test_convert.py

"""
Reference: We use tested models from https://github.com/pytorch/pytorch/blob/master/test/jit/test_models.py.
"""

import os
import sys
import unittest

import numpy as np
import torch
import torch.nn.functional as F
import torchvision

import nni.retiarii.nn.pytorch as nn
from nni.retiarii import basic_unit
from nni.retiarii.converter import convert_to_graph
from nni.retiarii.codegen import model_to_pytorch_script

class MnistNet(nn.Module):
    def __init__(self):
        super(MnistNet, self).__init__()
        self.conv1 = nn.Conv2d(1, 10, kernel_size=5)
        self.conv2 = nn.Conv2d(10, 20, kernel_size=5)
        self.conv2_drop = nn.Dropout2d()
        self.fc1 = nn.Linear(320, 50)
        self.fc2 = nn.Linear(50, 10)

    def forward(self, x):
        x = F.relu(F.max_pool2d(self.conv1(x), 2))
        x = F.relu(F.max_pool2d(self.conv2_drop(self.conv2(x)), 2))
        x = x.view(-1, 320)
        x = F.relu(self.fc1(x))
        x = F.dropout(x, training=self.training)
        x = self.fc2(x)
        return F.log_softmax(x, dim=1)

# NOTE: serialize module cannot be placed within class or function
@basic_unit
class Linear(nn.Module):
    def __init__(self, d_embed, d_proj):
        super().__init__()
        self.linear = nn.Linear(d_embed, d_proj)

    def forward(self, input):
        if len(input.size()) <= 2:
            return self.linear(input)
        size = input.size()[:2]
        out = self.linear(input.view(size[0] * size[1], -1))
        return out.view(size[0], size[1], -1)

class TestConvert(unittest.TestCase):
    @staticmethod
    def _match_state_dict(current_values, expected_format):
        result = {}
        for k, v in expected_format.items():
            for idx, cv in enumerate(current_values):
                if cv.shape == v.shape:
                    result[k] = cv
                    current_values.pop(idx)
                    break
        return result

    def checkExportImport(self, model, input):
        script_module = torch.jit.script(model)
        model_ir = convert_to_graph(script_module, model)
        model_code = model_to_pytorch_script(model_ir)

        exec_vars = {}
        exec(model_code + '\n\nconverted_model = _model()', exec_vars)
        converted_model = exec_vars['converted_model']
        converted_state_dict = self._match_state_dict(list(model.state_dict().values()),
                                                      dict(converted_model.state_dict()))
        converted_model.load_state_dict(converted_state_dict)
        with torch.no_grad():
            expected_output = model.eval()(*input)
            converted_output = converted_model.eval()(*input)
        self.assertEqual(len(converted_output), len(expected_output))
        for a, b in zip(converted_output, expected_output):
            self.assertLess((a - b).abs().max().item(), 1E-4)
        return converted_model

    def setUp(self):
        # FIXME
        import nni.retiarii.debug_configs
        nni.retiarii.debug_configs.framework = 'pytorch'

    def test_dcgan_models(self):
        class DCGANGenerator(nn.Module):
            def __init__(self, nz, ngf, nc):
                super(DCGANGenerator, self).__init__()
                self.main = nn.Sequential(
                    # input is Z, going into a convolution
                    nn.ConvTranspose2d(nz, ngf * 8, 4, 1, 0, bias=False),
                    nn.BatchNorm2d(ngf * 8),
                    nn.ReLU(True),
                    # state size. (ngf*8) x 4 x 4
                    nn.ConvTranspose2d(ngf * 8, ngf * 4, 4, 2, 1, bias=False),
                    nn.BatchNorm2d(ngf * 4),
                    nn.ReLU(True),
                    # state size. (ngf*4) x 8 x 8
                    nn.ConvTranspose2d(ngf * 4, ngf * 2, 4, 2, 1, bias=False),
                    nn.BatchNorm2d(ngf * 2),
                    nn.ReLU(True),
                    # state size. (ngf*2) x 16 x 16
                    nn.ConvTranspose2d(ngf * 2, ngf, 4, 2, 1, bias=False),
                    nn.BatchNorm2d(ngf),
                    nn.ReLU(True),
                    # state size. (ngf) x 32 x 32
                    nn.ConvTranspose2d(ngf, nc, 4, 2, 1, bias=False),
                    nn.Tanh()
                    # state size. (nc) x 64 x 64
                )

            def forward(self, input):
                return self.main(input)

