diff --git a/README.md b/README.md index 52b914f..4ac4bb5 100644 --- a/README.md +++ b/README.md @@ -1 +1,61 @@ -# causalvae + +## Deep Causal Varitional Autoencoder + +To train a supervised variational autoencoder using Deepmind's [dSprites](https://github.com/deepmind/dsprites-dataset) dataset. + +dSprites is a dataset of sprites, which are 2D shapes procedurally generated from 5 ground truth independent "factors." These factors are color, shape, scale, rotation, x and y positions of a sprite. + +All possible combinations of these variables are present exactly once, generating N = 737280 total images. + +Factors and their values: + +* Shape: 3 values {square, ellipse, heart} +* Scale: 6 values linearly spaced in (0.5, 1) +* Orientation: 40 values in (0, 2pi) +* Position X: 32 values in (0, 1) +* Position Y: 32 values in (0, 1) + +There is a sixth factor for color, but it is white for every image in this dataset. + +The purpose of this dataset was to evaluate the ability of disentanglement methods. In these methods, you treat these factors as latent and then try to "disentangle" them in the latent representation. + +However, in this project, these factors are not treated as latent, but are included as labels in the model training. Further, a causal story is invented that relates these factors and the images in a DAG + +![vae_dag](figs/dag.png) + +Structural causal model is of the form: + +![scm_eq](figs/scm.png) + +The image variable will be a 64 x 64 array. The noise term for the image variable will be the traditional Gaussian random variable. The structural assignment *g* for the image variable will be the decoder. + + +## Work: +* Built a Structural causal model that articulates a causal story relating shape, orientation, scale, X, Y, and the data. +* Resampled the dataset to get a new dataset with an empirical distribution that is faithful to the DAG and is entailed by the SCM +* To implement a causal VAE using [Pyro](http://pyro.ai/) by extending the primitive version of VAE. The VAE is fully supervised. +* Finally used the trained model to answer some counterfactual queries, for example, "given this image of a heart with this orientation, position, and scale, what would it have looked like if it were a square?" + +## Optimization: +* The code is made compatible for GPU for faster processing. +* The learned weights are saved to avoid training frequently to enhance development efficiency. + +## Results +* Achieved good reconstruction accuracy using vanilla VAE - +![vanilla_vae](figs/vae-recons.png) + +* Trained VAE and made sure it recognises changes in the latent dimensions(Manually changed latent variables before training) - +![vae_latent_manual](figs/manual-intervention.png) + +* Built structural causal model and verfied for reconstruction accuracy - +![scm_vae](figs/scm-conditioned.png) + +* Counterfactual queries (1) - Intervention on shape - Given a oval and certain (x,y) co-ordinates and orientation, how would it look it was a sqaure? +![scm_intervention_vae](figs/intervention.png) + +* Counterfactual queries (1) - Intervention on shape, position of (x,y) - +![scm_intervention_vae](figs/intervention-2.png) + +## Applications +* DeepFakes :[Structured Disentangled Representations](https://arxiv.org/pdf/1804.02086.pdf) +![dp](deep-fakes.jpg) diff --git a/causal_vae_dsprites.ipynb b/causal_vae_dsprites.ipynb new file mode 100644 index 0000000..ec540fe --- /dev/null +++ b/causal_vae_dsprites.ipynb @@ -0,0 +1,1851 @@ +{ + "nbformat": 4, + "nbformat_minor": 0, + "metadata": { + "kernelspec": { + "name": "python3", + "display_name": "Python 3" + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.6.8" + }, + "colab": { + "name": "causal_vae_dsprites.ipynb", + "version": "0.3.2", + "provenance": [], + "collapsed_sections": [], + "machine_shape": "hm", + "include_colab_link": true + }, + "accelerator": "GPU" + }, + "cells": [ + { + "cell_type": "markdown", + "metadata": { + "id": "view-in-github", + "colab_type": "text" + }, + "source": [ + "\"Open" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "zSOMAHldq2xR", + "colab_type": "text" + }, + "source": [ + "## Deep Causal Variational Inference\n", + "\n", + "### Introduction:\n", + "To train a supervised variational autoencoder using Deepmind's [dSprites](https://github.com/deepmind/dsprites-dataset) dataset.\n", + "\n", + "dSprites is a dataset of sprites, which are 2D shapes procedurally generated from 5 ground truth independent \"factors.\" These factors are color, shape, scale, rotation, x and y positions of a sprite.\n", + "\n", + "All possible combinations of these variables are present exactly once, generating N = 737280 total images.\n", + "\n", + "Factors and their values:\n", + "\n", + "* Shape: 3 values {square, ellipse, heart}\n", + "* Scale: 6 values linearly spaced in (0.5, 1)\n", + "* Orientation: 40 values in (0, 2$\\pi$)\n", + "* Position X: 32 values in (0, 1)\n", + "* Position Y: 32 values in (0, 1)\n", + "\n", + "\n", + "Further, the objective of any generative model is essentially to capture underlying data generative factors, the disentangled representation would mean a single latent unit being sensitive to variations in single generative factors\n", + "\n", + "\n", + "### Goal:\n", + "To include the latent factors as labels in the training and to invent a causal story that relates these factors and the images in a DAG.\n", + "\n", + "Reference \n", + "\n", + "[Structured Disentangled Representation](https://arxiv.org/pdf/1804.02086.pdf)" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "CMhHGoIwCwkN", + "colab_type": "code", + "colab": {} + }, + "source": [ + "#Install dependencies\n", + "!pip3 install pyro-ppl\n", + "!pip3 install torch torchvision\n", + "!pip3 install pydrive --upgrade\n", + "!pip3 install tqdm" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "uNa3MO8GCqWg", + "colab_type": "code", + "colab": {} + }, + "source": [ + "# Load necessary libraries\n", + "from matplotlib import pyplot as plt\n", + "import numpy as np\n", + "import seaborn as sns\n", + "\n", + "import os\n", + "from collections import defaultdict\n", + "\n", + "import torch\n", + "import torch.nn as nn\n", + "\n", + "from tqdm import tqdm\n", + "import pyro\n", + "import pyro.distributions as dist\n", + "from pyro.infer import SVI, Trace_ELBO, TraceEnum_ELBO, config_enumerate, EmpiricalMarginal\n", + "from pyro.optim import Adam, SGD\n", + "import torch.distributions.constraints as constraints\n", + "\n", + "# Change figure aesthetics\n", + "%matplotlib inline\n", + "sns.set_context('talk', font_scale=1.2, rc={'lines.linewidth': 1.5})\n", + "\n", + "from ipywidgets import interact, interactive, fixed, interact_manual\n", + "import ipywidgets as widgets\n", + "\n", + "#to utilize GPU capabilities\n", + "USE_CUDA = True\n", + "\n", + "pyro.enable_validation(True)\n", + "pyro.distributions.enable_validation(False)" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "nkXRHmSqDFTy", + "colab_type": "code", + "outputId": "40eb56b2-2dcd-42d2-942f-ccf0e4c08982", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 124 + } + }, + "source": [ + "# Mount Google drive to load data\n", + "from google.colab import drive\n", + "drive.mount('/content/gdrive')" + ], + "execution_count": 0, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Go to this URL in a browser: https://accounts.google.com/o/oauth2/auth?client_id=947318989803-6bn6qk8qdgf4n4g3pfee6491hc0brc4i.apps.googleusercontent.com&redirect_uri=urn%3Aietf%3Awg%3Aoauth%3A2.0%3Aoob&scope=email%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdocs.test%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive.photos.readonly%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fpeopleapi.readonly&response_type=code\n", + "\n", + "Enter your authorization code:\n", + "··········\n", + "Mounted at /content/gdrive\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "KRCLhKI7d2r1", + "colab_type": "code", + "colab": {} + }, + "source": [ + "# Mount G drive to access files\n", + "from pydrive.auth import GoogleAuth\n", + "from pydrive.drive import GoogleDrive\n", + "from google.colab import auth\n", + "from oauth2client.client import GoogleCredentials\n", + "\n", + "auth.authenticate_user()\n", + "gauth = GoogleAuth()\n", + "gauth.credentials = GoogleCredentials.get_application_default()\n", + "drive = GoogleDrive(gauth)" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "cqyc4SqyDLIq", + "colab_type": "code", + "outputId": "3d558d65-597e-47dc-fe0f-7a9f9dad6945", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 262 + } + }, + "source": [ + "# Hack to get all available GPU ram.\n", + "\n", + "import tensorflow as tf\n", + "tf.test.gpu_device_name()\n", + "\n", + "!ln -sf /opt/bin/nvidia-smi /usr/bin/nvidia-smi\n", + "!pip install gputil\n", + "!pip install psutil\n", + "!pip install humanize\n", + "import psutil\n", + "import humanize\n", + "import os\n", + "import GPUtil as GPU\n", + "GPUs = GPU.getGPUs()\n", + "# XXX: only one GPU on Colab and isn’t guaranteed\n", + "gpu = GPUs[0]\n", + "def printm():\n", + " process = psutil.Process(os.getpid())\n", + " print(\"Gen RAM Free: \" + humanize.naturalsize( psutil.virtual_memory().available ), \" | Proc size: \" + humanize.naturalsize( process.memory_info().rss))\n", + " print(\"GPU RAM Free: {0:.0f}MB | Used: {1:.0f}MB | Util {2:3.0f}% | Total {3:.0f}MB\".format(gpu.memoryFree, gpu.memoryUsed, gpu.memoryUtil*100, gpu.memoryTotal))\n", + "printm()" + ], + "execution_count": 0, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Collecting gputil\n", + " Downloading https://files.pythonhosted.org/packages/ed/0e/5c61eedde9f6c87713e89d794f01e378cfd9565847d4576fa627d758c554/GPUtil-1.4.0.tar.gz\n", + "Building wheels for collected packages: gputil\n", + " Building wheel for gputil (setup.py) ... \u001b[?25l\u001b[?25hdone\n", + " Created wheel for gputil: filename=GPUtil-1.4.0-cp36-none-any.whl size=7411 sha256=0ecd5811e41f36cf83f484d14611cf7cf4232041ade307328b392bc99b263545\n", + " Stored in directory: /root/.cache/pip/wheels/3d/77/07/80562de4bb0786e5ea186911a2c831fdd0018bda69beab71fd\n", + "Successfully built gputil\n", + "Installing collected packages: gputil\n", + "Successfully installed gputil-1.4.0\n", + "Requirement already satisfied: psutil in /usr/local/lib/python3.6/dist-packages (5.4.8)\n", + "Requirement already satisfied: humanize in /usr/local/lib/python3.6/dist-packages (0.5.1)\n", + "Gen RAM Free: 25.2 GB | Proc size: 2.2 GB\n", + "GPU RAM Free: 11111MB | Used: 330MB | Util 3% | Total 11441MB\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "nSkXleG1CqWj", + "colab_type": "code", + "colab": {} + }, + "source": [ + "class Encoder(nn.Module):\n", + "\t\"\"\"\n", + " MLPs (multi-layered perceptrons or simple feed-forward networks)\n", + " where the provided activation parameter is used on every linear layer except\n", + " for the output layer where we use the provided output_activation parameter\n", + "\t\"\"\"\n", + "\tdef __init__(self, image_dim, label_dim, z_dim):\n", + "\t\tsuper(Encoder, self).__init__()\n", + "\t\t#setup image and label dimensions from the dataset\n", + "\t\tself.image_dim = image_dim\n", + "\t\tself.label_dim = label_dim\n", + "\t\tself.z_dim = z_dim\n", + "\t\t# setup the three linear transformations used\n", + "\t\tself.fc1 = nn.Linear(self.image_dim+self.label_dim, 1000)\n", + "\t\tself.fc2 = nn.Linear(1000, 1000)\n", + "\t\tself.fc31 = nn.Linear(1000, z_dim) # mu values\n", + "\t\tself.fc32 = nn.Linear(1000, z_dim) # sigma values\n", + "\t\t# setup the non-linearities\n", + "\t\tself.softplus = nn.Softplus()\n", + "\n", + "\tdef forward(self, xs, ys):\n", + "\t\txs = xs.reshape(-1, self.image_dim)\n", + "\t\t#now concatenate the image and label\n", + "\t\tinputs = torch.cat((xs,ys), -1)\n", + "\t\t# then compute the hidden units\n", + "\t\thidden1 = self.softplus(self.fc1(inputs))\n", + "\t\thidden2 = self.softplus(self.fc2(hidden1))\n", + "\t\t# then return a mean vector and a (positive) square root covariance\n", + "\t\t# each of size batch_size x z_dim\n", + "\t\tz_loc = self.fc31(hidden2)\n", + "\t\tz_scale = torch.exp(self.fc32(hidden2))\n", + "\t\treturn z_loc, z_scale\n", + "\n", + "\n", + "class Decoder(nn.Module):\n", + "\tdef __init__(self, image_dim, label_dim, z_dim):\n", + "\t\tsuper(Decoder, self).__init__()\n", + "\t\t# setup the two linear transformations used\n", + "\t\thidden_dim = 1000\n", + "\t\tself.fc1 = nn.Linear(z_dim+label_dim, hidden_dim)\n", + "\t\tself.fc2 = nn.Linear(hidden_dim, hidden_dim)\n", + "\t\tself.fc3 = nn.Linear(hidden_dim, hidden_dim)\n", + "\t\tself.fc4 = nn.Linear(hidden_dim, image_dim)\n", + "\t\t# setup the non-linearities\n", + "\t\tself.softplus = nn.Softplus()\n", + "\t\tself.sigmoid = nn.Sigmoid()\n", + "\n", + "\tdef forward(self, zs, ys):\n", + "\t\tinputs = torch.cat((zs, ys),-1)\n", + "\t\t# then compute the hidden units\n", + "\t\thidden1 = self.softplus(self.fc1(inputs))\n", + "\t\thidden2 = self.softplus(self.fc2(hidden1))\n", + "\t\thidden3 = self.softplus(self.fc3(hidden2))\n", + "\t\t# return the parameter for the output Bernoulli\n", + "\t\t# each is of size batch_size x 784\n", + "\t\tloc_img = self.sigmoid(self.fc4(hidden3))\n", + "\t\treturn loc_img" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "Yqj05Mv8CqWm", + "colab_type": "code", + "colab": {} + }, + "source": [ + "class CVAE(nn.Module):\n", + "\t\"\"\"\n", + "\tThis class encapsulates the parameters (neural networks) and models & guides \n", + "\tneeded to train a supervised variational auto-encoder \n", + "\t\"\"\"\n", + "\tdef __init__(self, config_enum=None, use_cuda=False, aux_loss_multiplier=None):\n", + " \n", + "\t\tsuper(CVAE, self).