mirror of
https://github.com/NVIDIA/Model-Optimizer.git
synced 2026-11-10 09:34:01 +08:00
195 lines
7.6 KiB
Python
195 lines
7.6 KiB
Python
# SPDX-FileCopyrightText: Copyright (c) 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.
|
|
|
|
"""Tests of QuantPooling module."""
|
|
|
|
import pytest
|
|
import torch
|
|
import torch.nn.functional as F
|
|
|
|
from modelopt.torch.quantization import tensor_quant
|
|
from modelopt.torch.quantization.config import QuantizerAttributeConfig
|
|
from modelopt.torch.quantization.nn.modules import quant_pooling
|
|
|
|
|
|
class TestQuantMaxPool:
|
|
@pytest.mark.parametrize(
|
|
("pool_cls", "f_pool", "input_shape"),
|
|
[
|
|
(quant_pooling.QuantMaxPool1d, F.max_pool1d, (2, 5, 5)),
|
|
(quant_pooling.QuantMaxPool2d, F.max_pool2d, (2, 5, 5, 5)),
|
|
(quant_pooling.QuantMaxPool3d, F.max_pool3d, (2, 5, 5, 5, 5)),
|
|
],
|
|
)
|
|
def test_input_fake_quant(self, pool_cls, f_pool, input_shape):
|
|
quant_pooling_object = pool_cls(kernel_size=3, stride=1)
|
|
|
|
test_input = torch.randn(input_shape)
|
|
|
|
quant_input = tensor_quant.fake_tensor_quant(
|
|
test_input, torch.max(torch.abs(test_input)), None
|
|
)
|
|
|
|
out1 = f_pool(quant_input, 3, 1, 0, 1, False, False)
|
|
out2 = quant_pooling_object(test_input)
|
|
assert torch.allclose(out1, out2, rtol=0, atol=0)
|
|
|
|
def test_input_variable_bits(self):
|
|
# Repeat checking the output for variable number of bits to QuantizerAttributeConfig
|
|
for bits in [2, 4, 6]:
|
|
quant_desc_input = QuantizerAttributeConfig(num_bits=bits)
|
|
|
|
quant_pooling.QuantMaxPool2d.default_quant_desc_input = quant_desc_input
|
|
quant_pooling_object = quant_pooling.QuantMaxPool2d(kernel_size=3, stride=1)
|
|
|
|
test_input = torch.randn(1, 5, 5, 5)
|
|
|
|
quant_input = tensor_quant.fake_tensor_quant(
|
|
test_input, torch.max(torch.abs(test_input)), None, bits
|
|
)
|
|
|
|
out1 = F.max_pool2d(quant_input, 3, 1, 0, 1, False, False)
|
|
out2 = quant_pooling_object(test_input)
|
|
assert torch.allclose(out1, out2, rtol=0, atol=0)
|
|
|
|
def test_input_fake_quant_disable(self):
|
|
quant_pooling_object = quant_pooling.QuantMaxPool2d(kernel_size=3, stride=1)
|
|
|
|
test_input = torch.randn(1, 5, 5, 5, dtype=torch.double)
|
|
|
|
quant_pooling_object.input_quantizer.disable()
|
|
|
|
out1 = F.max_pool2d(test_input, 3, 1, 0, 1, False, False)
|
|
out2 = quant_pooling_object(test_input)
|
|
assert torch.allclose(out1, out2, rtol=0, atol=0)
|
|
|
|
def test_input_multi_axis(self):
|
|
quant_desc_input = QuantizerAttributeConfig(num_bits=8, axis=(0, 1))
|
|
|
|
quant_pooling.QuantMaxPool2d.default_quant_desc_input = quant_desc_input
|
|
quant_pooling_object = quant_pooling.QuantMaxPool2d(kernel_size=3, stride=1)
|
|
|
|
test_input = torch.randn(2, 7, 5, 5, dtype=torch.double)
|
|
input_amax = torch.amax(torch.abs(test_input), dim=(2, 3), keepdim=True)
|
|
quant_input = tensor_quant.fake_tensor_quant(test_input, input_amax, None)
|
|
|
|
out1 = F.max_pool2d(quant_input, 3, 1, 0, 1, False, False)
|
|
out2 = quant_pooling_object(test_input)
|
|
assert torch.allclose(out1, out2, rtol=0, atol=0)
|
|
|
|
|
|
class TestQuantAvgPoolNd:
|
|
@pytest.mark.parametrize(
|
|
("pool_cls", "f_pool", "input_shape"),
|
|
[
|
|
(quant_pooling.QuantAvgPool1d, F.avg_pool1d, (2, 5, 5)),
|
|
(quant_pooling.QuantAvgPool2d, F.avg_pool2d, (2, 5, 5, 5)),
|
|
(quant_pooling.QuantAvgPool3d, F.avg_pool3d, (2, 5, 5, 5, 5)),
|
|
],
|
|
)
|
|
def test_input_fake_quant(self, pool_cls, f_pool, input_shape):