        class DCGANDiscriminator(nn.Module):
            def __init__(self, nc, ndf):
                super(DCGANDiscriminator, self).__init__()
                self.main = nn.Sequential(
                    # input is (nc) x 64 x 64
                    nn.Conv2d(nc, ndf, 4, 2, 1, bias=False),
                    nn.LeakyReLU(0.2, inplace=True),
                    # state size. (ndf) x 32 x 32
                    nn.Conv2d(ndf, ndf * 2, 4, 2, 1, bias=False),
                    nn.BatchNorm2d(ndf * 2),
                    nn.LeakyReLU(0.2, inplace=True),
                    # state size. (ndf*2) x 16 x 16
                    nn.Conv2d(ndf * 2, ndf * 4, 4, 2, 1, bias=False),
                    nn.BatchNorm2d(ndf * 4),
                    nn.LeakyReLU(0.2, inplace=True),
                    # state size. (ndf*4) x 8 x 8
                    nn.Conv2d(ndf * 4, ndf * 8, 4, 2, 1, bias=False),
                    nn.BatchNorm2d(ndf * 8),
                    nn.LeakyReLU(0.2, inplace=True),
                    # state size. (ndf*8) x 4 x 4
                    nn.Conv2d(ndf * 8, 1, 4, 1, 0, bias=False),
                    nn.Sigmoid()
                )

            def forward(self, input):
                return self.main(input).view(-1, 1).squeeze(1)

        bs, nz, ngf, nc, ndf = 5, 6, 9, 3, 10
        input = (torch.rand(bs, nz, 1, 1),)
        model = DCGANGenerator(nz, ngf, nc)
        self.checkExportImport(model, input)

    def test_neural_style(self):
        class TransformerNet(nn.Module):
            def __init__(self):
                super(TransformerNet, self).__init__()
                # Initial convolution layers
                self.conv1 = ConvLayer(3, 32, kernel_size=9, stride=1)
                self.in1 = nn.InstanceNorm2d(32, affine=True)
                self.conv2 = ConvLayer(32, 64, kernel_size=3, stride=2)
                self.in2 = nn.InstanceNorm2d(64, affine=True)
                self.conv3 = ConvLayer(64, 128, kernel_size=3, stride=2)
                self.in3 = nn.InstanceNorm2d(128, affine=True)
                # Residual layers
                self.res1 = ResidualBlock(128)
                self.res2 = ResidualBlock(128)
                self.res3 = ResidualBlock(128)
                self.res4 = ResidualBlock(128)
                self.res5 = ResidualBlock(128)
                # Upsampling Layers
                self.deconv1 = UpsampleConvLayer(128, 64, kernel_size=3, stride=1, upsample=2)
                self.in4 = nn.InstanceNorm2d(64, affine=True)
                self.deconv2 = UpsampleConvLayer(64, 32, kernel_size=3, stride=1, upsample=2)
                self.in5 = nn.InstanceNorm2d(32, affine=True)
                self.deconv3 = ConvLayer(32, 3, kernel_size=9, stride=1)
                # Non-linearities
                self.relu = nn.ReLU()

            def forward(self, X):
                y = self.relu(self.in1(self.conv1(X)))
                y = self.relu(self.in2(self.conv2(y)))
                y = self.relu(self.in3(self.conv3(y)))
                y = self.res1(y)
                y = self.res2(y)
                y = self.res3(y)
                y = self.res4(y)
                y = self.res5(y)
                y = self.relu(self.in4(self.deconv1(y)))
                y = self.relu(self.in5(self.deconv2(y)))
                y = self.deconv3(y)
                return y

        class ConvLayer(nn.Module):
            def __init__(self, in_channels, out_channels, kernel_size, stride):
                super(ConvLayer, self).__init__()
                reflection_padding = kernel_size // 2
                self.reflection_pad = nn.ReflectionPad2d(reflection_padding)
                self.conv2d = nn.Conv2d(in_channels, out_channels, kernel_size, stride)

            def forward(self, x):
                out = self.reflection_pad(x)
                out = self.conv2d(out)
                return out

        class ResidualBlock(nn.Module):
            """ResidualBlock
            introduced in: https://arxiv.org/abs/1512.03385
            recommended architecture: http://torch.ch/blog/2016/02/04/resnets.html
            """

            def __init__(self, channels):
                super(ResidualBlock, self).__init__()
                self.conv1 = ConvLayer(channels, channels, kernel_size=3, stride=1)
                self.in1 = nn.InstanceNorm2d(channels, affine=True)
                self.conv2 = ConvLayer(channels, channels, kernel_size=3, stride=1)
                self.in2 = nn.InstanceNorm2d(channels, affine=True)
                self.relu = nn.ReLU()

