 * torch.ones(2, 2).
self._run_and_verify_hook(gpu_model, 8, 2 * torch.ones(2, 2))
def _test_ddp_comm_hook_allreduce_hook_nccl(
self, gradient_as_bucket_view=False, static_graph=False
):
"""
This unit test verifies whether a DDP communication hook that just calls
allreduce gives the same result with the case of no hook registered.
Without the then callback, the future_value in reducer is no longer
a PyObject, and this unit test verifies future_value is properly checked.
"""
process_group = self._get_process_group()
def allreduce_hook(
state: object, bucket: dist.GradBucket
) -> torch.futures.Future[torch.Tensor]:
tensors = [bucket.buffer() / self.world_size]
return (
process_group.allreduce(tensors)
.get_future()
.then(lambda fut: fut.value()[0])
)
# Get GPU model with allreduce_hook registered.
gpu_model = self._gpu_model_with_ddp_comm_hook(
process_group, allreduce_hook, gradient_as_bucket_view, static_graph
)
# check whether the grads are equal to what DDP without hook would return.
self._run_and_verify_hook(gpu_model, 8, 0.25 * torch.ones(2, 2))
def _test_default_ddp_comm_hooks_nccl(self, gradient_as_bucket_view=False):
"""
This unit test verifies whether default Python DDP communication hooks ALLREDUCE, FP16_COMPRESS
and BF16_COMPRESS, can give the same result with the case of no hook registered.
"""
process_group = self._get_process_group()
# For these default DDP comm hooks, the only state is process group.
state = process_group
hook_options = [default.allreduce_hook, default.fp16_compress_hook]
if (
not TEST_WITH_ROCM
and BFLOAT16_AVAILABLE
and c10d.is_nccl_available()
and torch.cuda.nccl.version() >= (2, 10)
):
hook_options.append(default.bf16_compress_hook)
for hook in hook_options:
# Get GPU model with the hook registered.
# The first arg 'process_group' is used for initializing the test environment,
# so it cannot be replaced by 'state', although they have the same value.
gpu_model = self._gpu_model_with_ddp_comm_hook(
process_group, hook, gradient_as_bucket_view, state
)
# check whether the grads are equal to what DDP without hook would return.
self._run_and_verify_hook(gpu_model, 8, 0.25 * torch.ones(2, 2))
def _test_fp16_compress_wrapper(self, gradient_as_bucket_view=False):
"""
This unit test verifies whether wrapping the ALLREDUCE and POWER_SGD hooks with
the FP16_WRAPPER can give the same result as when there is no hook registered.
"""
process_group = self._get_process_group()
powerSGD_state = powerSGD.PowerSGDState(process_group=process_group)
hook_args = [
(powerSGD.powerSGD_hook, powerSGD_state),
(default.allreduce_hook, process_group),
]
for hook, state in hook_args:
gpu_model = self._gpu_model_with_ddp_comm_hook(
process_group,
default.fp16_compress_wrapper(hook),
gradient_as_bucket_view,
state,
)
# check whether the grads are equal to what DDP without hook would return.
self._run_and_verify_hook(gpu_model, 8, 0.25 * torch.ones(2, 2))
def _test_bf16_compress_wrapper(self, gradient_as_bucket_view=False):
"""
This unit test verifies whether wrapping the ALLREDUCE and POWER_SGD hooks with
the BF16_WRAPPER can give the same result as when there is no hook registered.
"""
process_group = self._get_process_group()
powerSGD_state = powerSGD.PowerSGDState(process_group=process_group)
hook_args = [
(powerSGD.powerSGD_hook, powerSGD_state),
(default.allreduce_hook, process_group),
]
for hook, state in hook_args:
gpu_model = self._gpu_model_with_ddp_comm_hook(
process_group,
default.bf16_compress_wrapper(hook),
gradient_as_bucket_view,
state,
)
# check whether the grads are equal to what DDP without hook would return.
self._run_and_verify_hook(gpu_model, 8, 0.25 * torch.ones(2, 2))
def _test_powerSGD_ddp_comm_hook_nccl(self, gradient_as_bucket_view=False):
"""
This unit test verifies whether Python DDP communication hook POWER_SGD
can give the same result with the case of no hook registered.
"""
process_group = self._get_process_group()
# Get GPU model with the hook registered.
# Test the hook with different algorithmic configs.
for use_error_feedback, warm_start, batch_tensors_with_same_shape in product(
[True, False], [True, False], [True, False],
):
state = powerSGD.PowerSGDState(
process_group=process_group,
matrix_approximation_rank=1,
use_error_feedback=use_error_feedback,
warm_start=warm_start,
batch_tensors_with_same_shape=batch_tensors_with_same_shape,
)
for hook in [powerSGD.powerSGD_hook, powerSGD.batched_powerSGD_hook]:
gpu_model = self._gpu_model_with_ddp_comm_hook(
process_group, hook, gradient_as_bucket_view, state
)
# check whether the grads are equal to what DDP without hook would return.
self._run_and_verify_hook(gpu_model, 8, 0.25 * torch.ones(2, 2))
def _test_builtin_ddp_comm_hooks_nccl(self, gradient_as_bucket_view=False):
"""
This unit test verifies whether built-in C++ DDP communication hooks ALLREDUCE and FP16_COMPRESS
can give the same result with the case of no hook registered.