__init__()\n", + "\t\tself.image_dim = 64**2\n", + "\t\tself.label_shape = np.array((1,3,6,40,32,32))\n", + "\t\tself.label_names = np.array(('color', 'shape', 'scale', 'orientation', 'posX', 'posY'))\n", + "\t\tself.label_dim = np.sum(self.label_shape)\n", + "\t\tself.z_dim = 50 \n", + "\t\tself.allow_broadcast = config_enum == 'parallel'\n", + "\t\tself.use_cuda = use_cuda\n", + "\t\tself.aux_loss_multiplier = aux_loss_multiplier\n", + "\t # define and instantiate the neural networks representing\n", + "\t # the paramters of various distributions in the model\n", + "\t\tself.setup_networks()\n", + "\n", + "\tdef setup_networks(self):\n", + "\t\t\"\"\"\n", + "\t\tSetup and initialize Encoder and decoder units\n", + "\t\t\"\"\"\n", + "\t\tself.encoder = Encoder(self.image_dim, self.label_dim, self.z_dim)\n", + "\t\tself.decoder = Decoder(self.image_dim, self.label_dim, self.z_dim)\n", + "\t\t# using GPUs for faster training of the networks\n", + "\t\tif self.use_cuda:\n", + "\t\t self.cuda()\n", + "\n", + "\tdef model(self, xs, ys):\n", + "\t\tpyro.module(\"cvae\", self)\n", + "\t\tbatch_size = xs.size(0)\n", + "\t\toptions = dict(dtype=xs.dtype, device=xs.device)\n", + "\t\twith pyro.plate(\"data\"):\n", + "\t\t\tprior_loc = torch.zeros(batch_size, self.z_dim, **options)\n", + "\t\t\tprior_scale = torch.ones(batch_size, self.z_dim, **options)\n", + "\t\t\tzs = pyro.sample(\"z\", dist.Normal(prior_loc, prior_scale).to_event(1))\n", + "\t\t\t# if the label y (which digit to write) is supervised, sample from the\n", + "\t\t\t# constant prior, otherwise, observe the value (i.e. score it against the constant prior)\n", + "\t\t\tloc = self.decoder.forward(zs, self.remap_y(ys))\n", + "\t\t\tpyro.sample(\"x\", dist.Bernoulli(loc).to_event(1), obs=xs)\n", + "\t\t \t# return the loc so we can visualize it later\n", + "\t\t\treturn loc\n", + "\n", + "\tdef guide(self, xs, ys):\n", + "\t\twith pyro.plate(\"data\"):\n", + "\t\t\t# sample (and score) the latent handwriting-style with the variational\n", + "\t\t\t# distribution q(z|x) = normal(loc(x),scale(x))\n", + "\t\t\tloc, scale = self.encoder.forward(xs, self.remap_y(ys))\n", + "\t\t\tpyro.sample(\"z\", dist.Normal(loc, scale).to_event(1))\n", + "\n", + "\tdef remap_y(self, ys):\n", + "\t\tnew_ys = []\n", + "\t\toptions = dict(dtype=ys.dtype, device=ys.device)\n", + "\t\tfor i, label_length in enumerate(self.label_shape):\n", + "\t\t prior = torch.ones(ys.size(0), label_length, **options) / (1.0 * label_length)\n", + "\t\t new_ys.append(pyro.sample(\"y_%s\" % self.label_names[i], dist.OneHotCategorical(prior), \n", + "\t\t obs=torch.nn.functional.one_hot(ys[:,i].to(torch.int64), int(label_length))))\n", + "\t\tnew_ys = torch.cat(new_ys, -1)\n", + "\t\treturn new_ys.to(torch.float32)\n", + "\n", + "\tdef reconstruct_image(self, xs, ys):\n", + "\t\t# backward\n", + "\t\tsim_z_loc, sim_z_scale = self.encoder.forward(xs, self.remap_y(ys))\n", + "\t\tzs = dist.Normal(sim_z_loc, sim_z_scale).to_event(1).sample()\n", + "\t\t# forward\n", + "\t\tloc = self.decoder.forward(zs, self.remap_y(ys))\n", + "\t\treturn dist.Bernoulli(loc).to_event(1).sample()" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "1mmMe3D8CqWo", + "colab_type": "code", + "colab": {} + }, + "source": [ + "def setup_data_loaders(train_x, test_x, train_y, test_y, batch_size=128, use_cuda=False):\n", + "\ttrain_dset = torch.utils.data.TensorDataset(\n", + "\t torch.from_numpy(train_x.astype(np.float32)).reshape(-1, 4096),\n", + "\t torch.from_numpy(train_y.astype(np.float32))\n", + "\t)\n", + "\ttest_dset = torch.utils.data.TensorDataset(\n", + "\t torch.from_numpy(test_x.astype(np.float32)).reshape(-1, 4096),\n", + "\t torch.from_numpy(test_y.astype(np.float32))\n", + "\t) \n", + "\tkwargs = {'num_workers': 1, 'pin_memory': use_cuda}\n", + "\ttrain_loader = torch.utils.data.DataLoader(\n", + "\t dataset=train_dset, batch_size=batch_size, shuffle=False, **kwargs\n", + "\t)\n", + "\n", + "\ttest_loader = torch.utils.data.DataLoader(\n", + "\t dataset=test_dset, batch_size=batch_size, shuffle=False, **kwargs\n", + "\t)\n", + "\treturn {\"train\":train_loader, \"test\":test_loader}" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "azMSBtsnCqWr", + "colab_type": "code", + "colab": {} + }, + "source": [ + "dataset_zip = np.load(\n", + " '/content/gdrive/My Drive/data-science/causal-ml/projects/dsprites-dataset/dsprites_ndarray_co1sh3sc6or40x32y32_64x64.npz',\n", + " encoding = 'bytes',\n", + " allow_pickle=True\n", + ")\n", + "\n", + "imgs = dataset_zip['imgs']\n", + "labels = dataset_zip['latents_classes']\n", + "label_sizes = dataset_zip['metadata'][()][b'latents_sizes']\n", + "label_names = dataset_zip['metadata'][()][b'latents_names']\n", + "\n", + "# Sample imgs randomly\n", + "indices_sampled = np.arange(imgs.shape[0])\n", + "np.random.shuffle(indices_sampled)\n", + "imgs_sampled = imgs[indices_sampled]\n", + "labels_sampled = labels[indices_sampled]\n", + "\n", + "data_loaders = setup_data_loaders(\n", + " imgs_sampled[1000:],\n", + " imgs_sampled[:1000],\n", + " labels_sampled[1000:],\n", + " labels_sampled[:1000],\n", + " batch_size=256,\n", + " use_cuda=USE_CUDA\n", + ")" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "Eamf-n9hCqWt", + "colab_type": "code", + "colab": {} + }, + "source": [ + "def train(svi, train_loader, use_cuda=False):\n", + "\t# initialize loss accumulator\n", + "\tepoch_loss = 0.\n", + "\t# do a training epoch over each mini-batch x returned\n", + "\t# by the data loader\n", + "\tfor xs,ys in train_loader:\n", + "\t # if on GPU put mini-batch into CUDA memory\n", + "\t if use_cuda:\n", + "\t xs = xs.cuda()\n", + "\t ys = ys.cuda()\n", + "\t # do ELBO gradient and accumulate loss\n", + "\t epoch_loss += svi.step(xs, ys)\n", + "\t# return epoch loss\n", + "\tnormalizer_train = len(train_loader.dataset)\n", + "\ttotal_epoch_loss_train = epoch_loss / normalizer_train\n", + "\treturn total_epoch_loss_train\n", + "\n", + "def evaluate(svi, test_loader, use_cuda=False):\n", + "\t# initialize loss accumulator\n", + "\ttest_loss = 0.\n", + "\t# compute the loss over the entire test set\n", + "\tfor xs, ys in test_loader:\n", + "\t # if on GPU put mini-batch into CUDA memory\n", + "\t if use_cuda:\n", + "\t xs = xs.cuda()\n", + "\t ys = ys.cuda()\n", + "\t # compute ELBO estimate and accumulate loss\n", + "\t test_loss += svi.evaluate_loss(xs, ys)\n", + "\tnormalizer_test = len(test_loader.dataset)\n", + "\ttotal_epoch_loss_test = test_loss / normalizer_test\n", + "\treturn total_epoch_loss_test" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "LR-hW5UzCqWv", + "colab_type": "code", + "colab": {} + }, + "source": [ + "# Run options\n", + "LEARNING_RATE = 1.0e-3\n", + "\n", + "# Run only for a single iteration for testing\n", + "NUM_EPOCHS = 10\n", + "TEST_FREQUENCY = 5" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "puMbph4idkYg", + "colab_type": "code", + "outputId": "928e47f6-fd32-49af-a8fa-5fcc4b99a8c1", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 35 + } + }, + "source": [ + "#################################\n", + "### FOR SAVING AND LOADING MODEL\n", + "################################\n", + "# clear param store\n", + "\n", + "pyro.clear_param_store()\n", + "\n", + "network_path = \"/content/gdrive/My Drive/data-science/causal-ml/projects/trained_model.save\"\n", + "\n", + "PATH = \"trained_model.save\"\n", + "\n", + "# new model\n", + "# vae = CVAE(use_cuda=USE_CUDA)\n", + "\n", + "# save current model\n", + "# torch.save(vae.state_dict(), PATH)\n", + "\n", + "# to load params from trained model\n", + "vae = CVAE(use_cuda=USE_CUDA)\n", + "vae.load_state_dict(torch.load(network_path))" + ], + "execution_count": 0, + "outputs": [ + { + "output_type": "execute_result", + "data": { + "text/plain": [ + "" + ] + }, + "metadata": { + "tags": [] + }, + "execution_count": 12 + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "TtCW3_agE7N_", + "colab_type": "text" + }, + "source": [ + "### **DONT RUN THE BELOW CODE AS WE'VE ALREADY TRAINED THE MODEL AND WE'VE STORED THE NETWORK PARAMS**\n", + "\n", + "## ==================================================================================" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "hIuVKC2lCqWx", + "colab_type": "code", + "colab": {} + }, + "source": [ + "import warnings\n", + "warnings.filterwarnings('ignore')\n", + "\n", + "# clear param store\n", + "pyro.clear_param_store()\n", + "\n", + "# setup the VAE\n", + "vae = CVAE(use_cuda=USE_CUDA)\n", + "\n", + "# setup the optimizer\n", + "adam_args = {\"lr\": LEARNING_RATE}\n", + "optimizer = Adam(adam_args)\n", + "\n", + "# setup the inference algorithm\n", + "svi = SVI(vae.model, vae.guide, optimizer, loss=Trace_ELBO())\n", + "\n", + "train_elbo = []\n", + "test_elbo = []\n", + "# training loop\n", + "\n", + "VERBOSE = True\n", + "pbar = tqdm(range(NUM_EPOCHS))\n", + "for epoch in pbar:\n", + " total_epoch_loss_train = train(svi, data_loaders[\"train\"], use_cuda=USE_CUDA)\n", + " train_elbo.append(-total_epoch_loss_train)\n", + " if VERBOSE:\n", + " print(\"[epoch %03d] average training loss: %.4f\" % (epoch, total_epoch_loss_train))\n", + " if epoch % TEST_FREQUENCY == 0:\n", + " # report test diagnostics\n", + " total_epoch_loss_test = evaluate(svi, data_loaders[\"test\"], use_cuda=USE_CUDA)\n", + " test_elbo.append(-total_epoch_loss_test)\n", + " if VERBOSE:\n", + " print(\"[epoch %03d] average test loss: %.4f\" % (epoch, total_epoch_loss_test))" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "navbYZazFD0E", + "colab_type": "text" + }, + "source": [ + "## ==================================================================================" + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "aFnfYMd3CqW0", + "colab_type": "text" + }, + "source": [ + "## Visualizing the reconstruction accuracy of VAE" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "IZQ_mi_GMm_v", + "colab_type": "code", + "colab": {} + }, + "source": [ + "" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "nIt0T2eUCqW0", + "colab_type": "code", + "colab": {} + }, + "source": [ + "import warnings\n", + "warnings.filterwarnings(\"ignore\")\n", + "\n", + "data_iter = iter(data_loaders[\"train\"])\n", + "xs, ys = next(data_iter)\n", + "\n", + "if USE_CUDA:\n", + " xs = xs.cuda()\n", + " ys = ys.cuda()\n", + "rs = vae.reconstruct_image(xs, ys)\n", + "if USE_CUDA:\n", + " xs = xs.cpu()\n", + " rs = rs.cpu()\n", + "originals = xs.numpy().reshape(-1, 64,64)\n", + "recons = rs.reshape(-1,64,64)" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "1YBT2lo7CqW2", + "colab_type": "code", + "outputId": "13118f8b-c9b0-4ebd-98b4-0b404b40074c", + "colab": { + "resources": { + "http://localhost:8080/nbextensions/google.colab/colabwidgets/controls.css": { + "data": "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", + "ok": true, + "headers": [ + [ + "content-type", + "text/css" + ] + ], + "status": 200, + "status_text": "" + } + }, + "base_uri": "https://localhost:8080/", + "height": 270 + } + }, + "source": [ + "# [ 0, 2, 1, 34, 4, 24]\n", + "def recon_check(original, recon):\n", + " fig = plt.figure()\n", + " ax0 = fig.add_subplot(121)\n", + " plt.imshow(original, cmap='Greys_r', interpolation='nearest')\n", + " plt.axis('off')\n", + " ax1 = fig.add_subplot(122)\n", + " plt.imshow(recon , cmap='Greys_r', interpolation='nearest')\n", + " plt.axis('off')\n", + " \n", + "def f(x):\n", + " fig = plt.figure()\n", + " ax0 = fig.add_subplot(121)\n", + " plt.imshow(originals[x], cmap='Greys_r', interpolation='nearest')\n", + " plt.axis('off')\n", + " ax1 = fig.add_subplot(122)\n", + " plt.imshow(recons[x], cmap='Greys_r', interpolation='nearest')\n", + " plt.axis('off')\n", + " \n", + "interact(f, x=widgets.IntSlider(min=0, max=xs.shape[0], step=1, value=0))" + ], + "execution_count": 341, + "outputs": [ + { + "output_type": "display_data", + "data": { + "application/vnd.jupyter.widget-view+json": { + "model_id": "ef9cf802f2784126bef5615081093250", + "version_minor": 0, + "version_major": 2 + }, + "text/plain": [ + "interactive(children=(IntSlider(value=0, description='x', max=256), Output()), _dom_classes=('widget-interact'…" + ] + }, + "metadata": { + "tags": [] + } + }, + { + "output_type": "execute_result", + "data": { + "text/plain": [ + "" + ] + }, + "metadata": { + "tags": [] + }, + "execution_count": 341 + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "EVCzRnuIzNpL", + "colab_type": "code", + "outputId": "319975aa-e56c-4814-c72f-15f1d8187c91", + "colab": { + "resources": { + "http://localhost:8080/nbextensions/google.colab/colabwidgets/controls.css": { + "data": "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", + "ok": true, + "headers": [ + [ + "content-type", + "text/css" + ] + ], + "status": 200, + "status_text": "" + } + }, + "base_uri": "https://localhost:8080/", + "height": 464 + } + }, + "source": [ + "y_names = ['shape', 'scale', 'orientation', 'posX', 'posY']\n", + "y_shapes = np.array((3,6,40,32,32))\n", + "img_dict = {}\n", + "\n", + "for i, img in enumerate(imgs_sampled):\n", + " img_dict[tuple(labels_sampled[i])] = img\n", + "\n", + "def find_in_dataset(shape, scale, orient, posX, posY):\n", + " fig = plt.figure()\n", + " img = img_dict[(0, shape, scale, orient, posX, posY)]\n", + " plt.imshow(img.reshape(64,64), cmap='Greys_r', interpolation='nearest')\n", + " plt.axis('off')\n", + "\n", + "interact(find_in_dataset, \n", + " shape=widgets.IntSlider(min=0, max=2, step=1, value=0),\n", + " scale=widgets.IntSlider(min=0, max=5, step=1, value=0),\n", + " orient=widgets.IntSlider(min=0, max=39, step=1, value=0),\n", + " posX=widgets.IntSlider(min=0, max=31, step=1, value=0),\n", + " posY=widgets.IntSlider(min=0, max=31, step=1, value=0))" + ], + "execution_count": 0, + "outputs": [ + { + "output_type": "display_data", + "data": { + "application/vnd.jupyter.widget-view+json": { + "model_id": "e787b532e7d84bf7a7d614eba2f930ff", + "version_minor": 0, + "version_major": 2 + }, + "text/plain": [ + "interactive(children=(IntSlider(value=0, description='shape', max=2), IntSlider(value=0, description='scale', …" + ] + }, + "metadata": { + "tags": [] + } + }, + { + "output_type": "execute_result", + "data": { + "text/plain": [ + "" + ] + }, + "metadata": { + "tags": [] + }, + "execution_count": 15 + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "1EueLb0OzSDm", + "colab_type": "code", + "outputId": "e6365256-b1f9-4fc6-c1b3-959c4dc381cf", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 311 + } + }, + "source": [ + "def get_specific_data(args=dict(), cuda=False):\n", + " '''\n", + " use this function to get examples of data with specific class labels\n", + " inputs: \n", + " args - dictionary whose keys can include {shape, scale, orientation,\n", + " posX, posY} and values can include any integers less than the \n", + " corresponding size of that label dimension\n", + " cuda - bool to indicate whether the output should be placed on GPU\n", + " '''\n", + " names_dict = {'shape': 1, 'scale': 2, 'orientation': 3, 'posX': 4, 'posY': 5}\n", + " selected_ind = np.ones(imgs.shape[0], dtype=bool)\n", + " for k,v in args.items():\n", + " col_id = names_dict[k]\n", + " selected_ind = np.bitwise_and(selected_ind, labels[:, col_id] == v)\n", + " ind = np.random.choice(np.arange(imgs.shape[0])[selected_ind])\n", + " x = torch.from_numpy(imgs[ind].reshape(1,64**2).astype(np.float32))\n", + " y = torch.from_numpy(labels[ind].reshape(1,6).astype(np.float32))\n", + " if not cuda:\n", + " return x,y\n", + " x = x.cuda()\n", + " y = y.cuda()\n", + " return x,y\n", + "\n", + "def plot_image(x):\n", + " \"\"\"\n", + " helper to plot dSprites images\n", + " \"\"\"\n", + " x = x.cpu()\n", + " plt.figure()\n", + " plt.imshow(x.reshape(64,64), interpolation='nearest', cmap='Greys_r')\n", + " plt.axis('off')\n", + "\n", + "def see_specific_image(args=dict(), verbose=True):\n", + " '''\n", + " use this function to get examples of data with specific class labels\n", + " inputs: \n", + " args - dictionary whose keys can include {shape, scale, orientation,\n", + " posX, posY} and values can include any integers less than the \n", + " corresponding size of that label dimension\n", + " verbose - bool to indicate whether the full class label should be written \n", + " as the title of the plot\n", + " '''\n", + " x,y = get_specific_data(args, cuda=False)\n", + " plot_image(x)\n", + " if verbose:\n", + " string = ''\n", + " for i, s in enumerate(['Shape', 'Scale', 'Orientation', 'PosX', 'PosY']):\n", + " string += '%s: %d, ' % (s, int(y[0][i+1]))\n", + " if i == 2:\n", + " string = string[:-2] + '\\n'\n", + " plt.title(string[:-2], fontsize=12)\n", + " \n", + "def compare_reconstruction(original, recon):\n", + " \"\"\"\n", + " compare two images side by side\n", + " inputs:\n", + " original - array for original image\n", + " recon - array for recon image\n", + " \"\"\"\n", + " fig = plt.figure()\n", + " ax0 = fig.add_subplot(121)\n", + " plt.imshow(original.cpu().reshape(64,64), cmap='Greys_r', interpolation='nearest')\n", + " plt.axis('off')\n", + " plt.title('original')\n", + " ax1 = fig.add_subplot(122)\n", + " plt.imshow(recon.cpu().reshape(64,64), cmap='Greys_r', interpolation='nearest')\n", + " plt.axis('off')\n", + " plt.title('reconstruction')\n", + " \n", + "def compare_to_density(original, recons):\n", + " \"\"\"\n", + " compare two images side by side\n", + " inputs:\n", + " original - array for original image\n", + " recon - array of multiple recon images\n", + " \"\"\"\n", + " fig = plt.figure()\n", + " ax0 = fig.add_subplot(121)\n", + " plt.imshow(original.cpu().reshape(64,64), cmap='Greys_r', interpolation='nearest')\n", + " plt.axis('off')\n", + " plt.title('original')\n", + " ax1 = fig.add_subplot(122)\n", + " plt.imshow(torch.mean(recons.cpu(), 0).reshape(64,64), cmap='Greys_r', interpolation='nearest')\n", + " plt.axis('off')\n", + " plt.title('reconstructions')\n", + "\n", + " \n", + "see_specific_image()" + ], + "execution_count": 