TensorRT-LLMs/tests/functional/test_nccl.py
2024-12-24 15:58:43 +08:00

126 lines
4.6 KiB
Python

# SPDX-FileCopyrightText: Copyright (c) 2022-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: Apache-2.0
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import unittest
from itertools import product
import pytest
# isort: off
import torch
# isort: on
import os
import sys
from cuda import cudart
from parameterized import parameterized
import tensorrt_llm
from tensorrt_llm import Mapping, Tensor
from tensorrt_llm.functional import (AllReduceConfig, AllReduceParams,
AllReduceStrategy, allreduce)
from tensorrt_llm.plugin.plugin import current_all_reduce_helper
sys.path.append(os.path.join(os.path.dirname(__file__), '..'))
from utils.util import (create_session, run_session, skip_bf16_pre_ampere,
unittest_name_func)
class TestCommunicationPlugin(unittest.TestCase):
def setUp(self):
tensorrt_llm.logger.set_level('error')
self.world_size = tensorrt_llm.mpi_world_size()
self.rank = tensorrt_llm.mpi_rank()
torch.cuda.set_device(self.rank)
cudart.cudaSetDevice(self.rank)
self.reference_tensors = [
torch.full([10000000], i + 1, dtype=torch.float32, device="cuda")
for i in range(self.world_size)
]
self.mapping = Mapping(self.world_size,
self.rank,
self.world_size,
tp_size=self.world_size)
@parameterized.expand(list(
product(["bfloat16", 'float16', "float32"], [
AllReduceStrategy.NCCL, AllReduceStrategy.ONESHOT,
AllReduceStrategy.TWOSHOT
], [
AllReduceConfig(0),
AllReduceConfig.PUSH_MODE,
AllReduceConfig.USE_MEMCPY,
], [64 * 70000, 64 * 70, 64])),
name_func=unittest_name_func)
def test_allreduce(self, dtype: str, strategy: AllReduceStrategy,
config: AllReduceConfig, size: int):
skip_bf16_pre_ampere(dtype)
if self.world_size == 1:
pytest.skip("Skip single GPU NCCL")
if strategy == AllReduceStrategy.NCCL and config != AllReduceConfig(0):
pytest.skip("NCCL with specific config discarded")
workspace = None
torch_dtype = tensorrt_llm._utils.str_dtype_to_torch(dtype)
dtype_size = torch.finfo(torch_dtype).bits // 8
allreduce_ref = torch.zeros(self.reference_tensors[0][:size].shape,
dtype=torch_dtype,
device="cuda")
for i in range(self.world_size):
allreduce_ref = allreduce_ref + self.reference_tensors[i][:size].to(
torch_dtype)
# construct trt network
builder = tensorrt_llm.Builder()
network = builder.create_network()
network.plugin_config.set_nccl_plugin(dtype)
_, workspace = current_all_reduce_helper().allocate_workspace(
self.mapping, size * dtype_size)
input = self.reference_tensors[self.rank][:size].to(torch_dtype)
inner_loop = 5
with tensorrt_llm.net_guard(network):
x = Tensor(name='x',
shape=input.shape,
dtype=tensorrt_llm.str_dtype_to_trt(dtype))
current_all_reduce_helper().set_workspace_tensor(self.mapping)
current = x
for i in range(inner_loop):
current = allreduce(current,
self.mapping.tp_group,
all_reduce_params=AllReduceParams(
strategy=strategy, config=config))
current.mark_output('output', dtype)
# trt run
session = create_session(builder, network, precision=dtype)
inputs = {'x': input, 'all_reduce_workspace': workspace}
outputs = run_session(session, inputs)
# compare diff
torch.testing.assert_close(outputs['output'],
(self.mapping.tp_size**(inner_loop - 1)) *
allreduce_ref)