"""
process_group = self._get_process_group()
for comm_hook_type in [
dist.BuiltinCommHookType.ALLREDUCE,
dist.BuiltinCommHookType.FP16_COMPRESS,
]:
# Get GPU model with the built-in communication hook.
gpu_model = self._gpu_model_with_builtin_ddp_comm_hook(
process_group, comm_hook_type, gradient_as_bucket_view
)
# check whether the grads are equal to what DDP without hook would return.
self._run_and_verify_hook(gpu_model, 8, 0.25 * torch.ones(2, 2))
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_ddp_comm_hook_allreduce_hook_nccl(self):
self._test_ddp_comm_hook_allreduce_hook_nccl()
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_default_ddp_comm_hooks_nccl(self):
self._test_default_ddp_comm_hooks_nccl()
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_fp16_compress_wrapper_nccl(self):
self._test_fp16_compress_wrapper()
@requires_nccl()
@requires_nccl_version((2, 10), "Need NCCL 2.10+ for BF16_COMPRESS")
@skip_but_pass_in_sandcastle_if(
not BFLOAT16_AVAILABLE,
"BFloat16 is only supported by CUDA 11+",
)
@skip_if_lt_x_gpu(2)
def test_bf16_compress_wrapper_nccl(self):
self._test_bf16_compress_wrapper()
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_builtin_ddp_comm_hooks_nccl(self):
self._test_builtin_ddp_comm_hooks_nccl()
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_powerSGD_ddp_comm_hook_nccl(self):
self._test_powerSGD_ddp_comm_hook_nccl()
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_ddp_comm_hook_allreduce_hook_nccl_grad_is_view(self):
self._test_ddp_comm_hook_allreduce_hook_nccl(gradient_as_bucket_view=True)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_ddp_comm_hook_allreduce_hook_nccl_static_graph(self):
self._test_ddp_comm_hook_allreduce_hook_nccl(static_graph=True)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_default_ddp_comm_hooks_nccl_is_view(self):
self._test_default_ddp_comm_hooks_nccl(gradient_as_bucket_view=True)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_fp16_compress_wrapper_is_view(self):
self._test_fp16_compress_wrapper(gradient_as_bucket_view=True)
@requires_nccl()
@requires_nccl_version((2, 10), "Need NCCL 2.10+ for BF16_COMPRESS")
@skip_but_pass_in_sandcastle_if(
not BFLOAT16_AVAILABLE,
"BFloat16 is only supported by CUDA 11+",
)
@skip_if_lt_x_gpu(2)
def test_bf16_compress_wrapper_is_view(self):
self._test_bf16_compress_wrapper(gradient_as_bucket_view=True)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_builtin_ddp_comm_hooks_nccl_grad_is_view(self):
self._test_builtin_ddp_comm_hooks_nccl(gradient_as_bucket_view=True)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_powerSGD_ddp_comm_hook_nccl_grad_is_view(self):
self._test_powerSGD_ddp_comm_hook_nccl(gradient_as_bucket_view=True)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_ddp_comm_hook_allreduce_with_then_hook_nccl(self):
"""
This unit test verifies whether a DDP communication hook that calls allreduce and then
multiplies the result by ten and divides by two gives the expected result.
"""
process_group = self._get_process_group()
def allreduce_with_then_hook(
state: object, bucket: dist.GradBucket
) -> torch.futures.Future[torch.Tensor]:
tensors = [bucket.buffer() / self.world_size]
fut = process_group.allreduce(tensors).get_future()
def mult(fut):
# Multiply the result by 10.
return 10 * fut.value()[0]
def div(fut):
# Divide the result by 2.
return 0.5 * fut.value()
return fut.then(mult).then(div)
# Get GPU model with allreduce_with_then_hook registered.
gpu_model = self._gpu_model_with_ddp_comm_hook(
process_group, allreduce_with_then_hook
)
# check whether the grads are equal to what allreduce returns multiplied by 5.
# without the comm_hook, result would be still 0.25 * torch.ones(2, 2).
self._run_and_verify_hook(gpu_model, 8, 1.25 * torch.ones(2, 2))
class AcceptsParam(torch.nn.Module):
def __init__(self, p, factor):
super().__init__()
self.a = p
self.f = factor
def forward(self, input):
return input + self.a * self.f
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_ddp_weight_sharing(self):
process_group = self._get_process_group()
size = 2048 * 2048
dev = self.rank
world = self.world_size
p = torch.nn.Parameter(torch.randn(size, requires_grad=True))
for try_set_to_none, use_bucket_view in product((False, True), (False, True)):
m = torch.nn.Sequential(
self.AcceptsParam(p, dev + 1), self.AcceptsParam(p, dev + 1)
).cuda(dev)
m = torch.nn.parallel.DistributedDataParallel(
m,
bucket_cap_mb=1,
gradient_as_bucket_view=use_bucket_view,
device_ids=[dev],
process_group=process_group,
)
for i in range(3):
m.zero_grad(set_to_none=try_set_to_none)
m(1).sum().backward()
# Each param value is multiplied by "rank + 1" twice in forward, so the grad
# values produced by a particular rank should be 2. * (rank + 1).
# Summing these over ranks and dividing by world size gives the expected result:
analytic = torch.full_like(
p, 2.0 * (world * (world + 1.0) / 2.0) / world, device=dev
)
for name, p in m.named_parameters():
self.assertEqual(
p.grad,
analytic,
"mismatch at "
+ name
+ ".grad for "
+ f"set_to_none = {try_set_to_none}, use_bucket_view = {use_bucket_view}",
)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_ddp_packed_sequence(self):
"""
Tests that DDP with ``device_ids`` specified can run a forward and
backward pass with ``PackedSequence`` s with parity compared to a local
version of the model.