0, + "outputs": [ + { + "output_type": "display_data", + "data": { + "image/png": "iVBORw0KGgoAAAANSUhEUgAAAQ0AAAEmCAYAAAB4ecX9AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAEYFJREFUeJzt3Xu0HWV5x/HvQ0KAYMiFUCoQwkIU\nBBbFG8hSaorYNipKW61FBKJ4iS4WlVLRskChFFl1VRAviOIlUKJ4Q6FRFFEjFQQvy9qaEjVows1Q\nQhLwBLxg3v7xvgeGvc4+2U9Ccsjh+1lrr5w9756Zd2bP/Gbm3fNOopSCJA1qm7GugKSti6EhKcXQ\nkJRiaEhKMTQkpRgaklKeMKEREfMi4jtjXY+tQUQsjojXj3U9NlZEXBwRZ451PQAi4vCI+OlY1+Ox\nNK5CIyKeHxE3RsR9EbE6Im6IiOeMdb02JCJeHRErImJdRHwpImYkxj09In4ZEUMRcUdEfGZz1jUr\nIg6OiB9GxAPt34MT40ZEvC0ifh4RD0bEbRFxXkRsN9p4pZT5pZRzHoO6z4mIO5LjlIjYp1OX/yyl\n7LupdRlgvk+LiKsi4p627X8tIvbt+cwpEbEyIu6PiE9saD32M25CIyJ2AhYBHwBmALsDZwO/Hct6\nbUhEHAB8BDgO2BV4ALhowHFPaOMdWUp5EvBs4BubqappETEJuAq4HJgOXApc1YYP4v3AG4HjgSnA\nXOCFwGdHmeeETanzVmwacDWwL3U7+h513QMQEX8BvIO6/mYDe1P3j7xSyrh4UXeYtaOUzwO+A/wb\nsAb4JTC3U/5a4Bbg18AvgDd1yuYAdwCnA6uA5cCxnfLt2nRvA+4GLgZ2GLDe7wY+1Xn/FOB3wJQB\nxv0g8L5RymcAnwTuasv8pTZ8OjVg72nDFwF7dMZbDLy+8/51bd2sAb4GzB5w2f4cuBOIzrDbgL8c\nYNynAn8ADukZPot6IDiivV8AfBj4CrAOOLIN+5fOOC8F/gtYC9wIHNQpWw78I/DfwH3AZ4DtgR2B\nB4H1wFB77QYcAny3TetX7TuY1KZ1PVBaPYaAVw1vO535Pb2t37XAEuBlnbIFwIeAL7ft8GbgKRu5\nP8xoddm5vf8U8O5O+QuBlRs17c29M2+pF7ATcC/1aDYXmN5TPg/4PfAGYALw5rYzRSt/CXWHDeAF\n1CP+M1vZHOAh4HxqQLygbRj7tvILqCk/g3pE/A/gvM681wLP71Pvq4C39wwbAp41wDK/BlgNvI0a\nmhN6yr/cdoLpwLbAC9rwnYG/ASa3+n6OFiitfDEtNICXA8vaxj4ROAO4sfPZRcA7+tTvFOCanmGL\ngFMHWLb5wIo+Zd8eXr9tR7sPeB71zHl7OqEBPAP4P+DQ9r2fQA2K7Vr5cupRebf2/d0CzO9873f0\nzPtZwHPbutirff6tnfIC7NN5//A02newjHrwmQQcQQ2HfTvLci81mCYCC4ErBlnXI6yjo4Ffdd7/\nGHhV5/1MOqGS2tfGemd/LF9tw15APSt4iLoj79rK5gHLOp+d3FbaH/eZ1peAv+988Q8BO3bKPwuc\nSQ2ZdXSOCMBhwC8HrPM3hjfSzrA7gTkDjn8scF2rw720AAKeTD1KTh9gGgcDazrvF/NIaFwDnNgp\n24YaqLMHmO6Z3Y2+DVsInDXAuGcAN/UpuwK4pP29ALisp3wBj4TGh4Fzesp/yiMBuhx4TafsPcDF\nne/9jg3U863AFzvvRwuNw4GVwDad8k8Pr49W7491yl4MLN2I/WCPtg0d0xl2K50zPGqAFWCv7PTH\nTZsGQCnlllLKvFLKHsCB1KPH+zofWdn57APtzycBRMTciLipNSKtpX5hMzvjrimlrOu8X9Gmvws1\ngH4YEWvbuF9twwcxRD1L6tqJegTaoFLKwlLKkdRr2vnAOe36dRawupSypneciJgcER9pja/3U0+r\np/VpD5gNXNhZttXUoNx9My/bKmrwjeTJrXzY7aNMZzZw6nD92zLMon53w1Z2/n6Atk2MpDU4Lhpu\nUKReXs7s9/keuwG3l1LWd4at4NHrcuC69KnfLsC1wEWllE93inq/i+G/B9rOusZVaHSVUpZSk/vA\nDX22tSJ/gdousWspZRr1Gjk6H5seETt23u9JvbxZRb32PaCUMq29ppbaMDmIJcCfdOqyN/US6GcD\njg9AKeX3pZTPUa/ND6TuSDMiYtoIHz+V2mB2aCllJ+BPh2c/wmdvp7bvTOu8diil3DhAtZYAB0VE\nd7oHteEb8k1gVkQc0h0YEbOolwfdBt/RumrfDpzbU//JPTtUPyNN98PAUuCpbd2dzsjrbSR3UZep\nu9/tST0r2GQRMZ0aGFeXUs7tKX7Udtb+vruUcm92PuMmNCJiv4g4NSL2aO9nAccANw0w+iTqjnoP\n8FBEzKU24vU6OyImRcTh1Ma1z7WjxiXABRHxR23eu7ej/SAWAke13/N3BP4ZuLKU8us2rQURsaDP\nMs+LiJdExJSI2KbV+wDg5lLKr6iXFhdFxPSI2DYihsNhCjXo1rafd981Sv0uBv6p/cpDREyNiFcO\nuGyLqY2ZJ0fEdhFxUhv+zU79l480YinlZ23eCyPiuRExodXhC8B1pZTrBqzDJcD8iDi0/YS74/A6\nG2Dcu4GdI2JqZ9gU4H5gKCL2o7aN9Y6zd5/p3Uw9ezitfR9zgKOol1ubpP16+DXghlLKO0b4yGXA\niRGxfzuQnEE9qKaNm9CgnmYdCtwcEeuoYfET6lF1VG0HPZnaTrEGeDW1PaRrZSu7i7qjz29nMwBv\npzZw3dROWa+jHskBaPdQHN5n3kuolxULqQ12U4C3dD4yC7ihT9Xvpx7pbqM2tr4HeHMpZfgmtuOo\njb9L27Tf2oa/D9iBepZ0E/VyakSllC8C/wpc0ZbtJ9SG5uFluyYiTu8z7u+oDXLHt/q9Dji6Dd/Q\nsgGcBHyM+pPtUKvnYmoj7kBKKT+gNn5/kPr9LaO2bw0y7lJqm8Mv2qXNbtRfWl5N3d4uoTY0d50F\nXNo+/7c90/sdNSTmUtf9RcDxne1oVKOta+CvgOcAr23b2/Brzzbvr1K3j29Rt5cVjH6w6F+P1iii\nUbQjwuWtrWRLzncStdX7oFLK77fkvLeEiLiW2th8y1jXRYObONYVUH/tyPT0sa7H5lJKGekSUI9z\n4+nyRNIW4OWJpBTPNCSlGBqSUgyNx7GIWB61S/hQRNzd7tlI3SHYM71ntG7R+3SGPav9PLjXgNM4\nJyL+JyIeioizkvNfHBG/acuzKiKujIh+d31uaFoREddHxLt6hh8fEbdGxOQBpnFBRCyLiF9HxC0R\ncWynbNeoj1m4t62fGyPisI2p63hjaDz+HdXuLn0mtVPaGRs7oVLKj6j3K1zSdrptgU8A7yylLB9w\nMsuA06id4TbGSW15nka99f2CjZlIqY1xrwdO6dx4tgvwXmq/mQdGG78ZonZUnEq9h+RDnTtQ76f2\nfN6F2uHvvcDV8cTtev8wQ2MrUUq5k3qH54EAEbFbRFwdta/Msoh4w/BnI+KQiPhBO6u4OyLO70zq\nbGrfjTdSbwwbogbJoPW4tJRyDRvRZ6FnOqupd3cOL8/UiLgs6kNkVkTEGdFut46IfSLi21EfrrQq\n2oOG2l2j5wIfb599P/CFUsq3BqzDmaWUn5ZS1pdSvkvtNn9YK3twuIx6m/h6ah+Tqf2n+MTgfRpb\niai3xb8YuLINuoJ6d+ZuwH7A1yPi1lLKN4ELgQtLKf/eLmce7n9TSvltRJxIPVPYhvq8ivWd+VzU\nPte9K3VzLM9M6p2dP2qDPkDdIfemdt2/lvq8io8D57T3f0a95f/ZnUmdD7wC+HwbfkBnHscBp5RS\nnjlAfSa38c/vGb6EelY0kdr7dXVyUcefbLdYX1vuRe22PUS9BXsF9bbjHai3X/+BzoN6gPOABe3v\n66lnFDP7THcq9TbmGzahbpczQBf3nnEWU/terKV20lpIPf2fQH3w0P6dz74JWNz+vgz4KJ0HBfVM\n9wBq57KXb+SyRFueRX3Kt6c+guC4sd4mHg8vL08e/44utWfm7FLKW0opD1LPLlaX1qmt6XaxPpF6\ndFwaEd+PiJf2TPO91AfZ7BERf7e5F6DHyW15di+lHFtKuYd62r8tdRmGdZfnNOqO/b2IWBIRr+tO\nsNT+OzBY79mRnE9dX8eMVFhK+U0pZSFw5nD7yROZlydbp7uo3d6ndILj4S7WpZSfA8e06/y/Bj4f\nETuXUtZFxJHAy4D9qU+I+mREXFvG9rR7FbVj3Wzgf9uw7vKspHY6IyKeD1wXEdeXUpZt6owj4lzq\no+/m9ITwSCZRL582NpzGBc80tkKllNupjXbnRcT2EXEQ9ezicoCIeE1E7FJqW8XaNtr6qF3vP0q9\nzl9VSvkK8HUSv2BE7dK9PXXbmdjmP6GV7RX1adx7JZfnD9QexudG7eY/G/iHzvK8MtojD6g9VQu1\nYXKTRP1vDl4BvKg3NCPisIh4XlveHVrv0hnA9zd1vlu9sb4+8tX/RW3TOLJP2R7UZ0aupj7KbX6n\n7HJqV/gh6lHx6Db8QuArPdOZ2T77ovb+Ytrj7vrMdwF1p+2+5rWyw1udt+0z7mI6DyzuKZve6n0P\n9cE576Q9Fo/apfvOtjy3Am8cYfxHPWavDTsB+HGf+U1o4/yWRx4cPASc1sqPoD7QaKit42/R5zmv\nT7SXfU/0mImIM4B7SikfGeu6aPMxNCSl2KYhKcXQkJRiaEhKMTQkpWyVN3dFhK230mZWShnx/3Px\nTENSiqEhKcXQkJRiaEhKMTQkpRgaklIMDUkphoakFENDUoqhISnF0JCUYmhISjE0JKUYGpJSDA1J\nKYaGpBRDQ1KKoSEpxdCQlGJoSEoxNCSlGBqSUgwNSSmGhqQUQ0NSiqEhKcXQkJRiaEhKMTQkpRga\nklIMDUkphoakFENDUoqhISnF0JCUYmhISjE0JKUYGpJSDA1JKYaGpBRDQ1KKoSEpxdCQlGJoSEox\nNCSlGBqSUgwNSSmGhqQUQ0NSiqEhKcXQkJRiaEhKMTQkpRgaklIMDUkphoakFENDUoqhISnF0JCU\nYmhISjE0JKUYGpJSDA1JKYaGpBRDQ1KKoSEpxdCQlGJoSEoxNCSlGBqSUgwNSSmGhqQUQ0NSiqEh\nKcXQkJRiaEhKMTQkpRgaklIMDUkphoakFENDUoqhISnF0JCUYmhISjE0JKUYGpJSDA1JKYaGpBRD\nQ1KKoSEpxdCQlGJoSEoxNCSlGBqSUgwNSSmGhqQUQ0NSiqEhKcXQkJRiaEhKMTQkpRgaklIMDUkp\nhoakFENDUoqhISnF0JCUYmhISjE0JKUYGpJSDA1JKYaGpBRDQ1KKoSEpxdCQlGJoSEoxNCSlTBzr\nCoxXpZSxrgIAETHWVdA445mGpBRDQ1KKoSEpxdCQlGJoSEoxNCSlGBqSUgwNSSmGhqQUQ0NSiqEh\nKcXQkJRiaEhKMTQkpRgaklIMDUkphoakFENDUoqhISnF0JCUYmhISjE0JKUYGpJSDA1JKYaGpBRD\nQ1KK/y3jZtL97xAfL/9Fo/RY8ExDUoqhISnF0JCUYpvGVqLbRiKNJc80JKUYGpJSvDzZAry00Hji\nmYakFENDUoqhISnF0JCUYmhISjE0JKUYGpJSDA1JKYaGpBRDQ1KKoSEpxdCQlGJoSEoxNCSlGBqS\nUgwNSSmGhqQUQ0NSiqEhKcXQkJRiaEhKMTQkpRgaklIMDUkphoakFENDUoqhISnF0JCUYmhISjE0\nJKUYGpJSDA1JKYaGpBRDQ1KKoSEpxdCQlGJoSEoxNCSlGBqSUgwNSSmGhqQUQ0NSiqEhKcXQkJRi\naEhKMTQkpRgaklIMDUkphoakFENDUoqhISnF0JCUYmhISjE0JKUYGpJSDA1JKYaGpBRDQ1KKoSEp\nxdCQlGJoSEoxNCSlGBqSUgwNSSlRShnrOkjainimISnF0JCUYmhISjE0JKUYGpJSDA1JKYaGpBRD\nQ1KKoSEpxdCQlGJoSEoxNCSlGBqSUgwNSSmGhqQUQ0NSiqEhKcXQkJRiaEhKMTQkpRgaklIMDUkp\nhoakFENDUoqhISnF0JCUYmhISjE0JKUYGpJSDA1JKYaGpBRDQ1KKoSEpxdCQlGJoSEoxNCSlGBqS\nUgwNSSn/D1nKrt3W//HwAAAAAElFTkSuQmCC\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "OkB6V4LjESxV", + "colab_type": "code", + "outputId": "a923d9d6-233e-4066-e9ba-f60200c1a05a", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 35 + } + }, + "source": [ + "label_dims = vae.label_shape\n", + "label_dim_offsets = np.cumsum(label_dims)\n", + "label_dim_offsets" + ], + "execution_count": 0, + "outputs": [ + { + "output_type": "execute_result", + "data": { + "text/plain": [ + "array([ 1, 4, 10, 50, 82, 114])" + ] + }, + "metadata": { + "tags": [] + }, + "execution_count": 17 + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "VXhaIdcBEHnS", + "colab_type": "code", + "colab": {} + }, + "source": [ + "class SCM():\n", + " \"\"\"\n", + " Structural causal model\n", + " \n", + " args: \n", + " vae: instance of vae\n", + " mu: loc of q(z|x) given by the vae encoder\n", + " sigma: scale of q(z|x) given by the vae encoder\n", + " \n", + " \"\"\"\n", + " def __init__(self, vae, mu, sigma):\n", + " \"\"\"\n", + " Constructor\n", + " \n", + " Intializes :\n", + " image dimensions - 4096(64*64), \n", + " z dimensions: size of the tensor representing the latent random variable z, \n", + " label dimensions: 114 labels y that correspond to an image(one hot encoded)\n", + " f(x) = p(x|y,z)\n", + " Noise variables in the model N_#\n", + " \"\"\"\n", + " self.vae = vae\n", + " self.image_dim = vae.image_dim\n", + " self.z_dim = vae.z_dim\n", + " # these are used for f_X\n", + " self.label_dims = vae.label_shape\n", + " \n", + " def f_X(Y, Z, N):\n", + " \"\"\"\n", + " Generating one hots for the factors\n", + " \"\"\" \n", + " zs = Z.cuda()\n", + " # convert the labels to one hot\n", + " ys = [torch.tensor([0])]\n", + " ys.append(torch.nn.functional.one_hot(torch.round(Y[0]).to(torch.long), int(self.label_dims[1])))\n", + " ys.append(torch.nn.functional.one_hot(torch.round(Y[1]).to(torch.long), int(self.label_dims[2])))\n", + " ys.append(torch.nn.functional.one_hot(torch.round(Y[2]).to(torch.long), int(self.label_dims[3])))\n", + " ys.append(torch.nn.functional.one_hot(torch.round(Y[3]).to(torch.long), int(self.label_dims[4])))\n", + " ys.append(torch.nn.functional.one_hot(torch.round(Y[4]).to(torch.long), int(self.label_dims[5])))\n", + " ys = torch.cat(ys).to(torch.float32).reshape(1,-1).cuda()\n", + " p = vae.decoder.forward(zs, ys)\n", + " return (N < p.cpu()).type(torch.float)\n", + " \n", + " def f_Y(N):\n", + " \"\"\"\n", + " Gumbel distribution - to model the distribution of the maximum of a number of samples\n", + " m = Gumbel(torch.tensor([1.0]), torch.tensor([2.0])).sample() # sample from Gumbel distribution with loc=1, scale=2\n", + " tensor([ 1.0124])\n", + " \n", + " https://pytorch.org/docs/stable/_modules/torch/distributions/gumbel.html\n", + " \"\"\"\n", + "# m = torch.distributions.gumbel.Gumbel(torch.zeros(N.size(0)), torch.ones(N.size(0)))\n", + " beta = 12\n", + " indices = torch.tensor(np.arange(N.size(0))).to(torch.float32)\n", + " smax = nn.functional.softmax(beta*N)\n", + " argmax_ind = torch.sum(smax*indices)\n", + " return argmax_ind\n", + " \n", + " def f_Z(N):\n", + " \"\"\"\n", + " Z ~ Normal(mu, sigma) \n", + " \"\"\"\n", + " return N * sigma + mu\n", + " \n", + " def model(noise): \n", + " \"\"\"\n", + " The model corresponds to a generative process\n", + " \n", + " args: noise variables\n", + " return: X(image), Y(labels), Z(latents) \n", + " \"\"\"\n", + " N_X = pyro.sample( 'N_X', noise['N_X'].to_event(1) )\n", + " # denoted using the index in the sequence \n", + " # that they are stored in as vae.label_names:\n", + " # ['shape', 'scale', 'orientation', 'posX', 'posY']\n", + " N_Y_1 = pyro.sample( 'N_Y_1', noise['N_Y_1'].to_event(1) )\n", + " N_Y_2 = pyro.sample( 'N_Y_2', noise['N_Y_2'].to_event(1) )\n", + " N_Y_3 = pyro.sample( 'N_Y_3', noise['N_Y_3'].to_event(1) )\n", + " N_Y_4 = pyro.sample( 'N_Y_4', noise['N_Y_4'].to_event(1) )\n", + " N_Y_5 = pyro.sample( 'N_Y_5', noise['N_Y_5'].to_event(1) )\n", + " \n", + " # Z ~ Normal(Nx_mu, Nx_sigma) \n", + " N_Z = pyro.sample( 'N_Z', noise['N_Z'].to_event(1) )\n", + " Z = pyro.sample('Z', dist.Normal( f_Z( N_Z ), 1e-1).to_event(1) )\n", + " \n", + " # Y ~ Gumbel max of Ny \n", + "# Y_1_mu = f_Y(N_Y_1)\n", + "# Y_2_mu = f_Y(N_Y_2)\n", + "# Y_3_mu = f_Y(N_Y_3)\n", + "# Y_4_mu = f_Y(N_Y_4)\n", + "# Y_5_mu = f_Y(N_Y_5)\n", + " \n", + " Y_1 = pyro.sample('Y_1', dist.Normal( f_Y(N_Y_1), 1e-2) )\n", + " Y_2 = pyro.sample('Y_2', dist.Normal( f_Y(N_Y_2), 1e-1) )\n", + " Y_3 = pyro.sample('Y_3', dist.Normal( f_Y(N_Y_3), 1e-1) )\n", + " Y_4 = pyro.sample('Y_4', dist.Normal( f_Y(N_Y_4), 1e-1) )\n", + " Y_5 = pyro.sample('Y_5', dist.Normal( f_Y(N_Y_5), 1e-1) )\n", + " \n", + "# Y_mu = (Y_1_mu, Y_2_mu, Y_3_mu, Y_4_mu, Y_5_mu)\n", + " \n", + " # X ~ p(x|y,z) = bernoulli(loc(y,z)) \n", + " X = pyro.sample('X', dist.Normal( f_X( (Y_1, Y_2, Y_3,Y_4,Y_5), Z, N_X ), 1e-2).to_event(1))\n", + " \n", + " # return noise and variables\n", + " noise_samples = N_X, (N_Y_1, N_Y_2, N_Y_3, N_Y_4, N_Y_5), N_Z\n", + " variable_samples = X, (Y_1, Y_2, Y_3, Y_4, Y_5), Z\n", + " return variable_samples, noise_samples\n", + " \n", + " self.model = model\n", + " #Initialize all noise variables in the model \n", + " self.init_noise = {\n", + " 'N_X' : dist.Uniform(torch.zeros(vae.image_dim), torch.ones(vae.image_dim)),\n", + " 'N_Z' : dist.Normal(torch.zeros(vae.z_dim), torch.ones(vae.z_dim)),\n", + " 'N_Y_1' : dist.Uniform(torch.zeros(self.label_dims[1]),torch.ones(self.label_dims[1])),\n", + " 'N_Y_2' : dist.Uniform(torch.zeros(self.label_dims[2]),torch.ones(self.label_dims[2])),\n", + " 'N_Y_3' : dist.Uniform(torch.zeros(self.label_dims[3]),torch.ones(self.label_dims[3])),\n", + " 'N_Y_4' : dist.Uniform(torch.zeros(self.label_dims[4]),torch.ones(self.label_dims[4])),\n", + " 'N_Y_5' : dist.Uniform(torch.zeros(self.label_dims[5]),torch.ones(self.label_dims[5])) \n", + " }\n", + " \n", + " def update_noise_svi(self, obs_data, intervened_model=None):\n", + " \"\"\"\n", + " Use svi to find out the mu, sigma of the distributionsfor the \n", + " condition outlined in obs_data\n", + " \"\"\"\n", + " \n", + " def guide(noise):\n", + " \"\"\"\n", + " The guide serves as an approximation to the posterior p(z|x). \n", + " The guide provides a valid joint probability density over all the \n", + " latent random variables in the model.\n", + " \n", + " https://pyro.ai/examples/svi_part_i.html\n", + " \"\"\"\n", + " # create params with constraints\n", + " mu = {\n", + " 'N_X': pyro.param('N_X_mu', 0.5*torch.ones(self.image_dim),constraint = constraints.interval(0., 1.)),\n", + " 'N_Z': pyro.param('N_Z_mu', torch.zeros(self.z_dim),constraint = constraints.interval(-3., 3.)),\n", + " 'N_Y_1': pyro.param('N_Y_1_mu', 0.5*torch.ones(self.label_dims[1]),constraint = constraints.interval(0., 1.)),\n", + " 'N_Y_2': pyro.param('N_Y_2_mu', 0.5*torch.ones(self.label_dims[2]),constraint = constraints.interval(0., 1.)),\n", + " 'N_Y_3': pyro.param('N_Y_3_mu', 0.5*torch.ones(self.label_dims[3]),constraint = constraints.interval(0., 1.)),\n", + " 'N_Y_4': pyro.param('N_Y_4_mu', 0.5*torch.ones(self.label_dims[4]),constraint = constraints.interval(0., 1.)),\n", + " 'N_Y_5': pyro.param('N_Y_5_mu', 0.5*torch.ones(self.label_dims[5]),constraint = constraints.interval(0., 1.))