TensorRT-LLMs/tests/functional/test_matmul.py
2024-07-17 20:45:02 +08:00

142 lines
4.9 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
# isort: off
import torch
# isort: on
import os
import sys
from parameterized import parameterized
import tensorrt_llm
from tensorrt_llm import Tensor
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 TestMatmul(unittest.TestCase):
def setUp(self):
torch.backends.cudnn.allow_tf32 = False
tensorrt_llm.logger.set_level('error')
def _matmul(self, m, n, k, dtype, ta, tb):
shape1 = (k, m) if ta else (m, k)
mat1 = torch.randn(shape1,
dtype=tensorrt_llm._utils.str_dtype_to_torch(dtype),
device="cuda") * 1e-1
shape2 = (n, k) if tb else (k, n)
mat2 = torch.randn(shape2,
dtype=tensorrt_llm._utils.str_dtype_to_torch(dtype),
device="cuda") * 1e-1
# construct trt network
builder = tensorrt_llm.Builder()
network = builder.create_network()
with tensorrt_llm.net_guard(network):
x = Tensor(name='x',
shape=mat1.shape,
dtype=tensorrt_llm.str_dtype_to_trt(dtype))
y = Tensor(name='y',
shape=mat2.shape,
dtype=tensorrt_llm.str_dtype_to_trt(dtype))
output = tensorrt_llm.functional.matmul(x, y, transa=ta, transb=tb)
output.mark_output('output', dtype)
# trt run
session = create_session(builder, network, precision=dtype)
inputs = {'x': mat1, 'y': mat2}
outputs = run_session(session, inputs)
tols = {
"float32": {
"rtol": 4e-4,
"atol": 1e-02
},
"float16": {
"rtol": 1e-02,
"atol": 1e-02
},
"bfloat16": {
"rtol": 1e-02,
"atol": 1e-02
},
}
# pytorch run
if ta:
mat1 = mat1.transpose(0, 1)
if tb:
mat2 = mat2.transpose(0, 1)
ref = torch.matmul(mat1, mat2)
torch.testing.assert_close(ref, outputs['output'], **tols[dtype])
@parameterized.expand([('float16', False, False), ('float16', False, True),
('float16', True, False), ('float16', True, True),
('bfloat16', True, False), ('bfloat16', True, True),
('float32', False, False), ('float32', False, True),
('float32', True, False), ('float32', True, True)],
name_func=unittest_name_func)
def test_matmul(self, dtype, transa, transb):
# Skip tests that are not supported in pre-ampere architecture
skip_bf16_pre_ampere(dtype)
bs = 2
inseq = 16
hidden_size = 768
tp = 1
# qkv_gemm
self._matmul(bs * inseq, 3 * hidden_size // tp, hidden_size, dtype,
transa, transb)
# mlp_gemm_1
self._matmul(bs * inseq, 4 * hidden_size // tp, hidden_size, dtype,
transa, transb)
def test_matmul_broadcast(self):
dtype = 'float32'
x_data = torch.randn(16, 4, 4, 5, device="cuda")
y_data = torch.randn(16, 1, 5, 4, device="cuda")
# construct trt network
builder = tensorrt_llm.Builder()
network = builder.create_network()
with tensorrt_llm.net_guard(network):
x = Tensor(name='x',
shape=x_data.shape,
dtype=tensorrt_llm.str_dtype_to_trt(dtype))
y = Tensor(name='y',
shape=y_data.shape,
dtype=tensorrt_llm.str_dtype_to_trt(dtype))
output = tensorrt_llm.functional.matmul(x, y)
output.mark_output('output', dtype)
# trt run
session = create_session(builder, network, precision=dtype)
inputs = {'x': x_data, 'y': y_data}
outputs = run_session(session, inputs)
# pytorch run
ref = torch.matmul(x_data, y_data)
# compare diff
torch.testing.assert_close(ref, outputs['output'])