"""
store = c10d.FileStore(self.file_name, self.world_size)
process_group = dist.init_process_group(
"nccl",
world_size=self.world_size,
rank=self.rank,
store=store,
)
seqs = ["sequence_sequence", "seq", "sequence"]
vocab = ["<pad>"] + sorted({ch for seq in seqs for ch in seq})
vectorized_seqs = [[vocab.index(tok) for tok in seq] for seq in seqs]
# Set the seed to make the embedding and LSTM deterministic (even
# across ranks since DDP broadcasts parameters from rank 0)
torch.manual_seed(0)
embed = nn.Embedding(len(vocab), 4) # keep on CPU
lstm = nn.LSTM(input_size=4, hidden_size=2, batch_first=True).to(self.rank)
lstm_ddp = DistributedDataParallel(
copy.deepcopy(lstm),
device_ids=[self.rank],
process_group=process_group,
)
for p1, p2 in zip(lstm.parameters(), lstm_ddp.module.parameters()):
self.assertEqual(p1, p2)
seq_lengths = torch.LongTensor(list(map(len, vectorized_seqs)))
seq_tensor = torch.Tensor(
torch.zeros((len(vectorized_seqs), seq_lengths.max()))
).long()
for i, (seq, seq_len) in enumerate(zip(vectorized_seqs, seq_lengths)):
seq_tensor[i, :seq_len] = torch.LongTensor(seq)
seq_lengths, permutation_idx = seq_lengths.sort(0, descending=True)
seq_tensor = seq_tensor[permutation_idx]
embedded_seq_tensor = embed(seq_tensor)
packed_input = torch.nn.utils.rnn.pack_padded_sequence(
embedded_seq_tensor, seq_lengths, batch_first=True,
)
packed_input_ddp = torch.nn.utils.rnn.pack_padded_sequence(
embedded_seq_tensor.detach().clone(), seq_lengths, batch_first=True,
)
# Move the input to GPU explicitly for the local model
packed_output, (ht, ct) = lstm(packed_input.to(self.rank))
# Let DDP move the input to GPU internally
packed_output_ddp, (ht_ddp, ct_ddp) = lstm_ddp(packed_input_ddp)
self.assertEqual(packed_output.data, packed_output_ddp.data)
self.assertEqual(ht, ht_ddp)
self.assertEqual(ct, ct_ddp)
packed_output.data.sum().backward()
packed_output_ddp.data.sum().backward()
for p1, p2 in zip(lstm.parameters(), lstm_ddp.parameters()):
self.assertEqual(p1.grad, p2.grad)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_channels_last_contig(self):
process_group = self._get_process_group()
device = torch.device(f"cuda:{self.rank}")
tensor = torch.ones((2, 16, 768, 1152), dtype=torch.float32, device=device).to(memory_format=torch.channels_last)
process_group.broadcast([tensor]).wait()
class WorkHookTest(MultiProcessTestCase):
@property
def world_size(self):
return 2
def setUp(self):
super().setUp()
# set NCCL_ENABLE_TIMING to enable timing for CUDAEvents
# in ProcessGroup Work
os.environ["NCCL_ENABLE_TIMING"] = "1"
self._spawn_processes()
def tearDown(self):
super().tearDown()
del os.environ["NCCL_ENABLE_TIMING"]
try:
os.remove(self.file_name)
except OSError:
pass
def _get_store(self):
return dist.FileStore(self.file_name, self.world_size)
def _get_process_group(self):
store = self._get_store()
c10d.init_process_group("nccl", store=store, rank=self.rank, world_size=self.world_size)
return c10d.distributed_c10d._get_default_group()
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_on_completion_hook_broadcast(self):
pg = self._get_process_group()
num_hook_fired = 0
durations: List[float] = []
def hook(work_info: torch._C._distributed_c10d.WorkInfo):
nonlocal num_hook_fired, durations
num_hook_fired += 1
durations.append(work_info.active_duration.total_seconds())
pg._register_on_completion_hook(hook)
tensor = torch.ones([2, 3]).cuda(self.rank) * self.rank
pg.broadcast([tensor]).wait()
pg.broadcast([tensor]).wait()
# N.B.: destroy_process_group is necessary to wait for
# all pending works to finish.
c10d.destroy_process_group(pg)
self.assertEqual(num_hook_fired, 2)
self.assertEqual(len(durations), 2)
for duration in durations:
self.assertTrue(duration > 0)
self.assertEqual(tensor, torch.zeros([2, 3]).cuda(self.rank))
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_on_completion_hook_mixed_ops(self):
pg = self._get_process_group()
num_hook_fired = 0
durations: List[float] = []
def hook(work_info: torch._C._distributed_c10d.WorkInfo):
nonlocal num_hook_fired, durations
num_hook_fired += 1
durations.append(work_info.active_duration.total_seconds())
pg._register_on_completion_hook(hook)
tensor = torch.ones([2, 3]).cuda(self.rank)
tensor_list = [torch.empty_like(tensor) for _ in range(self.world_size)]
# intentionall using async ops.
pg.allreduce(tensor)
pg.allgather(tensor_list, tensor)
pg.allreduce(tensor)
# N.B.: destroy_process_group is necessary to wait for
# all pending works to finish.
c10d.destroy_process_group(pg)
self.assertEqual(num_hook_fired, 3)
self.assertEqual(len(durations), 3)
for duration in durations:
self.assertTrue(duration > 0)
self.assertEqual(
tensor,
torch.ones([2, 3]).cuda(self.rank) * self.world_size * self.world_size,
)
self.assertEqual(
tensor_list,
[torch.ones([2, 3]).cuda(self.rank) * self.world_size for _ in range(self.world_size)],
)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_on_completion_hook_with_ddp(self):
pg = self._get_process_group()
num_hook_fired: Dict[int, int] = {}
durations: Dict[OpType, List[float]] = {}
def hook(work_info: torch._C._distributed_c10d.WorkInfo):
nonlocal num_hook_fired, durations
op_type = work_info.op_type
if op_type not in num_hook_fired:
num_hook_fired[op_type] = 0
durations[op_type] = []
num_hook_fired[op_type] += 1
durations[op_type].append(work_info.active_duration.total_seconds())
pg._register_on_completion_hook(hook)
nlayers = 10
net = nn.Sequential(
*[nn.Linear(1000, 1000, bias=False) for _ in range(nlayers)]
).to(self.rank)
ddp = DistributedDataParallel(
net,
device_ids=[self.rank],
process_group=pg,
bucket_cap_mb=1,
)
pg._wait_for_pending_works()