\n", + " }\n", + " sigma = {\n", + " 'N_X': pyro.param('N_X_sigma', 0.1*torch.ones(self.image_dim),constraint = constraints.interval(0.0001, 0.5)),\n", + " 'N_Z': pyro.param('N_Z_sigma', torch.ones(self.z_dim),constraint = constraints.interval(0.0001, 3.)),\n", + " 'N_Y_1': pyro.param('N_Y_1_sigma', 0.1*torch.ones(self.label_dims[1]),constraint = constraints.interval(0.0001, 0.5)),\n", + " 'N_Y_2': pyro.param('N_Y_2_sigma', 0.1*torch.ones(self.label_dims[2]),constraint = constraints.interval(0.0001, 0.5)),\n", + " 'N_Y_3': pyro.param('N_Y_3_sigma', 0.1*torch.ones(self.label_dims[3]),constraint = constraints.interval(0.0001, 0.5)),\n", + " 'N_Y_4': pyro.param('N_Y_4_sigma', 0.1*torch.ones(self.label_dims[4]),constraint = constraints.interval(0.0001, 0.5)),\n", + " 'N_Y_5': pyro.param('N_Y_5_sigma', 0.1*torch.ones(self.label_dims[5]),constraint = constraints.interval(0.0001, 0.5))\n", + " }\n", + " for noise_term in noise.keys():\n", + " pyro.sample(noise_term, dist.Normal(mu[noise_term], sigma[noise_term]).to_event(1))\n", + " \n", + " # Condition the model\n", + " if intervened_model is not None:\n", + " obs_model = pyro.condition(intervened_model, obs_data)\n", + " else:\n", + " obs_model = pyro.condition(self.model, obs_data)\n", + " \n", + " pyro.clear_param_store()\n", + "\n", + " # Once we’ve specified a guide, we’re ready to proceed to inference. \n", + " # Now, this an optimization problem where each iteration of training takes \n", + " # a step that moves the guide closer to the exact posterior \n", + " \n", + " # https://arxiv.org/pdf/1601.00670.pdf\n", + " svi = SVI(\n", + " model= obs_model,\n", + " guide= guide,\n", + " optim= SGD({\"lr\": 1e-5, 'momentum': 0.1}),\n", + " loss=Trace_ELBO(retain_graph=True)\n", + " )\n", + " \n", + " num_steps = 1500\n", + " samples = defaultdict(list)\n", + " for t in range(num_steps):\n", + " loss = svi.step(self.init_noise)\n", + "# if t % 100 == 0:\n", + "# print(\"step %d: loss of %.2f\" % (t, loss))\n", + " for noise in self.init_noise.keys():\n", + " mu = '{}_mu'.format(noise)\n", + " sigma = '{}_sigma'.format(noise)\n", + " samples[mu].append(pyro.param(mu).detach().numpy())\n", + " samples[sigma].append(pyro.param(sigma).detach().numpy())\n", + " means = {k: torch.tensor(np.array(v).mean(axis=0)) for k, v in samples.items()}\n", + " \n", + " # update the inferred noise\n", + " updated_noise = {\n", + " 'N_X' : dist.Normal(means['N_X_mu'], means['N_X_sigma']),\n", + " 'N_Z' : dist.Normal(means['N_Z_mu'], means['N_Z_sigma']),\n", + " 'N_Y_1': dist.Normal(means['N_Y_1_mu'], means['N_Y_1_sigma']),\n", + " 'N_Y_2': dist.Normal(means['N_Y_2_mu'], means['N_Y_2_sigma']),\n", + " 'N_Y_3': dist.Normal(means['N_Y_3_mu'], means['N_Y_3_sigma']),\n", + " 'N_Y_4': dist.Normal(means['N_Y_4_mu'], means['N_Y_4_sigma']),\n", + " 'N_Y_5': dist.Normal(means['N_Y_5_mu'], means['N_Y_5_sigma']),\n", + " }\n", + " return updated_noise\n", + " \n", + " def __call__(self):\n", + " return self.model(self.init_noise)" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "lk7vAe8byP5G", + "colab_type": "text" + }, + "source": [ + "## Sanity check: 1\n", + "### Making sure VAE works" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "TDI4enwfEPCY", + "colab_type": "code", + "outputId": "921d5efc-4d33-4d7f-e167-a60e70e848f8", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 303 + } + }, + "source": [ + "# Generate an instance of dSprites image \n", + "ox, y = get_specific_data(cuda=True)\n", + "plot_image(ox)\n", + "# Pass it through VAE to get q(z|x) => N(mu, sigma)\n", + "mu, sigma = vae.encoder.forward(ox,vae.remap_y(y))\n", + "# Feed these params to our custom SCM\n", + "scm = SCM(vae, mu.cpu(), sigma.cpu())\n", + "print(y)\n", + "# Check for reconstruction\n" + ], + "execution_count": 392, + "outputs": [ + { + "output_type": "stream", + "text": [ + "tensor([[ 0., 2., 1., 19., 13., 10.]], device='cuda:0')\n" + ], + "name": "stdout" + }, + { + "output_type": "display_data", + "data": { + "image/png": "iVBORw0KGgoAAAANSUhEUgAAAQ0AAAENCAYAAAAVEjAIAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "B1rkIEyL-357", + "colab_type": "text" + }, + "source": [ + "## Sanity check 2\n", + "\n", + "### To check if the decoder is able to generate the image if the latents are changed:\n", + "#### To achieve this we manually change the labels in the code and run it through the decoder and check for reconstruction" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "1CcO36iqEf_T", + "colab_type": "code", + "outputId": "8dc36b9f-0ee7-44e2-ba86-8299afa4c6f4", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 492 + } + }, + "source": [ + "original, y_original = get_specific_data(cuda=True)\n", + "print('top: ',y_original)\n", + "mu, sigma = vae.encoder.forward(original,vae.remap_y(y_original))\n", + "B = 100\n", + "zs = torch.cat([dist.Normal(mu.cpu(), sigma.cpu()).sample() for a in range(B)], 0)\n", + "ys = torch.cat([vae.remap_y(y_original) for a in range(B)], 0)\n", + "rs = vae.decoder.forward(zs.cuda(), ys).detach()\n", + "compare_to_density(original,rs)\n", + "\n", + "y_new = torch.tensor(y_original)\n", + "y_new[0,1] = (y_original[0,1] + 1) % 2\n", + "print('bottom: ', y_new)\n", + "zs = torch.cat([dist.Normal(mu.cpu(), sigma.cpu()).sample() for a in range(B)], 0)\n", + "ys = torch.cat([vae.remap_y(y_new) for a in range(B)], 0)\n", + "rs = vae.decoder.forward(zs.cuda(), ys).detach()\n", + "compare_to_density(original,rs)" + ], + "execution_count": 393, + "outputs": [ + { + "output_type": "stream", + "text": [ + "top: tensor([[ 0., 0., 2., 16., 31., 5.]], device='cuda:0')\n", + "bottom: tensor([[ 0., 1., 2., 16., 31., 5.]], device='cuda:0')\n" + ], + "name": "stdout" + }, + { + "output_type": "display_data", + "data": { + "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAADcCAYAAACBHI1wAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAIT+H7Ibd/Pt/SDNAAAAAElFTkSuQmCC\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + }, + { + "output_type": "display_data", + "data": { + "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAADcCAYAAACBHI1wAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nsAAAhAgLAEDo/wF7mAhYszWNWQAAAABJRU5ErkJggg==\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "bwyk5S5a7knN", + "colab_type": "text" + }, + "source": [ + "## Conditioning the model" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "tQu4lzTMp8QH", + "colab_type": "code", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 35 + }, + "outputId": "6a85b4b4-9af1-4097-eb7f-6adc13f1431d" + }, + "source": [ + "cond_data = {}\n", + "for i in range(1, 6):\n", + " cond_data[\"Y_{}\".format(i)] = torch.tensor(y[0,i].cpu()).to(torch.float32)\n", + "print(cond_data)" + ], + "execution_count": 394, + "outputs": [ + { + "output_type": "stream", + "text": [ + "{'Y_1': tensor(2.), 'Y_2': tensor(1.), 'Y_3': tensor(19.), 'Y_4': tensor(13.), 'Y_5': tensor(10.)}\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "rtJRtQgJve2A", + "colab_type": "code", + "outputId": "93aa7f5c-97bc-434b-c564-18ffebbe3b5a", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 35 + } + }, + "source": [ + "# cond_data['Y_1'] = torch.tensor(1.)\n", + "# cond_data['Y_2'] = torch.tensor(4.)\n", + "conditioned_model = pyro.condition(scm.model, data=cond_data)\n", + "cond_noise = scm.update_noise_svi(cond_data)\n", + "print(cond_data)" + ], + "execution_count": 395, + "outputs": [ + { + "output_type": "stream", + "text": [ + "{'Y_1': tensor(2.), 'Y_2': tensor(1.), 'Y_3': tensor(19.), 'Y_4': tensor(13.), 'Y_5': tensor(10.)}\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "0r2HePhs9DnB", + "colab_type": "code", + "outputId": "a95abb00-a531-4b8c-ca29-501c340e38d2", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 277 + } + }, + "source": [ + "rxs = []\n", + "for i in range(100):\n", + " (rx,ry,_), _ = scm.model(cond_noise)\n", + " rxs.append(rx)\n", + "compare_to_density(ox, torch.cat(rxs))\n", + "_ =plt.suptitle(\"SCM Conditioned on Original\", fontsize=18, fontstyle='italic')\n" + ], + "execution_count": 396, + "outputs": [ + { + "output_type": "display_data", + "data": { + "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEDCAYAAADEAyg+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "T9q5R5Az7qUO", + "colab_type": "text" + }, + "source": [ + "## Counterfactuals" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "4NA_mpKv7t4P", + "colab_type": "code", + "colab": {} + }, + "source": [ + "# intervening on Shape, posX and PosY \n", + "intervened_model = pyro.do(scm.model, data={\n", + " \"Y_1\": torch.tensor(0.),\n", + "# \"Y_4\": torch.tensor(5.),\n", + "# \"Y_5\": torch.tensor(25.),\n", + "})\n", + "noise_data = {}\n", + "for term, d in cond_noise.items():\n", + " noise_data[term] = d.loc\n", + "# intervened_noise = scm.update_noise_svi(noise_data, intervened_model)" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "rffN8m75cKrd", + "colab_type": "code", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 254 + }, + "outputId": "93cf401b-156a-4bb7-9796-30096fa82c46" + }, + "source": [ + "(rx1,ry,_), _ = intervened_model(scm.init_noise)\n", + "compare_to_density(ox, rx1)\n", + "print(ry)" + ], + "execution_count": 434, + "outputs": [ + { + "output_type": "stream", + "text": [ + "(tensor(0.), tensor(0.9218), tensor(15.6666), tensor(21.0625), tensor(18.7714))\n" + ], + "name": "stdout" + }, + { + "output_type": "display_data", + "data": { + "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAADcCAYAAACBHI1wAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "WwvqXamE4hqb", + "colab_type": "code", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 276 + }, + "outputId": "364feed3-eb07-427f-964f-88c5b07bc240" + }, + "source": [ + "rxs = []\n", + "for i in range(100):\n", + " (cfo1,ny1,nz1), _= intervened_model(cond_noise)\n", + " rxs.append(cfo1)\n", + "compare_to_density(ox, torch.cat(rxs))\n", + "_ =plt.suptitle(\"SCM intervened on shape\", fontsize=18, fontstyle='italic')" + ], + "execution_count": 435, + "outputs": [ + { + "output_type": "display_data", + "data": { + "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEDCAYAAADEAyg+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n+R9HuLwlzlJD7gAAAABJRU5ErkJggg==\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "MR4jlBojGdr1", + "colab_type": "text" + }, + "source": [ + "## ======== Test code =======" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "HNzpNC4EzkA6", + "colab_type": "code", + "outputId": "4c161478-3e67-478b-b2f8-e077615cfb52", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 52 + } + }, + "source": [ + "import warnings\n", + "warnings.filterwarnings(\"ignore\")\n", + "\n", + "xs, ys = next(data_iter)\n", + "x = xs[0].reshape(1,-1).cuda()\n", + "y = ys[0].reshape(1,-1).cuda()\n", + "mu, sigma = vae.encoder.forward(x,vae.remap_y(y))\n", + "mu = mu.cpu()\n", + "sigma = sigma.cpu()\n", + "recon_x1, y1, z1 = OldSCM(vae, mu, sigma)\n", + "print(y1)" + ], + "execution_count": 0, + "outputs": [ + { + "output_type": "stream", + "text": [ + "tensor([ 1.9037e-04, 9.9998e-01, 4.9999e+00, -1.2039e-04, 2.1000e+01,\n", + " 4.0000e+00])\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "sLOw_NXQziDa", + "colab_type": "code", + "colab": {} + }, + "source": [ + "def OldSCM(vae, mu, sigma):\n", + " z_dim = vae.z_dim\n", + " Ny, Y, ys = [], [], []\n", + " Nx = pyro.sample(\"Nx\", dist.Uniform(torch.zeros(vae.image_dim), torch.ones(vae.image_dim)))\n", + " Nz = pyro.sample(\"Nz\", dist.Normal(torch.zeros(z_dim), torch.ones(z_dim)))\n", + " m = torch.distributions.gumbel.Gumbel(torch.tensor(0.0), torch.tensor(1.0))\n", + " for label_id in range(6):\n", + " name = vae.label_names[label_id]\n", + " length = vae.label_shape[label_id]\n", + " new = pyro.sample(\"Ny_%s\"%name, dist.Uniform(torch.zeros(length), torch.ones(length)) )\n", + " Ny.append(new)\n", + " gumbel_vars = torch.tensor([m.sample() for _ in range(length)])\n", + " max_ind = torch.argmax(torch.log(new) + gumbel_vars).item()\n", + " Y.append(pyro.sample(\"Y_%s\"%name, dist.Normal(torch.tensor(max_ind * 1.0), 1e-4)))\n", + "# Y.append(pyro.sample(\"Y_%s\"%name, dist.Delta(torch.tensor(max_ind*1.0))))\n", + " ys.append(torch.nn.functional.one_hot(torch.tensor(max_ind), int(length))) \n", + " Y = torch.tensor(Y)\n", + " ys = torch.cat(ys).to(torch.float32).reshape(1,-1).cuda()\n", + " Z = pyro.sample(\"Z\", dist.Normal(mu + Nz*sigma, 1e-4))\n", + "# Z = pyro.sample(\"Z\", dist.Delta(mu + Nz*sigma))\n", + " zs = Z.cuda()\n", + " p = vae.decoder.forward(zs,ys)\n", + " X = pyro.sample(\"X\", dist.Normal((Nx < p.cpu()).type(torch.float), 1e-4))\n", + "# X = pyro.sample(\"X\", dist.Delta((Nx < p.cpu()).type(torch.float)))\n", + " return X, Y, Z" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "_0EX3FetiTi-", + "colab_type": "code", + "outputId": "3007827d-15f3-4f55-f959-081e2122f249", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 255 + } + }, + "source": [ + "from pyro.infer.importance import Importance\n", + "from pyro.infer.mcmc import MCMC\n", + "from pyro.infer.mcmc.nuts import HMC\n", + "from pyro.infer import SVI\n", + "\n", + "intervened_model = pyro.do(OldSCM, data={\"Y_shape\": torch.tensor(0.)})\n", + "conditioned_model = pyro.condition(OldSCM, data={\n", + " \"X\": recon_x1,\n", + "# \"Z\": z1,\n", + " \"Y_color\":y1[0],\n", + " \"Y_shape\": y1[1],\n", + " \"Y_scale\": y1[2],\n", + " \"Y_orientation\": y1[3],\n", + " \"Y_posX\": y1[4],\n", + " \"Y_posY\": y1[5]\n", + " })\n", + "\n", + "# change to svi\n", + "posterior = pyro.infer.Importance(conditioned_model, num_samples = 100)\n", + "posterior.run(vae, mu, sigma)\n", + "\n", + "result = []\n", + "for i in range(500):\n", + " trace = posterior()\n", + " x = trace.nodes['Nx']['value']\n", + " ny_shape = trace.nodes['Ny_shape']['value']\n", + " ny_scale = trace.nodes['Ny_scale']['value']\n", + " ny_orientation = trace.nodes['Ny_orientation']['value']\n", + " ny_posX = trace.nodes['Ny_posX']['value']\n", + " ny_posY = trace.nodes['Ny_posY']['value']\n", + " z = trace.nodes['Nz']['value']\n", + " con_obj = pyro.condition(intervened_model, data = {\n", + " \"Nx\": x,\n", + " \"Ny_shape\": ny_shape, \n", + " \"Ny_scale\": ny_scale, \n", + " \"Ny_orientation\": ny_orientation, \n", + " \"Ny_posX\": ny_posX, \n", + " \"Ny_posY\": ny_posY, \n", + " \"Nz\": z\n", + " })\n", + " \n", + "recon_x2,y2,z2 = con_obj(vae, mu, sigma)\n", + "print(y2)\n", + "recon_check(recon_x1.reshape(-1, 64, 64)[0], recon_x2.reshape(-1, 64, 64)[0])" + ], + "execution_count": 0, + "outputs": [ + { + "output_type": "stream", + "text": [ + "tensor([-9.3067e-05, 0.0000e+00, 2.0000e+00, 3.9999e+00, 2.1000e+01,\n", + " 1.6000e+01])\n" + ], + "name": "stdout" + }, + { + "output_type": "display_data", + "data": { + "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAADMCAYAAAB+36QhAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRK5CYII=\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "Y1MuTqXb2in_", + "colab_type": "code", + "colab": {} + }, + "source": [ + "result = []\n", + "for i in range(500):\n", + " trace = posterior()\n", + " x = trace.nodes['Nx']['value']\n", + " ny_shape = trace.nodes['Ny_shape']['value']\n", + " ny_scale = trace.nodes['Ny_scale']['value']\n", + " ny_orientation = trace.nodes['Ny_orientation']['value']\n", + " ny_posX = trace.nodes['Ny_posX']['value']\n", + " ny_posY = trace.nodes['Ny_posY']['value']\n", + " z = trace.nodes['Nz']['value']\n", + " con_obj = pyro.condition(intervened_model, data = {\n", + " \"Nx\": x,\n", + " \"Ny_shape\": ny_shape, \n", + " \"Ny_scale\": ny_scale, \n", + " \"Ny_orientation\": ny_orientation, \n", + " \"Ny_posX\": ny_posX, \n", + " \"Ny_posY\": ny_posY, \n", + " \"Nz\": z\n", + " })\n", + " \n", + "recon_x2,y2,z2 = con_obj(vae, mu, sigma)\n", + "print(y2)\n", + "recon_check(recon_x1.reshape(-1, 64, 64)[0], recon_x2.reshape(-1, 64, 64)[0])" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "cou0AUQhejcz", + "colab_type": "code", + "colab": {} + }, + "source": [ + "vae.label_names" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "H7A27SBfl8aM", + "colab_type": "code", + "outputId": "feb97a2d-c831-436b-fab8-aa3f49052e4e", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 390 + } + }, + "source": [ + "adam_params = {\"lr\": 0.00042, \"betas\": (0.9, 0.999)}\n", + "optimizer = Adam(adam_params)\n", + "# set up the loss(es) for inference. wrapping the guide in config_enumerate builds\n", + "# the loss as a sum\n", + "# by enumerating each class label for the sampled discrete categorical distribution\n", + "# in the model\n", + "guide = config_enumerate(guide, \"parallel\", expand=True)\n", + "elbo = Trace_ELBO(max_plate_nesting=1)\n", + "loss_basic = SVI(sup_vae.model, guide, optimizer, loss=elbo)\n", + "\n", + "# build a list of all losses considered\n", + "losses = [loss_basic]\n", + "loss_aux = SVI(sup_vae.model_classify, sup_vae.guide_classify, optimizer, loss=elbo)\n", + "losses.append(loss_aux)\n", + "\n", + "\n", + "\n", + "from pyro.optim import Adam\n", + "from pyro.infer import SVI, Trace_ELBO\n", + "\n", + "adam = Adam({\"lr\": 0.005, \"betas\": (0.90, 0.999)})\n", + "svi = SVI(vae.model, vae.guide, adam, loss=Trace_ELBO())\n", + "\n", + "\n", + "param_vals = []\n", + "pbar = tqdm(range(5))\n", + "for _ in pbar:\n", + " xs, ys = next(data_iter)\n", + " xs = xs.cuda()\n", + " ys = ys.cuda()\n", + " svi.step(xs, ys)\n", + "# param_vals.append({k: param(k).item() for k in [\"fl\", \"ia\"]})\n", + "\n", + "# pd.DataFrame(param_vals).plot(subplots=True)" + ], + "execution_count": 0, + "outputs": [ + { + "output_type": "stream", + "text": [ + "\n", + " 0%| | 0/5 [00:00\u001b[0m in \u001b[0;36m\u001b[0;34m()\u001b[0m\n\u001b[1;32m 9\u001b[0m \u001b[0mpbar\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtqdm\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m5\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 10\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0m_\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mpbar\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 11\u001b[0;31m \u001b[0mxs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mys\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnext\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdata_iter\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 12\u001b[0m \u001b[0mxs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mxs\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcuda\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 13\u001b[0m \u001b[0mys\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mys\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcuda\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", + "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/torch/utils/data/dataloader.py\u001b[0m in \u001b[0;36m__next__\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m 792\u001b[0m \u001b[0;31m# no valid `self.rcvd_idx` is found (i.e., didn't break)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 793\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_shutdown_workers\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 794\u001b[0;31m \u001b[0;32mraise\u001b[0m \u001b[0mStopIteration\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m 795\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m 796\u001b[0m \u001b[0;31m# Now `self.rcvd_idx` is the batch index we want to fetch\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n", + "\u001b[0;31mStopIteration\u001b[0m: " + ] + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "pa3ygX4zCYdH", + "colab_type": "code", + "colab": {} + }, + "source": [ + "" + ], + "execution_count": 0, + "outputs": [] + } + ] +} \ No newline at end of file diff --git a/causal_vae_dsprites.py b/causal_vae_dsprites.py new file mode 100644 index 0000000..13f2044 --- /dev/null +++ b/causal_vae_dsprites.py @@ -0,0 +1,972 @@ +# -*- coding: utf-8 -*- +"""causal_vae_dsprites.ipynb + +Automatically generated by Colaboratory. + +Original file is located at + https://colab.research.google.com/drive/17TusqgELDORkovcWZxzlyScvCa7CgJyP + +## Deep Causal Variational Inference + +### Introduction: +To train a supervised variational autoencoder using Deepmind's [dSprites](https://github.com/deepmind/dsprites-dataset) dataset. + +dSprites is a dataset of sprites, which are 2D shapes procedurally generated from 5 ground truth independent "factors." These factors are color, shape, scale, rotation, x and y positions of a sprite. + +All possible combinations of these variables are present exactly once, generating N = 737280 total images. + +Factors and their values: + +* Shape: 3 values {square, ellipse, heart} +* Scale: 6 values linearly spaced in (0.5, 1) +* Orientation: 40 values in (0, 2$\pi$) +* Position X: 32 values in (0, 1) +* Position Y: 32 values in (0, 1) + + +Further, the objective of any generative model is essentially to capture underlying data generative factors, the disentangled representation would mean a single latent unit being sensitive to variations in single generative factors + + +### Goal: +To include the latent factors as labels in the training and to invent a causal story that relates these factors and the images in a DAG. + +Reference + +[Structured Disentangled Representation](https://arxiv.org/pdf/1804.02086.pdf) +""" + +#Install dependencies +!pip3 install pyro-ppl +!pip3 install torch torchvision +!pip3 install pydrive --upgrade +!pip3 install tqdm + +# Commented out IPython magic to ensure Python compatibility. +# Load necessary libraries +from matplotlib import pyplot as plt +import numpy as np +import seaborn as sns + +import os +from collections import defaultdict + +import torch +import torch.nn as nn + +from tqdm import tqdm +import pyro +import pyro.distributions as dist +from pyro.infer import SVI, Trace_ELBO, TraceEnum_ELBO, config_enumerate, EmpiricalMarginal +from pyro.optim import Adam, SGD +import torch.distributions.constraints as constraints + +# Change figure aesthetics +# %matplotlib inline +sns.set_context('talk', font_scale=1.2, rc={'lines.linewidth': 1.5}) + +from ipywidgets import interact, interactive, fixed, interact_manual +import ipywidgets as widgets + +#to utilize GPU capabilities +USE_CUDA = True + +pyro.enable_validation(True) +pyro.distributions.enable_validation(False) + +# Mount Google drive to load data +from google.colab import drive +drive.mount('/content/gdrive') + +# Mount G drive to access files +from pydrive.auth import GoogleAuth +from pydrive.drive import GoogleDrive +from google.colab import auth +from oauth2client.client import GoogleCredentials + +auth.authenticate_user() +gauth = GoogleAuth() +gauth.credentials = GoogleCredentials.get_application_default() +drive = GoogleDrive(gauth) + +# Hack to get all available GPU ram. + +import tensorflow as tf +tf.test.gpu_device_name() + +!ln -sf /opt/bin/nvidia-smi /usr/bin/nvidia-smi +!pip install gputil +!pip install psutil +!pip install humanize +import psutil +import humanize +import os +import GPUtil as GPU +GPUs = GPU.getGPUs() +# XXX: only one GPU on Colab and isn’t guaranteed +gpu = GPUs[0] +def printm(): + process = psutil.Process(os.getpid()) + print("Gen RAM Free: " + humanize.naturalsize( psutil.virtual_memory().available ), " | Proc size: " + humanize.naturalsize( process.memory_info().rss)) + print("GPU RAM Free: {0:.0f}MB | Used: {1:.0f}MB | Util {2:3.0f}% | Total {3:.0f}MB".format(gpu.memoryFree, gpu.memoryUsed, gpu.memoryUtil*100, gpu.memoryTotal)) +printm() + +class Encoder(nn.Module): + """ + MLPs (multi-layered perceptrons or simple feed-forward networks) + where the provided activation parameter is used on every linear layer except + for the output layer where we use the provided output_activation parameter + """ + def __init__(self, image_dim, label_dim, z_dim): + super(Encoder, self).__init__() + #setup image and label dimensions from the dataset + self.image_dim = image_dim + self.label_dim = label_dim + self.z_dim = z_dim + # setup the three linear transformations used + self.fc1 = nn.Linear(self.image_dim+self.label_dim, 1000) + self.fc2 = nn.Linear(1000, 1000) + self.fc31 = nn.Linear(1000, z_dim) # mu values + self.fc32 = nn.Linear(1000, z_dim) # sigma values + # setup the non-linearities + self.softplus = nn.Softplus() + + def forward(self, xs, ys): + xs = xs.reshape(-1, self.image_dim) + #now concatenate the image and label + inputs = torch.cat((xs,ys), -1) + # then compute the hidden units + hidden1 = self.softplus(self.fc1(inputs)) + hidden2 = self.softplus(self.fc2(hidden1)) + # then return a mean vector and a (positive) square root covariance + # each of size batch_size x z_dim + z_loc = self.fc31(hidden2) + z_scale = torch.exp(self.fc32(hidden2)) + return z_loc, z_scale + + +class Decoder(nn.Module): + def __init__(self, image_dim, label_dim, z_dim): + super(Decoder, self).__init__() + # setup the two linear transformations used + hidden_dim = 1000 + self.fc1 = nn.Linear(z_dim+label_dim, hidden_dim) + self.fc2 = nn.Linear(hidden_dim, hidden_dim) + self.fc3 = nn.Linear(hidden_dim, hidden_dim) + self.fc4 = nn.Linear(hidden_dim, image_dim) + # setup the non-linearities + self.softplus = nn.Softplus() + self.sigmoid = nn.Sigmoid() + + def forward(self, zs, ys): + inputs = torch.cat((zs, ys),-1) + # then compute the hidden units + hidden1 = self.softplus(self.fc1(inputs)) + hidden2 = self.softplus(self.fc2(hidden1)) + hidden3 = self.softplus(self.fc3(hidden2)) + # return the parameter for the output Bernoulli + # each is of size batch_size x 784 + loc_img = self.sigmoid(self.fc4(hidden3)) + return loc_img + +class CVAE(nn.Module): + """ + This class encapsulates the parameters (neural networks) and models & guides + needed to train a supervised variational auto-encoder + """ + def __init__(self, config_enum=None, use_cuda=False, aux_loss_multiplier=None): + + super(CVAE, self).__init__() + self.image_dim = 64**2 + self.label_shape = np.array((1,3,6,40,32,32)) + self.label_names = np.array(('color', 'shape', 'scale', 'orientation', 'posX', 'posY')) + self.label_dim = np.sum(self.label_shape) + self.z_dim = 50 + self.allow_broadcast = config_enum == 'parallel' + self.use_cuda = use_cuda + self.aux_loss_multiplier = aux_loss_multiplier + # define and instantiate the neural networks representing + # the paramters of various distributions in the model + self.setup_networks() + + def setup_networks(self): + """ + Setup and initialize Encoder and decoder units + """ + self.encoder = Encoder(self.image_dim, self.label_dim, self.z_dim) + self.decoder = Decoder(self.image_dim, self.label_dim, self.z_dim) + # using GPUs for faster training of the networks + if self.use_cuda: + self.cuda() + + def model(self, xs, ys): + pyro.module("cvae", self) + batch_size = xs.size(0) + options = dict(dtype=xs.dtype, device=xs.device) + with pyro.plate("data"): + prior_loc = torch.zeros(batch_size, self.z_dim, **options) + prior_scale = torch.ones(batch_size, self.z_dim, **options) + zs = pyro.sample("z", dist.Normal(prior_loc, prior_scale).to_event(1)) + # if the label y (which digit to write) is supervised, sample from the + # constant prior, otherwise, observe the value (i.e. score it against the constant prior) + loc = self.decoder.forward(zs, self.remap_y(ys)) + pyro.sample("x", dist.Bernoulli(loc).to_event(1), obs=xs) + # return the loc so we can visualize it later + return loc + + def guide(self, xs, ys): + with pyro.plate("data"): + # sample (and score) the latent handwriting-style with the variational + # distribution q(z|x) = normal(loc(x),scale(x)) + loc, scale = self.encoder.forward(xs, self.remap_y(ys)) + pyro.sample("z", dist.Normal(loc, scale).to_event(1)) + + def remap_y(self, ys): + new_ys = [] + options = dict(dtype=ys.dtype, device=ys.device) + for i, label_length in enumerate(self.label_shape): + prior = torch.ones(ys.size(0), label_length, **options) / (1.0 * label_length) + new_ys.append(pyro.sample("y_%s" % self.label_names[i], dist.OneHotCategorical(prior), + obs=torch.nn.functional.one_hot(ys[:,i].to(torch.int64), int(label_length)))) + new_ys = torch.cat(new_ys, -1) + return new_ys.to(torch.float32) + + def reconstruct_image(self, xs, ys): + # backward + sim_z_loc, sim_z_scale = self.encoder.forward(xs, self.remap_y(ys)) + zs = dist.Normal(sim_z_loc, sim_z_scale).to_event(1).sample() + # forward + loc = self.decoder.forward(zs, self.remap_y(ys)) + return dist.Bernoulli(loc).to_event(1).sample() + +def setup_data_loaders(train_x, test_x, train_y, test_y, batch_size=128, use_cuda=False): + train_dset = torch.utils.data.TensorDataset( + torch.from_numpy(train_x.astype(np.float32)).reshape(-1, 4096), + torch.from_numpy(train_y.astype(np.float32)) + ) + test_dset = torch.utils.data.TensorDataset( + torch.from_numpy(test_x.astype(np.float32)).reshape(-1, 4096), + torch.from_numpy(test_y.astype(np.float32)) + ) + kwargs = {'num_workers': 1, 'pin_memory': use_cuda} + train_loader = torch.utils.data.DataLoader( + dataset=train_dset, batch_size=batch_size, shuffle=False, **kwargs + ) + + test_loader = torch.utils.data.DataLoader( + dataset=test_dset, batch_size=batch_size, shuffle=False, **kwargs + ) + return {"train":train_loader, "test":test_loader} + +dataset_zip = np.load( + '/content/gdrive/My Drive/data-science/causal-ml/projects/dsprites-dataset/dsprites_ndarray_co1sh3sc6or40x32y32_64x64.npz', + encoding = 'bytes', + allow_pickle=True +) + +imgs = dataset_zip['imgs'] +labels = dataset_zip['latents_classes'] +label_sizes = dataset_zip['metadata'][()][b'latents_sizes'] +label_names = dataset_zip['metadata'][()][b'latents_names'] + +# Sample imgs randomly +indices_sampled = np.arange(imgs.shape[0]) +np.random.shuffle(indices_sampled) +imgs_sampled = imgs[indices_sampled] +labels_sampled = labels[indices_sampled] + +data_loaders = setup_data_loaders( + imgs_sampled[1000:], + imgs_sampled[:1000], + labels_sampled[1000:], + labels_sampled[:1000], + batch_size=256, + use_cuda=USE_CUDA +) + +def train(svi, train_loader, use_cuda=False): + # initialize loss accumulator + epoch_loss = 0. + # do a training epoch over each mini-batch x returned + # by the data loader + for xs,ys in train_loader: + # if on GPU put mini-batch into CUDA memory + if use_cuda: + xs = xs.cuda() + ys = ys.cuda() + # do ELBO gradient and accumulate loss + epoch_loss += svi.step(xs, ys) + # return epoch loss + normalizer_train = len(train_loader.dataset) + total_epoch_loss_train = epoch_loss / normalizer_train + return total_epoch_loss_train + +def evaluate(svi, test_loader, use_cuda=False): + # initialize loss accumulator + test_loss = 0. + # compute the loss over the entire test set + for xs, ys in test_loader: + # if on GPU put mini-batch into CUDA memory + if use_cuda: + xs = xs.cuda() + ys = ys.cuda() + # compute ELBO estimate and accumulate loss + test_loss += svi.evaluate_loss(xs, ys) + normalizer_test = len(test_loader.dataset) + total_epoch_loss_test = test_loss / normalizer_test + return total_epoch_loss_test + +# Run options +LEARNING_RATE = 1.0e-3 + +# Run only for a single iteration for testing +NUM_EPOCHS = 10 +TEST_FREQUENCY = 5 + +################################# +### FOR SAVING AND LOADING MODEL +################################ +# clear param store + +pyro.clear_param_store() + +network_path = "/content/gdrive/My Drive/data-science/causal-ml/projects/trained_model.save" + +PATH = "trained_model.save" + +# new model +# vae = CVAE(use_cuda=USE_CUDA) + +# save current model +# torch.save(vae.state_dict(), PATH) + +# to load params from trained model +vae = CVAE(use_cuda=USE_CUDA) +vae.load_state_dict(torch.load(network_path)) + +"""### **DONT RUN THE BELOW CODE AS WE'VE ALREADY TRAINED THE MODEL AND WE'VE STORED THE NETWORK PARAMS** + +## ================================================================================== +""" + +import warnings +warnings.filterwarnings('ignore') + +# clear param store +pyro.clear_param_store() + +# setup the VAE +vae = CVAE(use_cuda=USE_CUDA) + +# setup the optimizer +adam_args = {"lr": LEARNING_RATE} +optimizer = Adam(adam_args) + +# setup the inference algorithm +svi = SVI(vae.model, vae.guide, optimizer, loss=Trace_ELBO()) + +train_elbo = [] +test_elbo = [] +# training loop + +VERBOSE = True +pbar = tqdm(range(NUM_EPOCHS)) +for epoch in pbar: + total_epoch_loss_train = train(svi, data_loaders["train"], use_cuda=USE_CUDA) + train_elbo.append(-total_epoch_loss_train) + if VERBOSE: + print("[epoch %03d] average training loss: %.4f" % (epoch, total_epoch_loss_train)) + if epoch % TEST_FREQUENCY == 0: + # report test diagnostics + total_epoch_loss_test = evaluate(svi, data_loaders["test"], use_cuda=USE_CUDA) + test_elbo.append(-total_epoch_loss_test) + if VERBOSE: + print("[epoch %03d] average test loss: %.4f" % (epoch, total_epoch_loss_test)) + +"""## ================================================================================== + +## Visualizing the reconstruction accuracy of VAE +""" + + + +import warnings +warnings.filterwarnings("ignore") + +data_iter = iter(data_loaders["train"]) +xs, ys = next(data_iter) + +if USE_CUDA: + xs = xs.cuda() + ys = ys.cuda() +rs = vae.reconstruct_image(xs, ys) +if USE_CUDA: + xs = xs.cpu() + rs = rs.cpu() +originals = xs.numpy().reshape(-1, 64,64) +recons = rs.reshape(-1,64,64) + +# [ 0, 2, 1, 34, 4, 24] +def recon_check(original, recon): + fig = plt.figure() + ax0 = fig.add_subplot(121) + plt.imshow(original, cmap='Greys_r', interpolation='nearest') + plt.axis('off') + ax1 = fig.add_subplot(122) + plt.imshow(recon , cmap='Greys_r', interpolation='nearest') + plt.axis('off') + +def f(x): + fig = plt.figure() + ax0 = fig.add_subplot(121) + plt.imshow(originals[x], cmap='Greys_r', interpolation='nearest') + plt.axis('off') + ax1 = fig.add_subplot(122) + plt.imshow(recons[x], cmap='Greys_r', interpolation='nearest') + plt.axis('off') + +interact(f, x=widgets.IntSlider(min=0, max=xs.shape[0], step=1, value=0)) + +y_names = ['shape', 'scale', 'orientation', 