            def forward(self, x):
                residual = x
                out = self.relu(self.in1(self.conv1(x)))
                out = self.in2(self.conv2(out))
                out = out + residual
                return out

        class UpsampleConvLayer(nn.Module):
            """UpsampleConvLayer
            Upsamples the input and then does a convolution. This method gives better results
            compared to ConvTranspose2d.
            ref: http://distill.pub/2016/deconv-checkerboard/
            """

            def __init__(self, in_channels, out_channels, kernel_size, stride, upsample=None):
                super(UpsampleConvLayer, self).__init__()
                self.upsample = upsample
                if upsample:
                    self.upsample_layer = nn.Upsample(mode='nearest', scale_factor=upsample)
                reflection_padding = kernel_size // 2
                self.reflection_pad = nn.ReflectionPad2d(reflection_padding)
                self.conv2d = nn.Conv2d(in_channels, out_channels, kernel_size, stride)

            def forward(self, x):
                x_in = x
                if self.upsample:
                    x_in = self.upsample_layer(x_in)
                out = self.reflection_pad(x_in)
                out = self.conv2d(out)
                return out

        model = TransformerNet()
        input = (torch.rand(5, 3, 16, 16),)
        self.checkExportImport(model, input)

    def test_mnist(self):
        # eval() is present because dropout makes this nondeterministic
        self.checkExportImport(MnistNet().eval(), (torch.rand(5, 1, 28, 28),))

    def test_reinforcement_learning(self):
        class Policy(nn.Module):
            def __init__(self):
                super(Policy, self).__init__()
                self.affine1 = nn.Linear(4, 128)
                self.affine2 = nn.Linear(128, 2)

            def forward(self, x):
                x = F.relu(self.affine1(x))
                action_scores = self.affine2(x)
                return F.softmax(action_scores, dim=1)

        self.checkExportImport(Policy(), (torch.rand(1, 4),))

    def test_snli(self):

        class Encoder(nn.Module):

            def __init__(self, config):
                super(Encoder, self).__init__()
                #self.config = config
                input_size = config["d_proj"] if config["projection"] else config["d_embed"]
                dropout = 0 if config["n_layers"] == 1 else config["dp_ratio"]
                self.rnn = nn.LSTM(input_size=input_size, hidden_size=config["d_hidden"],
                                   num_layers=config["n_layers"], dropout=dropout,
                                   bidirectional=config["birnn"])
                self.n_cells = config["n_cells"]
                self.d_hidden = config["d_hidden"]
                self.birnn = config["birnn"]

            def forward(self, inputs):
                batch_size = inputs.size()[1]
                state_shape = self.n_cells, batch_size, self.d_hidden
                h0 = c0 = inputs.new_zeros(state_shape)
                outputs, (ht, ct) = self.rnn(inputs, (h0, c0))
                return ht[-1] if not self.birnn else ht[-2:].transpose(0, 1).contiguous().view(batch_size, -1)

        class SNLIClassifier(nn.Module):

            def __init__(self, config):
                super(SNLIClassifier, self).__init__()
                self.embed = nn.Embedding(config["n_embed"], config["d_embed"])
                self.projection = Linear(config["d_embed"], config["d_proj"])
                self.encoder = Encoder(config)
                self.dropout = nn.Dropout(p=config["dp_ratio"])
                self.relu = nn.ReLU()
                seq_in_size = 2 * config["d_hidden"]
                if config["birnn"]:
                    seq_in_size *= 2
                lin_config = [seq_in_size] * 2
                self.out = nn.Sequential(
                    Linear(*lin_config),
                    self.relu,
                    self.dropout,
                    Linear(*lin_config),
                    self.relu,
                    self.dropout,
                    Linear(*lin_config),
                    self.relu,
                    self.dropout,
                    Linear(seq_in_size, config["d_out"]))
                self.fix_emb = config["fix_emb"]
                self.project = config["projection"]

            def forward(self, premise, hypothesis):
                prem_embed = self.embed(premise)
                hypo_embed = self.embed(hypothesis)
                if self.fix_emb:
                    prem_embed = prem_embed.detach()
                    hypo_embed = hypo_embed.detach()
                if self.project:
                    prem_embed = self.relu(self.projection(prem_embed))
                    hypo_embed = self.relu(self.projection(hypo_embed))
                premise = self.encoder(prem_embed)
                hypothesis = self.encoder(hypo_embed)
                scores = self.out(torch.cat([premise, hypothesis], 1))
                return scores