# DDP is expected to synchronize model parameter by broadcasting
# from rank0 to other ranks. However, this is DDP's internal implementation,
# which is subject to change in future versions.
self.assertTrue(num_hook_fired[OpType.BROADCAST] > 0)
ctor_allreduce = num_hook_fired[OpType.ALLREDUCE] if OpType.ALLREDUCE in num_hook_fired else 0
x = torch.zeros(2, 1000).cuda(self.rank)
ddp(x).sum().backward()
c10d.destroy_process_group(pg)
self.assertTrue(OpType.ALLREDUCE in num_hook_fired)
# The number of allreduce ops depend on DDP internal implementation, but
# there should be at least one allreduce.
self.assertTrue(num_hook_fired[OpType.ALLREDUCE] - ctor_allreduce > 0)
self.assertTrue(all(duration > 0 for duration in chain(*(durations.values()))))
# Not testing FSDP due to https://github.com/pytorch/pytorch/issues/90848.
# We cannot disable workCleanupLoop() as hooks are fired in that thread.
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_on_completion_hook_all_gather_object(self):
torch.cuda.set_device(self.rank)
pg = self._get_process_group()
num_hook_fired: Dict[int, int] = {}
durations: Dict[OpType, List[float]] = {}
def hook(work_info: torch._C._distributed_c10d.WorkInfo):
nonlocal num_hook_fired, durations
op_type = work_info.op_type
if op_type not in num_hook_fired:
num_hook_fired[op_type] = 0
durations[op_type] = []
num_hook_fired[op_type] += 1
durations[op_type].append(work_info.active_duration.total_seconds())
pg._register_on_completion_hook(hook)
obj = {"rank": self.rank, "world_size": self.world_size}
obj_list = [None for _ in range(self.world_size)]
c10d.all_gather_object(obj_list, obj, group=pg)
for r, o in enumerate(obj_list):
self.assertTrue(isinstance(o, dict))
self.assertTrue(set(o.keys()), {"rank", "world_size"})
self.assertEqual(o["rank"], r)
self.assertEqual(o["world_size"], self.world_size)
c10d.destroy_process_group(pg)
self.assertTrue(OpType.ALLGATHER in num_hook_fired)
self.assertEqual(len(num_hook_fired), 1)
# two allgathers, one for size and another for values
self.assertEqual(num_hook_fired[OpType.ALLGATHER], 2)
self.assertTrue(all(duration > 0 for duration in durations[OpType.ALLGATHER]))
class NcclErrorHandlingTest(MultiProcessTestCase):
def setUp(self):
super().setUp()
# Need to skip return code checking for these tests since the child
# processes don't exit cleanly.
self.skip_return_code_checks = [
self.test_nccl_errors_blocking_abort.__wrapped__,
self.test_nccl_errors_blocking_sigkill.__wrapped__,
self.test_nccl_errors_blocking_sigterm.__wrapped__,
self.test_nccl_errors_blocking_nonzero_exit.__wrapped__,
]
# NCCL_BLOCKING_WAIT overrides NCCL_ASYNC_ERROR_HANDLING hence tests
# that use NCCL_BLOCKING_WAIT will test it as expected.
os.environ["NCCL_ASYNC_ERROR_HANDLING"] = "1"
self._spawn_processes()
def tearDown(self):
super().tearDown()
try:
os.remove(self.file_name)
except OSError:
pass
@property
def op_timeout_sec(self):
return 1
@property
def world_size(self):
return 3
@property
def blocking_wait_error_msg(self):
return "timeout"
def _run_all_reduce(self, pg):
pg.allreduce(torch.rand(10).cuda(self.rank))
@requires_nccl()
@requires_nccl_version((2, 4, 0), "Need NCCL 2.4+ for error checking")
@skip_if_lt_x_gpu(3)
@skip_if_rocm
@skip_but_pass_in_sandcastle("Test does not pass when run locally")
def test_nccl_errors_nonblocking(self):
# Note: we unset and restore NCCL_ASYNC_ERROR_HANDLING for this test
# since test_c10d_common runs with async error handling by default, but this
# tests behavior when it is not enabled.
prev_nccl_async_error_handling = os.environ.get(
"NCCL_ASYNC_ERROR_HANDLING", None
)
os.environ["NCCL_ASYNC_ERROR_HANDLING"] = "0"
store = c10d.FileStore(self.file_name, self.world_size)
process_group = c10d.ProcessGroupNCCL(store, self.rank, self.world_size)
process_group.allreduce(torch.rand(10).cuda(self.rank))
if self.rank == 0:
# This allreduce does not block Python thread as allreduce enqueues
# the cuda operation, and then wait only blocks the current cuda
# stream.
work = process_group.allreduce(torch.rand(10).cuda(self.rank))
work.wait()
# Now the work scheduled next should hang forever since the previous
# allreduce will never complete.
t = threading.Thread(target=self._run_all_reduce, args=(process_group,))
t.daemon = True
t.start()
t.join(int(get_timeout(self.id()) / 5))
self.assertTrue(t.is_alive())
if prev_nccl_async_error_handling is not None:
os.environ["NCCL_ASYNC_ERROR_HANDLING"] = prev_nccl_async_error_handling
def _test_nccl_errors_blocking(self, func):
store = c10d.FileStore(self.file_name, self.world_size)
process_group = c10d.ProcessGroupNCCL(
store,
self.rank,
self.world_size,
timeout=timedelta(seconds=10),
)
process_group.allreduce(torch.rand(10).cuda(self.rank))
if self.rank == 0:
work = process_group.allreduce(torch.rand(10).cuda(self.rank))
with self.assertRaisesRegex(RuntimeError, self.blocking_wait_error_msg):
# Operation would time out in blocking mode.
work.wait(timeout=timedelta(seconds=self.op_timeout_sec))
