mirror of
https://github.com/NVIDIA/TensorRT-LLM.git
synced 2026-01-14 06:27:45 +08:00
92 lines
3.2 KiB
Python
92 lines
3.2 KiB
Python
# SPDX-FileCopyrightText: Copyright (c) 2022-2023 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
|
|
|
|
import numpy as np
|
|
import torch
|
|
from polygraphy.backend.trt import EngineFromNetwork, TrtRunner
|
|
|
|
import tensorrt_llm
|
|
|
|
|
|
class TestFunctional(unittest.TestCase):
|
|
|
|
def setUp(self):
|
|
tensorrt_llm.logger.set_level('error')
|
|
|
|
def test_arange_int(self):
|
|
# test data
|
|
start = 0
|
|
end = 128
|
|
dtype = 'int32'
|
|
|
|
# construct trt network
|
|
builder = tensorrt_llm.Builder()
|
|
net = builder.create_network()
|
|
with tensorrt_llm.net_guard(net):
|
|
network = tensorrt_llm.default_trtnet()
|
|
|
|
output = tensorrt_llm.functional.arange(start=start,
|
|
end=end,
|
|
dtype=dtype).trt_tensor
|
|
output.name = 'output'
|
|
network.mark_output(output)
|
|
output.dtype = tensorrt_llm.str_dtype_to_trt(dtype)
|
|
|
|
# trt run
|
|
build_engine = EngineFromNetwork((builder.trt_builder, net.trt_network))
|
|
with TrtRunner(build_engine) as runner:
|
|
outputs = runner.infer(feed_dict={})
|
|
|
|
ref = torch.arange(start, end).int()
|
|
np.testing.assert_allclose(ref.cpu().numpy(),
|
|
outputs['output'],
|
|
atol=1e-5)
|
|
|
|
def test_arange_tensor(self):
|
|
# test data
|
|
s = 0
|
|
e = 128
|
|
dtype = 'int32'
|
|
|
|
# construct trt network
|
|
builder = tensorrt_llm.Builder()
|
|
net = builder.create_network()
|
|
with tensorrt_llm.net_guard(net):
|
|
network = tensorrt_llm.default_trtnet()
|
|
|
|
start = tensorrt_llm.functional.constant(np.array(s,
|
|
dtype=np.int32))
|
|
end_tensor = tensorrt_llm.functional.constant(
|
|
np.array([0] * e, dtype=np.int32))
|
|
|
|
output = tensorrt_llm.functional.arange(
|
|
start=start,
|
|
end=tensorrt_llm.functional.shape(end_tensor, 0),
|
|
dtype=dtype).trt_tensor
|
|
output.name = 'output'
|
|
network.mark_output(output)
|
|
output.dtype = tensorrt_llm.str_dtype_to_trt(dtype)
|
|
|
|
# trt run
|
|
build_engine = EngineFromNetwork((builder.trt_builder, net.trt_network))
|
|
with TrtRunner(build_engine) as runner:
|
|
outputs = runner.infer(feed_dict={})
|
|
|
|
ref = torch.arange(s, e).int()
|
|
np.testing.assert_allclose(ref.cpu().numpy(),
|
|
outputs['output'],
|
|
atol=1e-5)
|