        Config = {
            "n_embed": 100,
            "d_embed": 100,
            "d_proj": 300,
            "dp_ratio": 0.0,  # For deterministic testing TOD": change by fixing seed in checkTrace?,
            "d_hidden": 30,
            "birnn": True,
            "d_out": 300,
            "fix_emb": True,
            "projection": True,
            "n_layers": 2,
            "n_cells": 4  # 2 * n_layers because birnn = True,
        }

        premise = torch.LongTensor(48, 64).random_(0, 100)
        hypothesis = torch.LongTensor(24, 64).random_(0, 100)

        self.checkExportImport(SNLIClassifier(Config), (premise, hypothesis))

    def test_super_resolution(self):
        class Net(nn.Module):

            def __init__(self, upscale_factor):
                super(Net, self).__init__()

                self.relu = nn.ReLU()
                self.conv1 = nn.Conv2d(1, 64, (5, 5), (1, 1), (2, 2))
                self.conv2 = nn.Conv2d(64, 64, (3, 3), (1, 1), (1, 1))
                self.conv3 = nn.Conv2d(64, 32, (3, 3), (1, 1), (1, 1))
                self.conv4 = nn.Conv2d(32, upscale_factor ** 2, (3, 3), (1, 1), (1, 1))
                self.pixel_shuffle = nn.PixelShuffle(upscale_factor)

            def forward(self, x):
                x = self.relu(self.conv1(x))
                x = self.relu(self.conv2(x))
                x = self.relu(self.conv3(x))
                x = self.pixel_shuffle(self.conv4(x))
                return x

        net = Net(upscale_factor=4)
        self.checkExportImport(net, (torch.rand(5, 1, 32, 32),))

    @unittest.skip('Need to support Loop')  # FIXME
    def test_time_sequence_prediction(self):
        class Sequence(nn.Module): #torch.jit.ScriptModule
            def __init__(self):
                super(Sequence, self).__init__()
                self.lstm1 = nn.LSTMCell(1, 51)
                self.lstm2 = nn.LSTMCell(51, 51)
                self.linear = nn.Linear(51, 1)

            #@torch.jit.script_method
            def forward(self, input):
                # TODO: add future as input with default val
                # see https://github.com/pytorch/pytorch/issues/8724
                outputs = torch.empty((3, 0))
                h_t = torch.zeros((3, 51))
                c_t = torch.zeros((3, 51))
                h_t2 = torch.zeros((3, 51))
                c_t2 = torch.zeros((3, 51))

                output = torch.zeros([3, 51])
                future = 2

                # TODO: chunk call should appear as the for loop iterable
                # We hard-code it to 4 for now.
                a, b, c, d = input.chunk(input.size(1), dim=1)
                for input_t in (a, b, c, d):
                    h_t, c_t = self.lstm1(input_t, (h_t, c_t))
                    h_t2, c_t2 = self.lstm2(h_t, (h_t2, c_t2))
                    output = self.linear(h_t2)
                    outputs = torch.cat((outputs, output), 1)
                for _ in range(future):  # if we should predict the future
                    h_t, c_t = self.lstm1(output, (h_t, c_t))
                    h_t2, c_t2 = self.lstm2(h_t, (h_t2, c_t2))
                    output = self.linear(h_t2)
                    outputs = torch.cat((outputs, output), 1)
                return outputs

        class Traced(nn.Module):
            def __init__(self):
                super(Traced, self).__init__()
                self.seq = Sequence()

            def forward(self, input):
                return self.seq.forward(input)

        self.checkExportImport(Traced(), (torch.rand(3, 4),))

    @unittest.skip('incorrectly assigned weights')  # FIXME
    def test_vae(self):
        class VAE(nn.Module):
            def __init__(self):
                super(VAE, self).__init__()

                self.fc1 = nn.Linear(784, 400)
                self.fc21 = nn.Linear(400, 20)
                self.fc22 = nn.Linear(400, 20)
                self.fc3 = nn.Linear(20, 400)
                self.fc4 = nn.Linear(400, 784)

            def encode(self, x):
                h1 = F.relu(self.fc1(x))
                return self.fc21(h1), self.fc22(h1)

            def reparameterize(self, mu, logvar):
                if self.training:
                    std = torch.exp(0.5 * logvar)
                    eps = torch.randn_like(std)
                    return eps.mul(std).add_(mu)
                else:
                    return mu

            def decode(self, z):
                h3 = F.relu(self.fc3(z))
                return torch.sigmoid(self.fc4(h3))

            def forward(self, x):
                mu, logvar = self.encode(x.view(-1, 784))
                z = self.reparameterize(mu, logvar)
                return self.decode(z), mu, logvar

        self.checkExportImport(VAE().eval(), (torch.rand(128, 1, 28, 28),))