# Run some GPU operations to make sure cuda has not gotten stuck.
# It was observed cuda could get stuck if NCCL communicators were
# not properly aborted before throwing RuntimeError.
a = torch.rand(10).cuda(self.rank)
elif self.rank == 1:
# Clean up structures (ex: files for FileStore before going down)
del process_group
func()
@with_nccl_blocking_wait
@requires_nccl()
@requires_nccl_version((2, 4, 0), "Need NCCL 2.4+ for error checking")
@skip_if_lt_x_gpu(3)
@skip_if_rocm
def test_nccl_errors_blocking_clean_exit(self):
self._test_nccl_errors_blocking(lambda: sys.exit(0))
@with_nccl_blocking_wait
@requires_nccl()
@requires_nccl_version((2, 4, 0), "Need NCCL 2.4+ for error checking")
@skip_if_lt_x_gpu(3)
@skip_if_rocm
def test_nccl_errors_blocking_nonzero_exit(self):
self._test_nccl_errors_blocking(lambda: sys.exit(1))
@with_nccl_blocking_wait
@requires_nccl()
@requires_nccl_version((2, 4, 0), "Need NCCL 2.4+ for error checking")
@skip_if_lt_x_gpu(3)
@skip_if_rocm
@skip_but_pass_in_sandcastle(
"Frequently times out see https://github.com/pytorch/pytorch/issues/58920"
)
def test_nccl_errors_blocking_abort(self):
self._test_nccl_errors_blocking(lambda: os.abort())
@with_nccl_blocking_wait
@requires_nccl()
@requires_nccl_version((2, 4, 0), "Need NCCL 2.4+ for error checking")
@skip_if_lt_x_gpu(3)
@skip_if_rocm
def test_nccl_errors_blocking_sigkill(self):
self._test_nccl_errors_blocking(lambda: os.kill(os.getpid(), signal.SIGKILL))
@with_nccl_blocking_wait
@requires_nccl()
@requires_nccl_version((2, 4, 0), "Need NCCL 2.4+ for error checking")
@skip_if_lt_x_gpu(3)
@skip_if_rocm
def test_nccl_errors_blocking_sigterm(self):
self._test_nccl_errors_blocking(lambda: os.kill(os.getpid(), signal.SIGTERM))
@with_nccl_blocking_wait
@requires_nccl()
@requires_nccl_version((2, 4, 0), "Need NCCL 2.4+ for error checking")
@skip_if_lt_x_gpu(3)
def test_nccl_blocking_wait_with_barrier(self):
store = c10d.FileStore(self.file_name, self.world_size)
process_group = c10d.ProcessGroupNCCL(
store,
self.rank,
self.world_size,
timeout=timedelta(seconds=10),
)
process_group.barrier().wait()
if self.rank == 0:
with self.assertRaisesRegex(RuntimeError, self.blocking_wait_error_msg):
# This should timeout
process_group.barrier().wait(timeout=timedelta(seconds=self.op_timeout_sec))
def _run_invalid_nccl_blocking_wait_env(self, val):
os.environ["NCCL_BLOCKING_WAIT"] = val
store = c10d.FileStore(self.file_name, self.world_size)
with self.assertRaises(RuntimeError):
process_group = c10d.ProcessGroupNCCL(store, self.rank, self.world_size)
@requires_nccl()
@skip_if_lt_x_gpu(3)
def test_invalid_nccl_blocking_wait_env(self):
self._run_invalid_nccl_blocking_wait_env("abc")
self._run_invalid_nccl_blocking_wait_env("-1")
self._run_invalid_nccl_blocking_wait_env("2147483647")
self._run_invalid_nccl_blocking_wait_env("4294967295")
@with_nccl_blocking_wait
@requires_nccl()
@requires_gloo()
@skip_if_lt_x_gpu(3)
def test_nccl_timeout(self):
store = c10d.FileStore(self.file_name, self.world_size)
# Initialize process_group.
process_group = c10d.ProcessGroupNCCL(
store, self.rank, self.world_size, timeout=timedelta(seconds=10)
)
# Control gloo pg used as go-ahead signal/barrier
# to coordinate btwn ranks.
pg_gloo = c10d.ProcessGroupGloo(store, self.rank, self.world_size)
failed_collective_timeout = timedelta(milliseconds=100)
process_group.allreduce(torch.rand(10).cuda(self.rank)).wait(timeout=timedelta(seconds=5))
if self.rank == 0:
# This should timeout in about 1 second.
# Watchdog may abort timed out work resulting in NCCL error instead of operation timed out.
with self.assertRaisesRegex(RuntimeError, self.blocking_wait_error_msg):
process_group.allreduce(torch.rand(10).cuda(self.rank)).wait(timeout=failed_collective_timeout)
# Now do a barrier to tell other rank to go ahead.
pg_gloo.barrier().wait()
else:
# Wait on rank 0 to fail.
try:
pg_gloo.barrier().wait()
except Exception as e:
raise ValueError(f"Rank {self.rank} barrier timed out waiting for rank 0 with error: {str(e)}") from e
class CommTest(test_c10d_common.AbstractCommTest, MultiProcessTestCase):
@property
def device(self):
return f"cuda:{self.rank}"
def setUp(self):
super().setUp()