|
|
quant_pooling_object = pool_cls(kernel_size=3, stride=1)
|
|
|
|
test_input = torch.randn(input_shape)
|
|
|
|
quant_input = tensor_quant.fake_tensor_quant(
|
|
test_input, torch.max(torch.abs(test_input)), None
|
|
)
|
|
|
|
out1 = f_pool(quant_input, 3, 1, 0, False, True)
|
|
out2 = quant_pooling_object(test_input)
|
|
assert torch.allclose(out1, out2, rtol=0, atol=0)
|
|
|
|
@pytest.mark.parametrize("bits", [2, 4, 6])
|
|
def test_input_variable_bits(self, bits):
|
|
# Repeat checking the output for variable number of bits to QuantizerAttributeConfig
|
|
quant_desc_input = QuantizerAttributeConfig(num_bits=bits)
|
|
|
|
quant_pooling.QuantAvgPool2d.default_quant_desc_input = quant_desc_input
|
|
quant_pooling_object = quant_pooling.QuantAvgPool2d(kernel_size=3, stride=1)
|
|
|
|
test_input = torch.randn(1, 5, 5, 5, dtype=torch.double)
|
|
|
|
quant_input = tensor_quant.fake_tensor_quant(
|
|
test_input, torch.max(torch.abs(test_input)), None, bits
|
|
)
|
|
|
|
out1 = F.avg_pool2d(quant_input, 3, 1, 0, False, True, None)
|
|
out2 = quant_pooling_object(test_input)
|
|
assert torch.allclose(out1, out2, rtol=0, atol=0)
|
|
|
|
def test_input_fake_quant_disable(self):
|
|
quant_pooling_object = quant_pooling.QuantAvgPool2d(kernel_size=3, stride=1)
|
|
|
|
test_input = torch.randn(1, 5, 5, 5, dtype=torch.double)
|
|
|
|
quant_pooling_object.input_quantizer.disable()
|
|
|
|
out1 = F.avg_pool2d(test_input, 3, 1, 0, False, True, None)
|
|
out2 = quant_pooling_object(test_input)
|
|
assert torch.allclose(out1, out2, rtol=0, atol=0)
|
|
|
|
|
|
class TestQuantAdaptiveAvgPoolNd:
|
|
@pytest.mark.parametrize(
|
|
("pool_cls", "f_pool", "input_shape"),
|
|
[
|
|
(quant_pooling.QuantAdaptiveAvgPool1d, F.adaptive_avg_pool1d, (2, 5, 5)),
|
|
(quant_pooling.QuantAdaptiveAvgPool2d, F.adaptive_avg_pool2d, (2, 5, 5, 5)),
|
|
(quant_pooling.QuantAdaptiveAvgPool3d, F.adaptive_avg_pool3d, (2, 5, 5, 5, 5)),
|
|
],
|
|
)
|
|
def test_input_fake_quant(self, pool_cls, f_pool, input_shape):
|
|
quant_pooling_object = pool_cls(output_size=3)
|
|
|
|
test_input = torch.randn(input_shape)
|
|
|
|
quant_input = tensor_quant.fake_tensor_quant(
|
|
test_input, torch.max(torch.abs(test_input)), None
|
|
)
|
|
|
|
out1 = f_pool(quant_input, 3)
|
|
out2 = quant_pooling_object(test_input)
|
|
assert torch.allclose(out1, out2, rtol=0, atol=0)
|
|
|
|
def test_input_variable_bits(self):
|
|
# Repeat checking the output for variable number of bits to QuantizerAttributeConfig
|
|
for bits in [2, 4, 6]:
|
|
quant_desc_input = QuantizerAttributeConfig(num_bits=bits)
|
|
|
|
quant_pooling.QuantAdaptiveAvgPool2d.default_quant_desc_input = quant_desc_input
|
|
quant_pooling_object = quant_pooling.QuantAdaptiveAvgPool2d(output_size=3)
|
|
|
|
test_input = torch.randn(1, 5, 5, 5, dtype=torch.double)
|
|
|
|
quant_input = tensor_quant.fake_tensor_quant(
|
|
test_input, torch.max(torch.abs(test_input)), None, bits
|
|
)
|
|
|
|
out1 = F.adaptive_avg_pool2d(quant_input, 3)
|
|
out2 = quant_pooling_object(test_input)
|
|
assert torch.allclose(out1, out2, rtol=0, atol=0)
|
|
|
|
def test_input_fake_quant_disable(self):
|
|
quant_pooling_object = quant_pooling.QuantAdaptiveAvgPool2d(output_size=3)
|
|
|
|
test_input = torch.randn(1, 5, 5, 5, dtype=torch.double)
|
|
|
|
quant_pooling_object.input_quantizer.disable()
|
|
|
|
out1 = F.adaptive_avg_pool2d(test_input, 3)
|
|
out2 = quant_pooling_object(test_input)
|
|
assert torch.allclose(out1, out2, rtol=0, atol=0)
|