    def test_torchvision_resnet18(self):
        from .inject_nn import inject_pytorch_nn, remove_inject_pytorch_nn
        try:
            inject_pytorch_nn()
            self.checkExportImport(torchvision.models.resnet18().eval(), (torch.ones(1, 3, 224, 224),))
        finally:
            remove_inject_pytorch_nn()

    def test_resnet(self):
        def conv1x1(in_planes, out_planes, stride=1):
            """1x1 convolution"""
            return nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride, bias=False)

        def conv3x3(in_planes, out_planes, stride=1):
            """3x3 convolution with padding"""
            return nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,
                             padding=1, bias=False)

        class BasicBlock(nn.Module): #torch.jit.ScriptModule
            expansion = 1
            __constants__ = ['downsample']

            def __init__(self, inplanes, planes, stride=1, downsample=None):
                super(BasicBlock, self).__init__()
                self.conv1 = conv3x3(inplanes, planes, stride)
                self.bn1 = nn.BatchNorm2d(planes)
                self.relu = nn.ReLU(inplace=True)
                self.conv2 = conv3x3(planes, planes)
                self.bn2 = nn.BatchNorm2d(planes)
                self.downsample = downsample
                self.stride = stride

            # NOTE: jit cannot be annotated, otherwise, module id is not matched for recorded arguments
            #@torch.jit.script_method
            def forward(self, x):
                residual = x

                out = self.conv1(x)
                out = self.bn1(out)
                out = self.relu(out)

                out = self.conv2(out)
                out = self.bn2(out)

                if self.downsample is not None:
                    residual = self.downsample(x)

                out += residual
                out = self.relu(out)

                return out

        # NOTE: cannot inherit torch.jit.ScriptModule, otherwise, there would be error: 'RecursiveScriptModule' object has no attribute 'graph'
        class ResNet(nn.Module): #torch.jit.ScriptModule
            __constants__ = ['layer1', 'layer2', 'layer3', 'layer4']

            def __init__(self, block, layers, num_classes=1000):
                super(ResNet, self).__init__()
                self.inplanes = 64
                self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3,
                                       bias=False)
                self.bn1 = nn.BatchNorm2d(64)
                self.relu = nn.ReLU(inplace=True)
                self.maxpool = nn.MaxPool2d(kernel_size=3, stride=2, padding=1)
                self.layer1 = self._make_layer(block, 64, layers[0])
                self.layer2 = self._make_layer(block, 128, layers[1], stride=2)
                self.layer3 = self._make_layer(block, 256, layers[2], stride=2)
                self.layer4 = self._make_layer(block, 512, layers[3], stride=2)
                self.avgpool = nn.AdaptiveAvgPool2d((1, 1))
                self.fc = nn.Linear(512 * block.expansion, num_classes)

                for m in self.modules():
                    if isinstance(m, nn.Conv2d):
                        torch.nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
                    elif isinstance(m, nn.BatchNorm2d):
                        torch.nn.init.constant_(m.weight, 1)
                        torch.nn.init.constant_(m.bias, 0)

            def _make_layer(self, block, planes, blocks, stride=1):
                downsample = None
                if stride != 1 or self.inplanes != planes * block.expansion:
                    downsample = nn.Sequential(
                        conv1x1(self.inplanes, planes * block.expansion, stride),
                        nn.BatchNorm2d(planes * block.expansion),
                    )

                layers = []
                layers.append(block(self.inplanes, planes, stride, downsample))
                self.inplanes = planes * block.expansion
                for _ in range(1, blocks):
                    layers.append(block(self.inplanes, planes))

                return nn.Sequential(*layers)

            # NOTE: jit cannot be annotated, otherwise, module id is not matched for recorded arguments
            #@torch.jit.script_method
            def forward(self, x):
                x = self.conv1(x)
                x = self.bn1(x)
                x = self.relu(x)
                x = self.maxpool(x)

                x = self.layer1(x)
                x = self.layer2(x)
                x = self.layer3(x)
                x = self.layer4(x)

                x = self.avgpool(x)
                x = x.view(x.size(0), -1)
                x = self.fc(x)

                return x

        resnet18 = ResNet(BasicBlock, [2, 2, 2, 2])

        self.checkExportImport(resnet18, (torch.randn(1, 3, 224, 224),))

    def test_alexnet(self):
        from .inject_nn import inject_pytorch_nn, remove_inject_pytorch_nn
        try:
            inject_pytorch_nn()
            x = torch.ones(1, 3, 224, 224)
            model = torchvision.models.AlexNet()
            self.checkExportImport(model, (x,))
        finally:
            remove_inject_pytorch_nn()