# NCCL_BLOCKING_WAIT overrides NCCL_ASYNC_ERROR_HANDLING hence tests
# that use NCCL_BLOCKING_WAIT will test it as expected.
os.environ["NCCL_ASYNC_ERROR_HANDLING"] = "1"
self._spawn_processes()
def tearDown(self):
super().tearDown()
try:
os.remove(self.file_name)
except OSError:
pass
def _test_broadcast_coalesced(self, process_group, device, root_rank):
half = torch.float16
# No support for float16 for CPU tensors
if device == torch.device("cpu"):
half = torch.float32
target = torch.arange(60, dtype=half, device=device).chunk(5)
target += torch.arange(60, dtype=torch.float32, device=device).chunk(5)
target += torch.arange(60, dtype=half, device=device).chunk(5)
target += torch.arange(60, dtype=torch.float64, device=device).chunk(5)
target += torch.arange(60, dtype=half, device=device).chunk(5)
target += torch.arange(60, dtype=torch.float32, device=device).chunk(5)
# The tensors to pass to broadcast are idential to the target
# only on the process that is the root of the broadcast.
if self.rank == root_rank:
tensors = [tensor.clone() for tensor in target]
else:
tensors = [torch.zeros_like(tensor) for tensor in target]
if self.rank != root_rank:
self.assertNotEqual(tensors, target)
c10d._broadcast_coalesced(
process_group, tensors, buffer_size=256, src=root_rank
)
if self.rank != root_rank:
self.assertEqual(tensors, target)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_broadcast_coalesced_nccl(self):
store = c10d.FileStore(self.file_name, self.world_size)
c10d.init_process_group(backend="nccl", store=store, rank=self.rank, world_size=self.world_size)
process_group = c10d.distributed_c10d._get_default_group()
device = torch.device("cuda:%d" % self.rank)
ranks = [0, 1]
for root_rank in ranks:
self._test_broadcast_coalesced(process_group, device, root_rank)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_all_reduce_coalesced_nccl(self):
store = c10d.FileStore(self.file_name, self.world_size)
c10d.init_process_group(backend="nccl", store=store, rank=self.rank, world_size=self.world_size)
process_group = c10d.distributed_c10d._get_default_group()
device = torch.device("cuda:%d" % self.rank)
tensors = [torch.full((60 + i,), self.rank + 1 + i, device=device, dtype=torch.float) for i in range(5)]
torch.distributed.all_reduce_coalesced(tensors, group=process_group)
for i, t in enumerate(tensors):
self.assertEqual(t, torch.full_like(t, self.world_size * (i + (self.world_size + 1.) / 2.)))
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_sequence_num_set_default_pg_nccl(self):
torch.cuda.set_device(self.rank)
self._test_sequence_num_set_default_pg(backend="nccl")
@skip_if_lt_x_gpu(2)
@requires_nccl()
def test_sequence_num_incremented_nccl_default(self):
self._test_sequence_num_incremented_default_group("nccl")
@skip_if_lt_x_gpu(4)
@requires_nccl()
def test_sequence_num_incremented_nccl_subgroup(self):
if self.world_size < 4:
return skip_but_pass_in_sandcastle("Test requires world_size of at least 4")
self._test_sequence_num_incremented_subgroup("nccl")
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_sequence_num_set_nccl_new_group(self):
torch.cuda.set_device(self.rank)
self._test_sequence_num_set_new_group(backend="nccl")
def _test_pass_nccl_options(self, pg_opts):
store = c10d.FileStore(self.file_name, self.world_size)
# Test init_process_group accepts options
dist.init_process_group(
"nccl",
world_size=self.world_size,
rank=self.rank,
store=store,
pg_options=pg_opts,
)
# Test with new_group
pg = c10d.new_group([0, 1], pg_options=pg_opts)
# test the process group works as expected
t = torch.tensor([self.rank + 1] * 10).cuda(self.rank)
pg.allreduce(t).wait()
expected_tensor = torch.tensor([3] * 10).cuda(self.rank)
self.assertEqual(expected_tensor, t)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_pass_nccl_options_high_priority_stream(self):
pg_opts = c10d.ProcessGroupNCCL.Options()
pg_opts.is_high_priority_stream = True
self._test_pass_nccl_options(pg_opts)
@requires_nccl()
@requires_nccl_version((2, 17), "Need NCCL 2.17+ for configuring NCCL communicators")
@skip_if_lt_x_gpu(2)
def test_pass_nccl_options_config(self):
pg_opts = c10d.ProcessGroupNCCL.Options()
pg_opts.config.max_ctas = 4
pg_opts.config.min_ctas = 2
pg_opts.config.cga_cluster_size = 2
pg_opts.config.net_name = "Socket"
nccl_debug_file = tempfile.NamedTemporaryFile()
os.environ["NCCL_DEBUG"] = "INFO"
os.environ["NCCL_DEBUG_FILE"] = nccl_debug_file.name
# Tests functionality when passing nccl config
self._test_pass_nccl_options(pg_opts)
# Tests if comms were configured
nccl_debug_file_content = nccl_debug_file.read()
max_ctas = re.search(rb'Max CTAs.*(\d+)|$', nccl_debug_file_content).group(1)
min_ctas = re.search(rb'Min CTAs.*(\d+)|$', nccl_debug_file_content).group(1)
cga_cluster_size = re.search(rb'CGA cluster.*(\d+)|$', nccl_debug_file_content).group(1)
net_name = re.search(rb'Using network.([a-zA-z]+)|$', nccl_debug_file_content).group(1)
self.assertEqual(pg_opts.config.max_ctas, int(max_ctas))