'posX', 'posY'] +y_shapes = np.array((3,6,40,32,32)) +img_dict = {} + +for i, img in enumerate(imgs_sampled): + img_dict[tuple(labels_sampled[i])] = img + +def find_in_dataset(shape, scale, orient, posX, posY): + fig = plt.figure() + img = img_dict[(0, shape, scale, orient, posX, posY)] + plt.imshow(img.reshape(64,64), cmap='Greys_r', interpolation='nearest') + plt.axis('off') + +interact(find_in_dataset, + shape=widgets.IntSlider(min=0, max=2, step=1, value=0), + scale=widgets.IntSlider(min=0, max=5, step=1, value=0), + orient=widgets.IntSlider(min=0, max=39, step=1, value=0), + posX=widgets.IntSlider(min=0, max=31, step=1, value=0), + posY=widgets.IntSlider(min=0, max=31, step=1, value=0)) + +def get_specific_data(args=dict(), cuda=False): + ''' + use this function to get examples of data with specific class labels + inputs: + args - dictionary whose keys can include {shape, scale, orientation, + posX, posY} and values can include any integers less than the + corresponding size of that label dimension + cuda - bool to indicate whether the output should be placed on GPU + ''' + names_dict = {'shape': 1, 'scale': 2, 'orientation': 3, 'posX': 4, 'posY': 5} + selected_ind = np.ones(imgs.shape[0], dtype=bool) + for k,v in args.items(): + col_id = names_dict[k] + selected_ind = np.bitwise_and(selected_ind, labels[:, col_id] == v) + ind = np.random.choice(np.arange(imgs.shape[0])[selected_ind]) + x = torch.from_numpy(imgs[ind].reshape(1,64**2).astype(np.float32)) + y = torch.from_numpy(labels[ind].reshape(1,6).astype(np.float32)) + if not cuda: + return x,y + x = x.cuda() + y = y.cuda() + return x,y + +def plot_image(x): + """ + helper to plot dSprites images + """ + x = x.cpu() + plt.figure() + plt.imshow(x.reshape(64,64), interpolation='nearest', cmap='Greys_r') + plt.axis('off') + +def see_specific_image(args=dict(), verbose=True): + ''' + use this function to get examples of data with specific class labels + inputs: + args - dictionary whose keys can include {shape, scale, orientation, + posX, posY} and values can include any integers less than the + corresponding size of that label dimension + verbose - bool to indicate whether the full class label should be written + as the title of the plot + ''' + x,y = get_specific_data(args, cuda=False) + plot_image(x) + if verbose: + string = '' + for i, s in enumerate(['Shape', 'Scale', 'Orientation', 'PosX', 'PosY']): + string += '%s: %d, ' % (s, int(y[0][i+1])) + if i == 2: + string = string[:-2] + '\n' + plt.title(string[:-2], fontsize=12) + +def compare_reconstruction(original, recon): + """ + compare two images side by side + inputs: + original - array for original image + recon - array for recon image + """ + fig = plt.figure() + ax0 = fig.add_subplot(121) + plt.imshow(original.cpu().reshape(64,64), cmap='Greys_r', interpolation='nearest') + plt.axis('off') + plt.title('original') + ax1 = fig.add_subplot(122) + plt.imshow(recon.cpu().reshape(64,64), cmap='Greys_r', interpolation='nearest') + plt.axis('off') + plt.title('reconstruction') + +def compare_to_density(original, recons): + """ + compare two images side by side + inputs: + original - array for original image + recon - array of multiple recon images + """ + fig = plt.figure() + ax0 = fig.add_subplot(121) + plt.imshow(original.cpu().reshape(64,64), cmap='Greys_r', interpolation='nearest') + plt.axis('off') + plt.title('original') + ax1 = fig.add_subplot(122) + plt.imshow(torch.mean(recons.cpu(), 0).reshape(64,64), cmap='Greys_r', interpolation='nearest') + plt.axis('off') + plt.title('reconstructions') + + +see_specific_image() + +label_dims = vae.label_shape +label_dim_offsets = np.cumsum(label_dims) +label_dim_offsets + +class SCM(): + """ + Structural causal model + + args: + vae: instance of vae + mu: loc of q(z|x) given by the vae encoder + sigma: scale of q(z|x) given by the vae encoder + + """ + def __init__(self, vae, mu, sigma): + """ + Constructor + + Intializes : + image dimensions - 4096(64*64), + z dimensions: size of the tensor representing the latent random variable z, + label dimensions: 114 labels y that correspond to an image(one hot encoded) + f(x) = p(x|y,z) + Noise variables in the model N_# + """ + self.vae = vae + self.image_dim = vae.image_dim + self.z_dim = vae.z_dim + # these are used for f_X + self.label_dims = vae.label_shape + + def f_X(Y, Z, N): + """ + Generating one hots for the factors + """ + zs = Z.cuda() + # convert the labels to one hot + ys = [torch.tensor([0])] + ys.append(torch.nn.functional.one_hot(torch.round(Y[0]).to(torch.long), int(self.label_dims[1]))) + ys.append(torch.nn.functional.one_hot(torch.round(Y[1]).to(torch.long), int(self.label_dims[2]))) + ys.append(torch.nn.functional.one_hot(torch.round(Y[2]).to(torch.long), int(self.label_dims[3]))) + ys.append(torch.nn.functional.one_hot(torch.round(Y[3]).to(torch.long), int(self.label_dims[4]))) + ys.append(torch.nn.functional.one_hot(torch.round(Y[4]).to(torch.long), int(self.label_dims[5]))) + ys = torch.cat(ys).to(torch.float32).reshape(1,-1).cuda() + p = vae.decoder.forward(zs, ys) + return (N < p.cpu()).type(torch.float) + + def f_Y(N): + """ + Gumbel distribution - to model the distribution of the maximum of a number of samples + m = Gumbel(torch.tensor([1.0]), torch.tensor([2.0])).sample() # sample from Gumbel distribution with loc=1, scale=2 + tensor([ 1.0124]) + + https://pytorch.org/docs/stable/_modules/torch/distributions/gumbel.html + """ +# m = torch.distributions.gumbel.Gumbel(torch.zeros(N.size(0)), torch.ones(N.size(0))) + beta = 12 + indices = torch.tensor(np.arange(N.size(0))).to(torch.float32) + smax = nn.functional.softmax(beta*N) + argmax_ind = torch.sum(smax*indices) + return argmax_ind + + def f_Z(N): + """ + Z ~ Normal(mu, sigma) + """ + return N * sigma + mu + + def model(noise): + """ + The model corresponds to a generative process + + args: noise variables + return: X(image), Y(labels), Z(latents) + """ + N_X = pyro.sample( 'N_X', noise['N_X'].to_event(1) ) + # denoted using the index in the sequence + # that they are stored in as vae.label_names: + # ['shape', 'scale', 'orientation', 'posX', 'posY'] + N_Y_1 = pyro.sample( 'N_Y_1', noise['N_Y_1'].to_event(1) ) + N_Y_2 = pyro.sample( 'N_Y_2', noise['N_Y_2'].to_event(1) ) + N_Y_3 = pyro.sample( 'N_Y_3', noise['N_Y_3'].to_event(1) ) + N_Y_4 = pyro.sample( 'N_Y_4', noise['N_Y_4'].to_event(1) ) + N_Y_5 = pyro.sample( 'N_Y_5', noise['N_Y_5'].to_event(1) ) + + # Z ~ Normal(Nx_mu, Nx_sigma) + N_Z = pyro.sample( 'N_Z', noise['N_Z'].to_event(1) ) + Z = pyro.sample('Z', dist.Normal( f_Z( N_Z ), 1e-1).to_event(1) ) + + # Y ~ Gumbel max of Ny +# Y_1_mu = f_Y(N_Y_1) +# Y_2_mu = f_Y(N_Y_2) +# Y_3_mu = f_Y(N_Y_3) +# Y_4_mu = f_Y(N_Y_4) +# Y_5_mu = f_Y(N_Y_5) + + Y_1 = pyro.sample('Y_1', dist.Normal( f_Y(N_Y_1), 1e-2) ) + Y_2 = pyro.sample('Y_2', dist.Normal( f_Y(N_Y_2), 1e-1) ) + Y_3 = pyro.sample('Y_3', dist.Normal( f_Y(N_Y_3), 1e-1) ) + Y_4 = pyro.sample('Y_4', dist.Normal( f_Y(N_Y_4), 1e-1) ) + Y_5 = pyro.sample('Y_5', dist.Normal( f_Y(N_Y_5), 1e-1) ) + +# Y_mu = (Y_1_mu, Y_2_mu, Y_3_mu, Y_4_mu, Y_5_mu) + + # X ~ p(x|y,z) = bernoulli(loc(y,z)) + X = pyro.sample('X', dist.Normal( f_X( (Y_1, Y_2, Y_3,Y_4,Y_5), Z, N_X ), 1e-2).to_event(1)) + + # return noise and variables + noise_samples = N_X, (N_Y_1, N_Y_2, N_Y_3, N_Y_4, N_Y_5), N_Z + variable_samples = X, (Y_1, Y_2, Y_3, Y_4, Y_5), Z + return variable_samples, noise_samples + + self.model = model + #Initialize all noise variables in the model + self.init_noise = { + 'N_X' : dist.Uniform(torch.zeros(vae.image_dim), torch.ones(vae.image_dim)), + 'N_Z' : dist.Normal(torch.zeros(vae.z_dim), torch.ones(vae.z_dim)), + 'N_Y_1' : dist.Uniform(torch.zeros(self.label_dims[1]),torch.ones(self.label_dims[1])), + 'N_Y_2' : dist.Uniform(torch.zeros(self.label_dims[2]),torch.ones(self.label_dims[2])), + 'N_Y_3' : dist.Uniform(torch.zeros(self.label_dims[3]),torch.ones(self.label_dims[3])), + 'N_Y_4' : dist.Uniform(torch.zeros(self.label_dims[4]),torch.ones(self.label_dims[4])), + 'N_Y_5' : dist.Uniform(torch.zeros(self.label_dims[5]),torch.ones(self.label_dims[5])) + } + + def update_noise_svi(self, obs_data, intervened_model=None): + """ + Use svi to find out the mu, sigma of the distributionsfor the + condition outlined in obs_data + """ + + def guide(noise): + """ + The guide serves as an approximation to the posterior p(z|x). + The guide provides a valid joint probability density over all the + latent random variables in the model. + + https://pyro.ai/examples/svi_part_i.html + """ + # create params with constraints + mu = { + 'N_X': pyro.param('N_X_mu', 0.5*torch.ones(self.image_dim),constraint = constraints.interval(0., 1.)), + 'N_Z': pyro.param('N_Z_mu', torch.zeros(self.z_dim),constraint = constraints.interval(-3., 3.)), + 'N_Y_1': pyro.param('N_Y_1_mu', 0.5*torch.ones(self.label_dims[1]),constraint = constraints.interval(0., 1.)), + 'N_Y_2': pyro.param('N_Y_2_mu', 0.5*torch.ones(self.label_dims[2]),constraint = constraints.interval(0., 1.)), + 'N_Y_3': pyro.param('N_Y_3_mu', 0.5*torch.ones(self.label_dims[3]),constraint = constraints.interval(0., 1.)), + 'N_Y_4': pyro.param('N_Y_4_mu', 0.5*torch.ones(self.label_dims[4]),constraint = constraints.interval(0., 1.)), + 'N_Y_5': pyro.param('N_Y_5_mu', 0.5*torch.ones(self.label_dims[5]),constraint = constraints.interval(0., 1.)) + } + sigma = { + 'N_X': pyro.param('N_X_sigma', 0.1*torch.ones(self.image_dim),constraint = constraints.interval(0.0001, 0.5)), + 'N_Z': pyro.param('N_Z_sigma', torch.ones(self.z_dim),constraint = constraints.interval(0.0001, 3.)), + 'N_Y_1': pyro.param('N_Y_1_sigma', 0.1*torch.ones(self.label_dims[1]),constraint = constraints.interval(0.0001, 0.5)), + 'N_Y_2': pyro.param('N_Y_2_sigma', 0.1*torch.ones(self.label_dims[2]),constraint = constraints.interval(0.0001, 0.5)), + 'N_Y_3': pyro.param('N_Y_3_sigma', 0.1*torch.ones(self.label_dims[3]),constraint = constraints.interval(0.0001, 0.5)), + 'N_Y_4': pyro.param('N_Y_4_sigma', 0.1*torch.ones(self.label_dims[4]),constraint = constraints.interval(0.0001, 0.5)), + 'N_Y_5': pyro.param('N_Y_5_sigma', 0.1*torch.ones(self.label_dims[5]),constraint = constraints.interval(0.0001, 0.5)) + } + for noise_term in noise.keys(): + pyro.sample(noise_term, dist.Normal(mu[noise_term], sigma[noise_term]).to_event(1)) + + # Condition the model + if intervened_model is not None: + obs_model = pyro.condition(intervened_model, obs_data) + else: + obs_model = pyro.condition(self.model, obs_data) + + pyro.clear_param_store() + + # Once we’ve specified a guide, we’re ready to proceed to inference. + # Now, this an optimization problem where each iteration of training takes + # a step that moves the guide closer to the exact posterior + + # https://arxiv.org/pdf/1601.00670.pdf + svi = SVI( + model= obs_model, + guide= guide, + optim= SGD({"lr": 1e-5, 'momentum': 0.1}), + loss=Trace_ELBO(retain_graph=True) + ) + + num_steps = 1500 + samples = defaultdict(list) + for t in range(num_steps): + loss = svi.step(self.init_noise) +# if t % 100 == 0: +# print("step %d: loss of %.2f" % (t, loss)) + for noise in self.init_noise.keys(): + mu = '{}_mu'.format(noise) + sigma = '{}_sigma'.format(noise) + samples[mu].append(pyro.param(mu).detach().numpy()) + samples[sigma].append(pyro.param(sigma).detach().numpy()) + means = {k: torch.tensor(np.array(v).mean(axis=0)) for k, v in samples.items()} + + # update the inferred noise + updated_noise = { + 'N_X' : dist.Normal(means['N_X_mu'], means['N_X_sigma']), + 'N_Z' : dist.Normal(means['N_Z_mu'], means['N_Z_sigma']), + 'N_Y_1': dist.Normal(means['N_Y_1_mu'], means['N_Y_1_sigma']), + 'N_Y_2': dist.Normal(means['N_Y_2_mu'], means['N_Y_2_sigma']), + 'N_Y_3': dist.Normal(means['N_Y_3_mu'], means['N_Y_3_sigma']), + 'N_Y_4': dist.Normal(means['N_Y_4_mu'], means['N_Y_4_sigma']), + 'N_Y_5': dist.Normal(means['N_Y_5_mu'], means['N_Y_5_sigma']), + } + return updated_noise + + def __call__(self): + return self.model(self.init_noise) + +"""## Sanity check: 1 +### Making sure VAE works +""" + +# Generate an instance of dSprites image +ox, y = get_specific_data(cuda=True) +plot_image(ox) +# Pass it through VAE to get q(z|x) => N(mu, sigma) +mu, sigma = vae.encoder.forward(ox,vae.remap_y(y)) +# Feed these params to our custom SCM +scm = SCM(vae, mu.cpu(), sigma.cpu()) +print(y) +# Check for reconstruction + +"""## Sanity check 2 + +### To check if the decoder is able to generate the image if the latents are changed: +#### To achieve this we manually change the labels in the code and run it through the decoder and check for reconstruction +""" + +original, y_original = get_specific_data(cuda=True) +print('top: ',y_original) +mu, sigma = vae.encoder.forward(original,vae.remap_y(y_original)) +B = 100 +zs = torch.cat([dist.Normal(mu.cpu(), sigma.cpu()).sample() for a in range(B)], 0) +ys = torch.cat([vae.remap_y(y_original) for a in range(B)], 0) +rs = vae.decoder.forward(zs.cuda(), ys).detach() +compare_to_density(original,rs) + +y_new = torch.tensor(y_original) +y_new[0,1] = (y_original[0,1] + 1) % 2 +print('bottom: ', y_new) +zs = torch.cat([dist.Normal(mu.cpu(), sigma.cpu()).sample() for a in range(B)], 0) +ys = torch.cat([vae.remap_y(y_new) for a in range(B)], 0) +rs = vae.decoder.forward(zs.cuda(), ys).detach() +compare_to_density(original,rs) + +"""## Conditioning the model""" + +cond_data = {} +for i in range(1, 6): + cond_data["Y_{}".format(i)] = torch.tensor(y[0,i].cpu()).to(torch.float32) +print(cond_data) + +# cond_data['Y_1'] = torch.tensor(1.) +# cond_data['Y_2'] = torch.tensor(4.) +conditioned_model = pyro.condition(scm.model, data=cond_data) +cond_noise = scm.update_noise_svi(cond_data) +print(cond_data) + +rxs = [] +for i in range(100): + (rx,ry,_), _ = scm.model(cond_noise) + rxs.append(rx) +compare_to_density(ox, torch.cat(rxs)) +_ =plt.suptitle("SCM Conditioned on Original", fontsize=18, fontstyle='italic') + +"""## Counterfactuals""" + +# intervening on Shape, posX and PosY +intervened_model = pyro.do(scm.model, data={ + "Y_1": torch.tensor(0.), + "Y_4": torch.tensor(30.), + "Y_5": torch.tensor(25.), +}) +noise_data = {} +for term, d in cond_noise.items(): + noise_data[term] = d.loc +# intervened_noise = scm.update_noise_svi(noise_data, intervened_model) + +(rx1,ry,_), _ = intervened_model(scm.init_noise) +compare_to_density(ox, rx1) +print(ry) + +rxs = [] +for i in range(5000): + (cfo1,ny1,nz1), _= intervened_model(cond_noise) + rxs.append(cfo1) +compare_to_density(ox, torch.cat(rxs)) +_ =plt.suptitle("SCM intervened on shape", fontsize=18, fontstyle='italic') + +"""## ======== Test code =======""" + +import warnings +warnings.filterwarnings("ignore") + +xs, ys = next(data_iter) +x = xs[0].reshape(1,-1).cuda() +y = ys[0].reshape(1,-1).cuda() +mu, sigma = vae.encoder.forward(x,vae.remap_y(y)) +mu = mu.cpu() +sigma = sigma.cpu() +recon_x1, y1, z1 = OldSCM(vae, mu, sigma) +print(y1) + +def OldSCM(vae, mu, sigma): + z_dim = vae.z_dim + Ny, Y, ys = [], [], [] + Nx = pyro.sample("Nx", dist.Uniform(torch.zeros(vae.image_dim), torch.ones(vae.image_dim))) + Nz = pyro.sample("Nz", dist.Normal(torch.zeros(z_dim), torch.ones(z_dim))) + m = torch.distributions.gumbel.Gumbel(torch.tensor(0.0), torch.tensor(1.0)) + for label_id in range(6): + name = vae.label_names[label_id] + length = vae.label_shape[label_id] + new = pyro.sample("Ny_%s"%name, dist.Uniform(torch.zeros(length), torch.ones(length)) ) + Ny.append(new) + gumbel_vars = torch.tensor([m.sample() for _ in range(length)]) + max_ind = torch.argmax(torch.log(new) + gumbel_vars).item() + Y.append(pyro.sample("Y_%s"%name, dist.Normal(torch.tensor(max_ind * 1.0), 1e-4))) +# Y.append(pyro.sample("Y_%s"%name, dist.Delta(torch.tensor(max_ind*1.0)))) + ys.append(torch.nn.functional.one_hot(torch.tensor(max_ind), int(length))) + Y = torch.tensor(Y) + ys = torch.cat(ys).to(torch.float32).reshape(1,-1).cuda() + Z = pyro.sample("Z", dist.Normal(mu + Nz*sigma, 1e-4)) +# Z = pyro.sample("Z", dist.Delta(mu + Nz*sigma)) + zs = Z.cuda() + p = vae.decoder.forward(zs,ys) + X = pyro.sample("X", dist.Normal((Nx < p.cpu()).type(torch.float), 1e-4)) +# X = pyro.sample("X", dist.Delta((Nx < p.cpu()).type(torch.float))) + return X, Y, Z + +from pyro.infer.importance import Importance +from pyro.infer.mcmc import MCMC +from pyro.infer.mcmc.nuts import HMC +from pyro.infer import SVI + +intervened_model = pyro.do(OldSCM, data={"Y_shape": torch.tensor(0.)