mirror of
https://github.com/NVIDIA/TensorRT-LLM.git
synced 2026-01-14 06:27:45 +08:00
0fd59d64ab
* add fmha repo Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com> * fix format Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com> * fix code style Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com> * fix header Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com> * fix header kernel_traits.h Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com> * add .gitignore file Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com> * add SLIDING_WINDOW_ATTENTION Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com> * fix style Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com> * fix format Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com> * update setup.py Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com> * update build_wheel.py Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com> --------- Signed-off-by: Qidi Sang <200703406+qsang-nv@users.noreply.github.com> Signed-off-by: qsang-nv <200703406+qsang-nv@users.noreply.github.com>
76 lines
2.5 KiB
Python
76 lines
2.5 KiB
Python
# SPDX-FileCopyrightText: Copyright (c) 2021-2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
|
|
# SPDX-License-Identifier: NVIDIA TensorRT Source Code License Agreement
|
|
#
|
|
# NVIDIA CORPORATION, its affiliates and licensors retain all intellectual
|
|
# property and proprietary rights in and to this material, related
|
|
# documentation and any modifications thereto. Any use, reproduction,
|
|
# disclosure or distribution of this material and related documentation
|
|
# without an express license agreement from NVIDIA CORPORATION or
|
|
# its affiliates is strictly prohibited.
|
|
#
|
|
# test of tests
|
|
|
|
import pytest
|
|
from filter_rules import *
|
|
from utils import *
|
|
|
|
fmha_exe_path = 'bin/fmha.exe'
|
|
|
|
|
|
# debug
|
|
@pytest.mark.debug
|
|
@pytest.mark.parametrize('rules', [[]])
|
|
@pytest.mark.parametrize('fmha_arg', [FmhaArgs()])
|
|
def test_fmha_single_case(fmha_arg, rules, dryrun, disable_rules,
|
|
gpu_compute_cap):
|
|
fmha_harness(fmha_exe_path,
|
|
fmha_arg,
|
|
rules,
|
|
dryrun,
|
|
gpu_compute_cap=gpu_compute_cap)
|
|
|
|
|
|
# debug
|
|
@pytest.mark.debug
|
|
@apply_rule(corner_case_rule)
|
|
@pytest.mark.parametrize('rules', [[]])
|
|
@pytest.mark.parametrize('fmha_arg',
|
|
[FmhaArgs()._replace(seq_len=16, head_dim=64)])
|
|
def test_corner_case_rule(fmha_arg, rules, dryrun, disable_rules,
|
|
gpu_compute_cap):
|
|
fmha_harness(fmha_exe_path,
|
|
fmha_arg,
|
|
rules,
|
|
dryrun,
|
|
gpu_compute_cap=gpu_compute_cap)
|
|
|
|
|
|
# debug
|
|
@pytest.mark.debug
|
|
@apply_rule(compute_cap_specific_rule)
|
|
@pytest.mark.parametrize('rules', [[]])
|
|
@pytest.mark.parametrize('fmha_arg', [FmhaArgs()])
|
|
def test_compute_cap_specific_rule(fmha_arg, rules, dryrun, disable_rules,
|
|
mock_gpu_compute_cap):
|
|
fmha_harness(fmha_exe_path,
|
|
fmha_arg,
|
|
rules,
|
|
dryrun,
|
|
gpu_compute_cap=mock_gpu_compute_cap)
|
|
|
|
|
|
# debug; try running large seq and large head size without error tolerance
|
|
@pytest.mark.debug
|
|
@apply_rule(sm80_only)
|
|
@pytest.mark.parametrize('rules', [[]])
|
|
@pytest.mark.parametrize('fmha_arg',
|
|
[FmhaArgs()._replace(seq_len=4096, head_dim=256)])
|
|
@pytest.mark.xfail
|
|
def test_controlled_validation_fail(fmha_arg, rules, dryrun, disable_rules,
|
|
gpu_compute_cap):
|
|
fmha_harness(fmha_exe_path,
|
|
fmha_arg,
|
|
rules,
|
|
dryrun,
|
|
gpu_compute_cap=gpu_compute_cap)
|