self.assertEqual(pg_opts.config.min_ctas, int(min_ctas))
self.assertEqual(pg_opts.config.cga_cluster_size, int(cga_cluster_size))
self.assertEqual(pg_opts.config.net_name, net_name.decode())
@requires_nccl()
@skip_if_lt_x_gpu(4)
def test_nccl_barrier(self):
store = c10d.FileStore(self.file_name, self.world_size)
c10d.init_process_group(
backend="nccl", rank=self.rank, world_size=self.world_size, store=store
)
t = torch.tensor([self.rank + 1] * 10).cuda(2 * self.rank)
c10d.all_reduce(t)
expected_tensor = torch.tensor([3] * 10).cuda(2 * self.rank)
self.assertEqual(expected_tensor, t)
# Test with new_group
pg = c10d.new_group([0, 1])
t = torch.tensor([self.rank + 1] * 10).cuda(2 * self.rank)
pg.allreduce(t).wait()
self.assertEqual(expected_tensor, t)
pg = c10d.new_group([0])
if self.rank == 0:
t = torch.tensor([self.rank + 1] * 10).cuda(2 * self.rank)
expected_tensor = torch.tensor([self.rank + 1] * 10).cuda(2 * self.rank)
pg.allreduce(t).wait()
self.assertEqual(expected_tensor, t)
pg = c10d.new_group([1])
if self.rank == 1:
t = torch.tensor([self.rank + 1] * 10).cuda(2 * self.rank)
expected_tensor = torch.tensor([self.rank + 1] * 10).cuda(2 * self.rank)
pg.allreduce(t).wait()
self.assertEqual(expected_tensor, t)
@requires_nccl()
@skip_if_lt_x_gpu(4)
def test_nccl_barrier_timeout(self):
os.environ["ENABLE_NCCL_HEALTH_CHECK"] = "1"
store = c10d.FileStore(self.file_name, self.world_size)
if self.rank == 0:
with self.assertRaisesRegex(
RuntimeError, "Health check failure"
):
c10d.init_process_group(
backend="nccl",
rank=self.rank,
world_size=self.world_size,
store=store,
timeout=timedelta(seconds=10),
)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_nccl_barrier_device_ids(self):
store = c10d.FileStore(self.file_name, self.world_size)
c10d.init_process_group(
backend="nccl", rank=self.rank, world_size=self.world_size, store=store
)
c10d.barrier(device_ids=[self.rank])
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_nccl_barrier_device_ids_function_argument(self):
store = c10d.FileStore(self.file_name, self.world_size)
c10d.init_process_group(
backend="nccl", rank=self.rank, world_size=self.world_size, store=store
)
with self.assertRaisesRegex(TypeError, "Invalid function argument"):
c10d.barrier(device_ids=self.rank)
@requires_nccl()
@skip_if_lt_x_gpu(2)
@with_dist_debug_levels(levels=["DETAIL"])
def test_nccl_warn_not_in_group_debug_detail(self):
self._test_warn_not_in_group(backend="nccl")
@requires_nccl()
@skip_if_lt_x_gpu(2)
@with_dist_debug_levels(levels=["INFO"])
def test_nccl_warn_not_in_group_debug_info(self):
self._test_warn_not_in_group(backend="nccl")
@requires_nccl()
@skip_if_lt_x_gpu(2)
@with_dist_debug_levels(levels=["OFF"])
def test_nccl_warn_not_in_group_debug_off(self):
self._test_warn_not_in_group(backend="nccl")
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_nncl_rank_membership(self):
self._test_rank_membership(backend="nccl")
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_tensor_dtype_mismatch(self):
self._test_tensor_dtype_mismatch(backend="nccl")
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_tensor_dtype_complex(self):
self._test_tensor_dtype_complex(backend="nccl")
class CompilerTest(test_c10d_common.CompilerTest):
@property
def world_size(self):
return 2
def _get_default_group(self):
store = c10d.FileStore(self.file_name, self.world_size)
dist.init_process_group(
backend="nccl",
rank=self.rank,
world_size=self.world_size,
store=store,
)
return dist.distributed_c10d._get_default_group()
@skip_if_lt_x_gpu(2)
def test_allreduce_work_wait_gpu(self):
self._test_allreduce_work_wait(
torch.ones(2, 2, device=self.rank) * self.rank,
)
@skip_if_lt_x_gpu(2)
def test_allgather_work_wait_gpu(self):
self._test_allgather_work_wait(
torch.ones(2, 2, device=self.rank) * self.rank
)
@skip_if_lt_x_gpu(2)
def test_allgather_into_tensor_work_wait_gpu(self):
self._test_allgather_into_tensor_work_wait(
torch.ones(2, 2, device=self.rank) * self.rank
)
@skip_if_lt_x_gpu(2)
def test_reduce_scatter_work_wait_gpu(self):
self._test_reduce_scatter_work_wait(
torch.ones(2, 2, device=self.rank) * self.rank
)
@skip_if_lt_x_gpu(2)
def test_reduce_scatter_tensor_work_wait_gpu(self):
self._test_reduce_scatter_tensor_work_wait(
torch.ones(4, 4, device=self.rank) * self.rank
)
@skip_if_lt_x_gpu(2)
def test_broadcast_work_wait_gpu(self):
self._test_broadcast_work_wait(
torch.ones(2, 2, device=self.rank) * self.rank
)
@skip_if_lt_x_gpu(2)
def test_scatter_work_wait_gpu(self):
self._test_scatter_work_wait(
torch.ones(2, 2, device=self.rank) * self.rank
)
@skip_if_lt_x_gpu(2)
def test_alltoall_work_wait_gpu(self):
self._test_alltoall_work_wait(
torch.ones(2, 2, device=self.rank) * self.rank
)
@skip_if_lt_x_gpu(2)
def test_nested_comm_tensor_wrapping(self):
self._test_nested_comm_tensor_wrapping(
torch.ones(2, 2, device=self.rank) * self.rank
)
@skip_if_lt_x_gpu(2)
def test_consecutive_comm_work_wait_gpu(self):
self._test_consecutive_comm_work_wait(
torch.ones(2, 2, device=self.rank) * self.rank