}) +conditioned_model = pyro.condition(OldSCM, data={ + "X": recon_x1, +# "Z": z1, + "Y_color":y1[0], + "Y_shape": y1[1], + "Y_scale": y1[2], + "Y_orientation": y1[3], + "Y_posX": y1[4], + "Y_posY": y1[5] + }) + +# change to svi +posterior = pyro.infer.Importance(conditioned_model, num_samples = 100) +posterior.run(vae, mu, sigma) + +result = [] +for i in range(500): + trace = posterior() + x = trace.nodes['Nx']['value'] + ny_shape = trace.nodes['Ny_shape']['value'] + ny_scale = trace.nodes['Ny_scale']['value'] + ny_orientation = trace.nodes['Ny_orientation']['value'] + ny_posX = trace.nodes['Ny_posX']['value'] + ny_posY = trace.nodes['Ny_posY']['value'] + z = trace.nodes['Nz']['value'] + con_obj = pyro.condition(intervened_model, data = { + "Nx": x, + "Ny_shape": ny_shape, + "Ny_scale": ny_scale, + "Ny_orientation": ny_orientation, + "Ny_posX": ny_posX, + "Ny_posY": ny_posY, + "Nz": z + }) + +recon_x2,y2,z2 = con_obj(vae, mu, sigma) +print(y2) +recon_check(recon_x1.reshape(-1, 64, 64)[0], recon_x2.reshape(-1, 64, 64)[0]) + +result = [] +for i in range(500): + trace = posterior() + x = trace.nodes['Nx']['value'] + ny_shape = trace.nodes['Ny_shape']['value'] + ny_scale = trace.nodes['Ny_scale']['value'] + ny_orientation = trace.nodes['Ny_orientation']['value'] + ny_posX = trace.nodes['Ny_posX']['value'] + ny_posY = trace.nodes['Ny_posY']['value'] + z = trace.nodes['Nz']['value'] + con_obj = pyro.condition(intervened_model, data = { + "Nx": x, + "Ny_shape": ny_shape, + "Ny_scale": ny_scale, + "Ny_orientation": ny_orientation, + "Ny_posX": ny_posX, + "Ny_posY": ny_posY, + "Nz": z + }) + +recon_x2,y2,z2 = con_obj(vae, mu, sigma) +print(y2) +recon_check(recon_x1.reshape(-1, 64, 64)[0], recon_x2.reshape(-1, 64, 64)[0]) + +vae.label_names + +adam_params = {"lr": 0.00042, "betas": (0.9, 0.999)} +optimizer = Adam(adam_params) +# set up the loss(es) for inference. wrapping the guide in config_enumerate builds +# the loss as a sum +# by enumerating each class label for the sampled discrete categorical distribution +# in the model +guide = config_enumerate(guide, "parallel", expand=True) +elbo = Trace_ELBO(max_plate_nesting=1) +loss_basic = SVI(sup_vae.model, guide, optimizer, loss=elbo) + +# build a list of all losses considered +losses = [loss_basic] +loss_aux = SVI(sup_vae.model_classify, sup_vae.guide_classify, optimizer, loss=elbo) +losses.append(loss_aux) + + + +from pyro.optim import Adam +from pyro.infer import SVI, Trace_ELBO + +adam = Adam({"lr": 0.005, "betas": (0.90, 0.999)}) +svi = SVI(vae.model, vae.guide, adam, loss=Trace_ELBO()) + + +param_vals = [] +pbar = tqdm(range(5)) +for _ in pbar: + xs, ys = next(data_iter) + xs = xs.cuda() + ys = ys.cuda() + svi.step(xs, ys) +# param_vals.append({k: param(k).item() for k in ["fl", "ia"]}) + +# pd.DataFrame(param_vals).plot(subplots=True) + diff --git a/deep-fakes.jpg b/deep-fakes.jpg new file mode 100644 index 0000000..6541c59 Binary files /dev/null and b/deep-fakes.jpg differ diff --git a/figs/cs-vae-1.png b/figs/cs-vae-1.png new file mode 100644 index 0000000..0c0ab7c Binary files /dev/null and b/figs/cs-vae-1.png differ diff --git a/figs/dag.png b/figs/dag.png new file mode 100644 index 0000000..afdc92d Binary files /dev/null and b/figs/dag.png differ diff --git a/figs/deep-fakes.jpg b/figs/deep-fakes.jpg new file mode 100644 index 0000000..6541c59 Binary files /dev/null and b/figs/deep-fakes.jpg differ diff --git a/figs/intervention-2.png b/figs/intervention-2.png new file mode 100644 index 0000000..9b4b489 Binary files /dev/null and b/figs/intervention-2.png differ diff --git a/figs/intervention.png b/figs/intervention.png new file mode 100644 index 0000000..26350e9 Binary files /dev/null and b/figs/intervention.png differ diff --git a/figs/manual-intervention.png b/figs/manual-intervention.png new file mode 100644 index 0000000..daaa308 Binary files /dev/null and b/figs/manual-intervention.png differ diff --git a/figs/scm-conditioned.png b/figs/scm-conditioned.png new file mode 100644 index 0000000..fdb14df Binary files /dev/null and b/figs/scm-conditioned.png differ diff --git a/figs/scm.png b/figs/scm.png new file mode 100644 index 0000000..4953807 Binary files /dev/null and b/figs/scm.png differ diff --git a/figs/vae capturing latent.png b/figs/vae capturing latent.png new file mode 100644 index 0000000..a18b095 Binary files /dev/null and b/figs/vae capturing latent.png differ diff --git a/figs/vae-recons.png b/figs/vae-recons.png new file mode 100644 index 0000000..23a565f Binary files /dev/null and b/figs/vae-recons.png differ diff --git a/projects.pdf b/projects.pdf new file mode 100644 index 0000000..b9db080 Binary files /dev/null and b/projects.pdf differ diff --git a/vae_ajeya.ipynb b/vae_ajeya.ipynb new file mode 100644 index 0000000..3cb3a1e --- /dev/null +++ b/vae_ajeya.ipynb @@ -0,0 +1,1135 @@ +{ + "nbformat": 4, + "nbformat_minor": 0, + "metadata": { + "colab": { + "name": "vae-ajeya.ipynb", + "version": "0.3.2", + "provenance": [], + "collapsed_sections": [], + "include_colab_link": true + }, + "language_info": { + "codemirror_mode": { + "name": "ipython", + "version": 3 + }, + "file_extension": ".py", + "mimetype": "text/x-python", + "name": "python", + "nbconvert_exporter": "python", + "pygments_lexer": "ipython3", + "version": "3.7.1" + }, + "kernelspec": { + "name": "python3", + "display_name": "Python 3" + }, + "accelerator": "GPU" + }, + "cells": [ + { + "cell_type": "markdown", + "metadata": { + "id": "view-in-github", + "colab_type": "text" + }, + "source": [ + "\"Open" + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "6rtbzfmeMGvr", + "colab_type": "code", + "outputId": "5511d557-1110-448e-ad34-ce6095067c64", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 496 + } + }, + "source": [ + "!pip3 install pyro-ppl\n", + "!pip3 install torch torchvision\n", + "!pip3 install pydrive --upgrade" + ], + "execution_count": 1, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Requirement already satisfied: pyro-ppl in /usr/local/lib/python3.6/dist-packages (0.3.4)\n", + "Requirement already satisfied: six>=1.10.0 in /usr/local/lib/python3.6/dist-packages (from pyro-ppl) (1.12.0)\n", + "Requirement already satisfied: contextlib2 in /usr/local/lib/python3.6/dist-packages (from pyro-ppl) (0.5.5)\n", + "Requirement already satisfied: tqdm>=4.31 in /usr/local/lib/python3.6/dist-packages (from pyro-ppl) (4.32.2)\n", + "Requirement already satisfied: opt-einsum>=2.3.2 in /usr/local/lib/python3.6/dist-packages (from pyro-ppl) (2.3.2)\n", + "Requirement already satisfied: torch>=1.1.0 in /usr/local/lib/python3.6/dist-packages (from pyro-ppl) (1.1.0)\n", + "Requirement already satisfied: graphviz>=0.8 in /usr/local/lib/python3.6/dist-packages (from pyro-ppl) (0.10.1)\n", + "Requirement already satisfied: numpy>=1.7 in /usr/local/lib/python3.6/dist-packages (from pyro-ppl) (1.16.4)\n", + "Requirement already satisfied: torch in /usr/local/lib/python3.6/dist-packages (1.1.0)\n", + "Requirement already satisfied: torchvision in /usr/local/lib/python3.6/dist-packages (0.3.0)\n", + "Requirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from torch) (1.16.4)\n", + "Requirement already satisfied: six in /usr/local/lib/python3.6/dist-packages (from torchvision) (1.12.0)\n", + "Requirement already satisfied: pillow>=4.1.1 in /usr/local/lib/python3.6/dist-packages (from torchvision) (4.3.0)\n", + "Requirement already satisfied: olefile in /usr/local/lib/python3.6/dist-packages (from pillow>=4.1.1->torchvision) (0.46)\n", + "Requirement already up-to-date: pydrive in /usr/local/lib/python3.6/dist-packages (1.3.1)\n", + "Requirement already satisfied, skipping upgrade: oauth2client>=4.0.0 in /usr/local/lib/python3.6/dist-packages (from pydrive) (4.1.3)\n", + "Requirement already satisfied, skipping upgrade: PyYAML>=3.0 in /usr/local/lib/python3.6/dist-packages (from pydrive) (3.13)\n", + "Requirement already satisfied, skipping upgrade: google-api-python-client>=1.2 in /usr/local/lib/python3.6/dist-packages (from pydrive) (1.7.9)\n", + "Requirement already satisfied, skipping upgrade: httplib2>=0.9.1 in /usr/local/lib/python3.6/dist-packages (from oauth2client>=4.0.0->pydrive) (0.11.3)\n", + "Requirement already satisfied, skipping upgrade: rsa>=3.1.4 in /usr/local/lib/python3.6/dist-packages (from oauth2client>=4.0.0->pydrive) (4.0)\n", + "Requirement already satisfied, skipping upgrade: pyasn1>=0.1.7 in /usr/local/lib/python3.6/dist-packages (from oauth2client>=4.0.0->pydrive) (0.4.5)\n", + "Requirement already satisfied, skipping upgrade: pyasn1-modules>=0.0.5 in /usr/local/lib/python3.6/dist-packages (from oauth2client>=4.0.0->pydrive) (0.2.5)\n", + "Requirement already satisfied, skipping upgrade: six>=1.6.1 in /usr/local/lib/python3.6/dist-packages (from oauth2client>=4.0.0->pydrive) (1.12.0)\n", + "Requirement already satisfied, skipping upgrade: google-auth>=1.4.1 in /usr/local/lib/python3.6/dist-packages (from google-api-python-client>=1.2->pydrive) (1.4.2)\n", + "Requirement already satisfied, skipping upgrade: uritemplate<4dev,>=3.0.0 in /usr/local/lib/python3.6/dist-packages (from google-api-python-client>=1.2->pydrive) (3.0.0)\n", + "Requirement already satisfied, skipping upgrade: google-auth-httplib2>=0.0.3 in /usr/local/lib/python3.6/dist-packages (from google-api-python-client>=1.2->pydrive) (0.0.3)\n", + "Requirement already satisfied, skipping upgrade: cachetools>=2.0.0 in /usr/local/lib/python3.6/dist-packages (from google-auth>=1.4.1->google-api-python-client>=1.2->pydrive) (3.1.1)\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "OouaeUpQMEjN", + "colab_type": "code", + "colab": {} + }, + "source": [ + "from __future__ import absolute_import\n", + "from __future__ import division\n", + "from __future__ import print_function\n", + "from matplotlib import pyplot as plt\n", + "import numpy as np\n", + "import seaborn as sns\n", + "\n", + "import os\n", + "\n", + "import torch\n", + "import torchvision.datasets as dset\n", + "import torch.nn as nn\n", + "import torchvision.transforms as transforms\n", + "\n", + "import pyro\n", + "from pyro.contrib.examples.util import print_and_log\n", + "import pyro.distributions as dist\n", + "from pyro.infer import SVI, Trace_ELBO, TraceEnum_ELBO, config_enumerate\n", + "from pyro.optim import Adam\n", + "\n", + "# Change figure aesthetics\n", + "%matplotlib inline\n", + "sns.set_context('talk', font_scale=1.2, rc={'lines.linewidth': 1.5})\n", + "\n", + "USE_CUDA = True\n", + "\n", + "pyro.enable_validation(True)\n", + "pyro.distributions.enable_validation(False)\n", + "\n", + "# from custom_mlp import MLP, Expxt" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "s1p7kozIT9YN", + "colab_type": "code", + "outputId": "b43bca1d-270d-4b0f-a5db-72480fc19511", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 34 + } + }, + "source": [ + "import tensorflow as tf\n", + "tf.test.gpu_device_name()" + ], + "execution_count": 3, + "outputs": [ + { + "output_type": "execute_result", + "data": { + "text/plain": [ + "'/device:GPU:0'" + ] + }, + "metadata": { + "tags": [] + }, + "execution_count": 3 + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "ux30_pwSUEJX", + "colab_type": "code", + "outputId": "84958572-1246-4eab-f56b-118092b21e7a", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 102 + } + }, + "source": [ + "# Hack to get all available GPU ram.\n", + "!ln -sf /opt/bin/nvidia-smi /usr/bin/nvidia-smi\n", + "!pip install gputil\n", + "!pip install psutil\n", + "!pip install humanize\n", + "import psutil\n", + "import humanize\n", + "import os\n", + "import GPUtil as GPU\n", + "GPUs = GPU.getGPUs()\n", + "# XXX: only one GPU on Colab and isn’t guaranteed\n", + "gpu = GPUs[0]\n", + "def printm():\n", + " process = psutil.Process(os.getpid())\n", + " print(\"Gen RAM Free: \" + humanize.naturalsize( psutil.virtual_memory().available ), \" | Proc size: \" + humanize.naturalsize( process.memory_info().rss))\n", + " print(\"GPU RAM Free: {0:.0f}MB | Used: {1:.0f}MB | Util {2:3.0f}% | Total {3:.0f}MB\".format(gpu.memoryFree, gpu.memoryUsed, gpu.memoryUtil*100, gpu.memoryTotal))\n", + "printm()" + ], + "execution_count": 4, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Requirement already satisfied: gputil in /usr/local/lib/python3.6/dist-packages (1.4.0)\n", + "Requirement already satisfied: psutil in /usr/local/lib/python3.6/dist-packages (5.4.8)\n", + "Requirement already satisfied: humanize in /usr/local/lib/python3.6/dist-packages (0.5.1)\n", + "Gen RAM Free: 10.6 GB | Proc size: 3.5 GB\n", + "GPU RAM Free: 14310MB | Used: 769MB | Util 5% | Total 15079MB\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "BPXd84WMYr3W", + "colab_type": "code", + "outputId": "62d05def-f7f5-4b45-ee41-b9632f8a7034", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 34 + } + }, + "source": [ + "from google.colab import drive\n", + "drive.mount('/content/gdrive')" + ], + "execution_count": 5, + "outputs": [ + { + "output_type": "stream", + "text": [ + "Drive already mounted at /content/gdrive; to attempt to forcibly remount, call drive.mount(\"/content/gdrive\", force_remount=True).\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "0GhM07OkM4_9", + "colab_type": "code", + "colab": {} + }, + "source": [ + "\n", + "\n", + "from pydrive.auth import GoogleAuth\n", + "from pydrive.drive import GoogleDrive\n", + "from google.colab import auth\n", + "from oauth2client.client import GoogleCredentials\n", + "\n", + "auth.authenticate_user()\n", + "gauth = GoogleAuth()\n", + "gauth.credentials = GoogleCredentials.get_application_default()\n", + "drive = GoogleDrive(gauth)" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "Y8E4_2q1NTZt", + "colab_type": "code", + "colab": {} + }, + "source": [ + "# https://drive.google.com/open?id=1mAtimHIWHJM2UJNxsIylPyI1jUebGIx-\n", + "custom_mlp_module = drive.CreateFile({'id':'1mAtimHIWHJM2UJNxsIylPyI1jUebGIx-'})" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "R-A-TTIFPfip", + "colab_type": "code", + "colab": {} + }, + "source": [ + "custom_mlp = custom_mlp_module.GetContentFile('custom_mlp.py')\n", + "from custom_mlp import MLP, Exp" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "Vomid1y5MEjP", + "colab_type": "code", + "colab": {} + }, + "source": [ + "class SVAE(nn.Module):\n", + " \"\"\"\n", + " This class encapsulates the parameters (neural networks) and models & guides needed to train a\n", + " semi-supervised variational auto-encoder on the Dsprites image dataset\n", + "\n", + " :param output_size: size of the tensor representing the class label \n", + " :param input_size: size of the tensor representing the image (64*64 = 4096 for our Dsprites dataset\n", + " since we flatten the images and scale the pixels to be in [0,1])\n", + " :param z_dim: size of the tensor representing the latent random variable z\n", + " (for our Dsprites dataset)\n", + " :param hidden_layers: a tuple (or list) of MLP layers to be used in the neural networks\n", + " representing the parameters of the distributions in our model\n", + " :param use_cuda: use GPUs for faster training\n", + " :param aux_loss_multiplier: the multiplier to use with the auxiliary loss\n", + " \"\"\"\n", + " def __init__(self, output_size=6, input_size=4096, z_dim=50, hidden_layers=(500,),\n", + " config_enum=None, use_cuda=USE_CUDA, aux_loss_multiplier=None):\n", + "\n", + " super(SVAE, self).