)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_reduce_scatter_base_k(self):
store = dist.FileStore(self.file_name, self.world_size)
dist.init_process_group(
"nccl",
world_size=self.world_size,
rank=self.rank,
store=store,
)
output_tensor = torch.zeros(2, dtype=torch.int64).to(self.rank)
input_tensors = torch.arange(self.world_size * 2, dtype=torch.int64).to(self.rank)
input_tensors = torch.reshape(input_tensors, (self.world_size, 2))
dist.reduce_scatter_tensor(output_tensor, input_tensors)
self.assertEqual(output_tensor, input_tensors[self.rank] * self.world_size)
@requires_nccl()
@skip_if_lt_x_gpu(2)
def test_reduce_scatter_tensor_coalesced(self):
store = dist.FileStore(self.file_name, self.world_size)
dist.init_process_group(
"nccl",
world_size=self.world_size,
rank=self.rank,
store=store,
)
output_tensors = torch.zeros(2, 2).to(self.rank)
input_tensors = [torch.ones(2, 2).to(self.rank) for _ in range(self.world_size)]
with dist._coalescing_manager():
for i in range(self.world_size):
dist.reduce_scatter_tensor(output_tensors[i], input_tensors[i])
self.assertEqual(output_tensors, input_tensors[self.rank] * self.world_size)
class NcclProcessGroupWithDispatchedCollectivesTests(test_c10d_common.ProcessGroupWithDispatchedCollectivesTests):
@requires_nccl()
@skip_if_lt_x_gpu(1)
def test_collectives(self):
self._test_collectives(backend="nccl")
@requires_nccl()
@skip_if_lt_x_gpu(1)
def test_allreduce_coalesced(self):
self._test_allreduce_coalesced(backend="nccl")
@requires_nccl()
@skip_if_lt_x_gpu(1)
def test_all_to_all_single(self):
self._test_all_to_all_single(backend="nccl")
@requires_nccl()
@skip_if_lt_x_gpu(1)
def test_allgather_base(self):
store = dist.FileStore(self.file_name, self.world_size)
dist.init_process_group(
"nccl",
world_size=self.world_size,
rank=self.rank,
store=store,
)
device = "cuda"
tensor = torch.ones(10, 10, device=torch.device(device))
output_tensor = torch.zeros(10, 10, device=torch.device(device))
dist.all_gather_into_tensor(output_tensor, tensor)
self.assertEqual(output_tensor, tensor)
class LargeCommTest(test_c10d_common.AbstractLargeCommTest, MultiProcessTestCase):
def setUp(self):
super().setUp()
# NCCL_BLOCKING_WAIT overrides NCCL_ASYNC_ERROR_HANDLING hence tests
# that use NCCL_BLOCKING_WAIT will test it as expected.
os.environ["NCCL_ASYNC_ERROR_HANDLING"] = "1"
self._spawn_processes()
def tearDown(self):
super().tearDown()
try:
os.remove(self.file_name)
except OSError:
pass
@property
def device(self):
return self.rank
@requires_nccl()
@skip_if_lt_x_gpu(4)
def test_new_group_local_sync(self):
self._test_new_group_local_sync(backend="nccl")
@requires_nccl()
@skip_if_lt_x_gpu(4)
def test_new_group_local_sync_sanity_check(self):
self._test_new_group_local_sync_sanity_check(backend="nccl")
@requires_nccl()
@skip_if_lt_x_gpu(4)
def test_new_group_local_sync_duplicated_pg(self):
self._test_new_group_local_sync_duplicate_pg(backend="nccl")
class SparseCollective(MultiProcessTestCase):
@property
def world_size(self):
return 1
def setUp(self):
super().setUp()
# NCCL_BLOCKING_WAIT overrides NCCL_ASYNC_ERROR_HANDLING hence tests
# that use NCCL_BLOCKING_WAIT will test it as expected.
os.environ["NCCL_ASYNC_ERROR_HANDLING"] = "1"
# self.num_gpus = torch.cuda.device_count()
self._spawn_processes()
def tearDown(self):
super().tearDown()
try:
os.remove(self.file_name)
except OSError:
pass
class ToyModel(nn.Module):
def __init__(self, rank, vocab_size, embedding_dim):
super().__init__()
self.embedding = nn.Embedding(vocab_size, embedding_dim, sparse=True).to(rank)
self.linear = nn.Linear(embedding_dim, 1).to(rank)
def forward(self, inputs):
embedded = self.embedding(inputs)
# embedded shape: (batch_size, sequence_length, embedding_dim)
flattened = torch.mean(embedded, dim=1)
# flattened shape: (batch_size, embedding_dim)
output = self.linear(flattened)
# output shape: (batch_size, 1)
return output
@requires_nccl()
@skip_if_lt_x_gpu(1)
def test_ddp_set_sparse_metadata(self):
store = dist.FileStore(self.file_name, self.world_size)
dist.init_process_group(
"nccl",
world_size=self.world_size,
rank=self.rank,
store=store,
)
vocab_size = 5
model = SparseCollective.ToyModel(self.rank, vocab_size=vocab_size, embedding_dim=10)
ddp_model = DistributedDataParallel(model)
inputs = torch.tensor([[1, 0, 0], [0, 0, 0], [0, 0, 0]]).to(self.rank)
# set sparse metadata on the DDP model
indices = torch.Tensor(list(range(vocab_size)))
ddp_model._set_sparse_metadata({"embedding.weight" : indices})
# forward pass
try:
output = ddp_model(inputs)
loss = output.sum()
# backward pass
loss.backward()
self.assertTrue(ddp_model.module.embedding.weight.grad.indices, indices)
except RuntimeError as e:
if "allreduce_sparse is only available in the NCCL experimental branch." in str(e):
pass
else:
# Rethrow the exception if it's a different error
raise
if __name__ == "__main__":
assert (
not torch.cuda._initialized
), "test_distributed must not have initialized CUDA context on main process"
run_tests()