__init__()\n", + "\n", + " # initialize the class with all arguments provided to the constructor\n", + " self.output_size = output_size\n", + " self.input_size = input_size\n", + " self.z_dim = z_dim\n", + " self.hidden_layers = hidden_layers\n", + " self.allow_broadcast = config_enum == 'parallel'\n", + " self.use_cuda = use_cuda\n", + " self.aux_loss_multiplier = aux_loss_multiplier\n", + "\n", + " # define and instantiate the neural networks representing\n", + " # the paramters of various distributions in the model\n", + " self.setup_networks()\n", + "\n", + " def setup_networks(self):\n", + "\n", + " z_dim = self.z_dim\n", + " hidden_sizes = self.hidden_layers\n", + "\n", + " # define the neural networks used later in the model and the guide.\n", + " # these networks are MLPs (multi-layered perceptrons or simple feed-forward networks)\n", + " # where the provided activation parameter is used on every linear layer except\n", + " # for the output layer where we use the provided output_activation parameter\n", + " self.encoder_y = MLP([self.input_size] + hidden_sizes + [self.output_size],\n", + " activation=nn.Softplus,\n", + " output_activation=nn.Softmax,\n", + " allow_broadcast=self.allow_broadcast,\n", + " use_cuda=self.use_cuda)\n", + "\n", + " # a split in the final layer's size is used for multiple outputs\n", + " # and potentially applying separate activation functions on them\n", + " # e.g. in this network the final output is of size [z_dim,z_dim]\n", + " # to produce loc and scale, and apply different activations [None,Exp] on them\n", + " self.encoder_z = MLP([self.input_size + self.output_size] +\n", + " hidden_sizes + [[z_dim, z_dim]],\n", + " activation=nn.Softplus,\n", + " output_activation=[None, Exp],\n", + " allow_broadcast=self.allow_broadcast,\n", + " use_cuda=self.use_cuda)\n", + "\n", + " self.decoder = MLP([z_dim + self.output_size] +\n", + " hidden_sizes + [self.input_size],\n", + " activation=nn.Softplus,\n", + " output_activation=nn.Sigmoid,\n", + " allow_broadcast=self.allow_broadcast,\n", + " use_cuda=self.use_cuda)\n", + "\n", + " # using GPUs for faster training of the networks\n", + " if self.use_cuda:\n", + " self.cuda()\n", + "\n", + " def model(self, xs, ys=None):\n", + " \"\"\"\n", + " The model corresponds to the following generative process:\n", + " p(z) = normal(0,I) # dsprites label (latent)\n", + " p(y|x) = categorical(I/10.) # which digit (supervised)\n", + " p(x|y,z) = bernoulli(loc(y,z)) # an image\n", + " loc is given by a neural network `decoder`\n", + "\n", + " :param xs: a batch of scaled vectors of pixels from an image\n", + " :param ys: (optional) a batch of the class labels i.e.\n", + " the digit corresponding to the image(s)\n", + " :return: None\n", + " \"\"\"\n", + " # register this pytorch module and all of its sub-modules with pyro\n", + " pyro.module(\"ss_vae\", self)\n", + "\n", + " batch_size = xs.size(0)\n", + " options = dict(dtype=xs.dtype, device=xs.device)\n", + " with pyro.plate(\"data\"):\n", + "\n", + " # sample the handwriting style from the constant prior distribution\n", + " prior_loc = torch.zeros(batch_size, self.z_dim, **options)\n", + " prior_scale = torch.ones(batch_size, self.z_dim, **options)\n", + " zs = pyro.sample(\"z\", dist.Normal(prior_loc, prior_scale).to_event(1))\n", + "\n", + " # if the label y (which digit to write) is supervised, sample from the\n", + " # constant prior, otherwise, observe the value (i.e. score it against the constant prior)\n", + " alpha_prior = torch.ones(batch_size, self.output_size, **options) / (1.0 * self.output_size)\n", + " ys = pyro.sample(\"y\", dist.OneHotCategorical(alpha_prior), obs=ys)\n", + "\n", + " # finally, score the image (x) using the handwriting style (z) and\n", + " # the class label y (which digit to write) against the\n", + " # parametrized distribution p(x|y,z) = bernoulli(decoder(y,z))\n", + " # where `decoder` is a neural network\n", + " loc = self.decoder.forward([zs, ys])\n", + " pyro.sample(\"x\", dist.Bernoulli(loc).to_event(1), obs=xs)\n", + " # return the loc so we can visualize it later\n", + " return loc\n", + "\n", + " def guide(self, xs, ys=None):\n", + " \"\"\"\n", + " The guide corresponds to the following:\n", + " q(y|x) = categorical(alpha(x)) # infer label from an image\n", + " q(z|x,y) = normal(loc(x,y),scale(x,y)) # infer latent class from an image and the label\n", + " loc, scale are given by a neural network `encoder_z`\n", + " alpha is given by a neural network `encoder_y`\n", + "\n", + " :param xs: a batch of scaled vectors of pixels from an image\n", + " :param ys: (optional) a batch of the class labels i.e.\n", + " the digit corresponding to the image(s)\n", + " :return: None\n", + " \"\"\"\n", + " # inform Pyro that the variables in the batch of xs, ys are conditionally independent\n", + " with pyro.plate(\"data\"):\n", + "\n", + " # if the class label (the digit) is not supervised, sample\n", + " # (and score) the digit with the variational distribution\n", + " # q(y|x) = categorical(alpha(x))\n", + " if ys is None: \n", + " alpha = self.encoder_y.forward(xs)\n", + " ys = pyro.sample(\"y\", dist.OneHotCategorical(alpha))\n", + "\n", + " # sample (and score) the latent handwriting-style with the variational\n", + " # distribution q(z|x,y) = normal(loc(x,y),scale(x,y))\n", + " loc, scale = self.encoder_z.forward([xs, ys])\n", + " pyro.sample(\"z\", dist.Normal(loc, scale).to_event(1))\n", + "\n", + " def classifier(self, xs):\n", + " \"\"\"\n", + " classify an image (or a batch of images)\n", + "\n", + " :param xs: a batch of scaled vectors of pixels from an image\n", + " :return: a batch of the corresponding class labels (as one-hots)\n", + " \"\"\"\n", + " # use the trained model q(y|x) = categorical(alpha(x))\n", + " # compute all class probabilities for the image(s)\n", + " alpha = self.encoder_y.forward(xs)\n", + "\n", + " # get the index (digit) that corresponds to\n", + " # the maximum predicted class probability\n", + " res, ind = torch.topk(alpha, 1)\n", + "\n", + " # convert the digit(s) to one-hot tensor(s)\n", + " ys = torch.zeros_like(alpha).scatter_(1, ind, 1.0)\n", + " return ys\n", + "\n", + " def model_classify(self, xs, ys=None):\n", + " \"\"\"\n", + " this model is used to add an auxiliary (supervised) loss as described in the\n", + " Kingma et al., \"Semi-Supervised Learning with Deep Generative Models\". It \n", + " probably isn't needed here.\n", + " \"\"\"\n", + " # register all pytorch (sub)modules with pyro\n", + " pyro.module(\"ss_vae\", self)\n", + "\n", + " # inform Pyro that the variables in the batch of xs, ys are conditionally independent\n", + " with pyro.plate(\"data\"):\n", + " # this here is the extra term to yield an auxiliary loss that we do gradient descent on\n", + " if ys is not None:\n", + " alpha = self.encoder_y.forward(xs)\n", + " with pyro.poutine.scale(scale=self.aux_loss_multiplier):\n", + " pyro.sample(\"y_aux\", dist.OneHotCategorical(alpha), obs=ys)\n", + "\n", + " def guide_classify(self, xs, ys=None):\n", + " \"\"\"\n", + " dummy guide function to accompany model_classify in inference\n", + " \"\"\"\n", + " pass" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "qPdgbk1kMEjR", + "colab_type": "code", + "colab": {} + }, + "source": [ + "def setup_data_loaders(train_x, train_y, test_x, test_y, batch_size=128, use_cuda=USE_CUDA):\n", + " train_dset = torch.utils.data.TensorDataset(\n", + " torch.from_numpy(train_x.astype(np.float32)).reshape(-1, 4096),\n", + " torch.from_numpy(train_y.astype(np.float32))\n", + " )\n", + " test_dset = torch.utils.data.TensorDataset(\n", + " torch.from_numpy(test_x.astype(np.float32)).reshape(-1, 4096),\n", + " torch.from_numpy(test_y.astype(np.float32))\n", + " ) \n", + " kwargs = {'num_workers': 1, 'pin_memory': use_cuda}\n", + " loader = {}\n", + " loader[\"sup\"] = torch.utils.data.DataLoader(\n", + " dataset=train_dset, batch_size=batch_size, shuffle=False, **kwargs\n", + " )\n", + " test_loader = torch.utils.data.DataLoader(\n", + " dataset=test_dset, batch_size=batch_size, shuffle=False, **kwargs\n", + " )\n", + " loader[\"valid\"] = test_loader\n", + " loader[\"test\"] = test_loader\n", + " return loader\n", + "\n", + "def run_supervized_inference_for_epoch(data_loaders, losses):\n", + " \"\"\"\n", + " runs the inference algorithm for an epoch\n", + " returns the values of all losses separately on supervised and unsupervised parts\n", + " \"\"\"\n", + " num_losses = len(losses)\n", + "\n", + " # compute number of batches for an epoch\n", + " batches_per_epoch = len(data_loaders[\"sup\"])\n", + "\n", + " # initialize variables to store loss values\n", + " epoch_losses = [0.] * num_losses\n", + "\n", + " # setup the iterators for training data loaders\n", + " sup_iter = iter(data_loaders[\"sup\"])\n", + "\n", + " for i in range(batches_per_epoch):\n", + "\n", + " # extract the corresponding batch\n", + " (xs, ys) = next(sup_iter)\n", + " xs = xs.cuda()\n", + " ys = ys.cuda()\n", + " # run the inference for each loss with supervised or un-supervised\n", + " # data as arguments\n", + " for loss_id in range(num_losses):\n", + " new_loss = losses[loss_id].step(xs, ys)\n", + " epoch_losses[loss_id] += new_loss\n", + "\n", + " # return the values of all losses\n", + " return epoch_losses\n" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "Ad6H4X0HMEjT", + "colab_type": "code", + "colab": {} + }, + "source": [ + "dataset_zip = np.load(\n", + " '/content/gdrive/My Drive/data-science/causal-ml/projects/dsprites-dataset/dsprites_ndarray_co1sh3sc6or40x32y32_64x64.npz',\n", + " encoding = 'bytes',\n", + " allow_pickle=True\n", + ")\n", + "imgs = dataset_zip['imgs']\n", + "latents_values = dataset_zip['latents_values']\n", + "latents_classes = dataset_zip['latents_classes']\n", + "metadata = dataset_zip['metadata'][()]\n", + "\n", + "latents_sizes = metadata[b'latents_sizes']\n", + "latents_bases = np.concatenate((latents_sizes[::-1].cumprod()[::-1][1:],\n", + " np.array([1,])))\n", + "\n", + "def latent_to_index(latents):\n", + " return np.dot(latents, latents_bases).astype(int)\n", + "\n", + "\n", + "def sample_latent(size=1):\n", + " samples = np.zeros((size, latents_sizes.size))\n", + " for lat_i, lat_size in enumerate(latents_sizes):\n", + " samples[:, lat_i] = np.random.randint(lat_size, size=size)\n", + "\n", + " return samples\n", + "\n", + "# Sample latents randomly\n", + "latents_sampled = sample_latent(size=70000)\n", + "\n", + "# Select images\n", + "indices_sampled = latent_to_index(latents_sampled)\n", + "imgs_sampled = imgs[indices_sampled]\n", + "\n", + "data_loaders = setup_data_loaders(\n", + " imgs_sampled[1000:], latents_sampled[1000:],\n", + " imgs_sampled[:1000], latents_sampled[:1000],\n", + " batch_size=256,\n", + " use_cuda=USE_CUDA\n", + ")" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "Xvn38eq0MEjU", + "colab_type": "code", + "colab": {} + }, + "source": [ + "def get_accuracy(data_loader, classifier_fn, batch_size):\n", + " \"\"\"\n", + " compute the accuracy over the supervised training set or the testing set\n", + " \"\"\"\n", + " predictions, actuals = [], []\n", + "\n", + " # use the appropriate data loader\n", + " \n", + " for (xs, ys) in data_loader:\n", + " xs = xs.cuda()\n", + " ys = ys.cuda()\n", + " # use classification function to compute all predictions for each batch\n", + " predictions.append(classifier_fn(xs))\n", + " actuals.append(ys)\n", + "\n", + " # compute the number of accurate predictions\n", + " accurate_preds = 0\n", + " for pred, act in zip(predictions, actuals):\n", + " for i in range(pred.size(0)):\n", + " v = torch.sum(pred[i] == act[i])\n", + " accurate_preds += (v.item() == 10)\n", + "\n", + " # calculate the accuracy between 0 and 1\n", + " accuracy = (accurate_preds * 1.0) / (len(predictions) * batch_size)\n", + " return accuracy\n", + "\n", + "\n", + "def visualize(s_vae, viz, test_loader):\n", + " if viz:\n", + " plot_conditional_samples_ssvae(s_vae, viz)\n", + " mnist_test_tsne_ssvae(ssvae=s_vae, test_loader=test_loader)" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "_pz3enLBMEjW", + "colab_type": "code", + "colab": {} + }, + "source": [ + "sup_vae = SVAE(\n", + " output_size=6,\n", + " input_size=4096,\n", + " z_dim=50,\n", + " hidden_layers=[500],\n", + " use_cuda=USE_CUDA,\n", + " config_enum=\"parallel\",\n", + " aux_loss_multiplier=46\n", + ")" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "T4kr9m3UMEja", + "colab_type": "code", + "colab": {} + }, + "source": [ + "adam_params = {\"lr\": 0.00042, \"betas\": (0.9, 0.999)}\n", + "optimizer = Adam(adam_params)\n", + "# set up the loss(es) for inference. wrapping the guide in config_enumerate builds\n", + "# the loss as a sum\n", + "# by enumerating each class label for the sampled discrete categorical distribution\n", + "# in the model\n", + "guide = config_enumerate(sup_vae.guide, \"parallel\", expand=True)\n", + "elbo = Trace_ELBO(max_plate_nesting=1)\n", + "loss_basic = SVI(sup_vae.model, guide, optimizer, loss=elbo)\n", + "\n", + "# build a list of all losses considered\n", + "losses = [loss_basic]\n", + "loss_aux = SVI(sup_vae.model_classify, sup_vae.guide_classify, optimizer, loss=elbo)\n", + "losses.append(loss_aux)" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "r0dRPWlWeG64", + "colab_type": "code", + "outputId": "270c8c1e-5d1d-4af3-9072-5b503f0f88f0", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 731 + } + }, + "source": [ + "sup_vae" + ], + "execution_count": 17, + "outputs": [ + { + "output_type": "execute_result", + "data": { + "text/plain": [ + "SVAE(\n", + " (encoder_y): MLP(\n", + " (sequential_mlp): Sequential(\n", + " (0): ConcatModule()\n", + " (1): DataParallel(\n", + " (module): Linear(in_features=4096, out_features=500, bias=True)\n", + " )\n", + " (2): Softplus(beta=1, threshold=20)\n", + " (3): Linear(in_features=500, out_features=6, bias=True)\n", + " (4): Softmax()\n", + " )\n", + " )\n", + " (encoder_z): MLP(\n", + " (sequential_mlp): Sequential(\n", + " (0): ConcatModule()\n", + " (1): DataParallel(\n", + " (module): Linear(in_features=4102, out_features=500, bias=True)\n", + " )\n", + " (2): Softplus(beta=1, threshold=20)\n", + " (3): ListOutModule(\n", + " (0): Sequential(\n", + " (0): Linear(in_features=500, out_features=50, bias=True)\n", + " )\n", + " (1): Sequential(\n", + " (0): Linear(in_features=500, out_features=50, bias=True)\n", + " (1): Exp()\n", + " )\n", + " )\n", + " )\n", + " )\n", + " (decoder): MLP(\n", + " (sequential_mlp): Sequential(\n", + " (0): ConcatModule()\n", + " (1): DataParallel(\n", + " (module): Linear(in_features=56, out_features=500, bias=True)\n", + " )\n", + " (2): Softplus(beta=1, threshold=20)\n", + " (3): Linear(in_features=500, out_features=4096, bias=True)\n", + " (4): Sigmoid()\n", + " )\n", + " )\n", + ")" + ] + }, + "metadata": { + "tags": [] + }, + "execution_count": 17 + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "qfoFJZPVMEjb", + "colab_type": "code", + "outputId": "1c457d2c-6478-412c-99f6-36969a8173f5", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 1000 + } + }, + "source": [ + "sup_num = 60000\n", + "batch_size = 256\n", + "num_epochs = 100\n", + "\n", + "logger = open(\"./tmp.log\", \"w\")\n", + "# Number of supervised number\n", + "\n", + "# initializing local variables to maintain the best validation accuracy\n", + "# seen across epochs over the supervised training set\n", + "# and the corresponding testing set and the state of the networks\n", + "best_valid_acc, corresponding_test_acc = 0.0, 0.0\n", + "\n", + "# run inference for a certain number of epochs\n", + "for i in range(0, num_epochs):\n", + "\n", + " # get the losses for an epoch\n", + " epoch_losses_sup = run_supervized_inference_for_epoch(\n", + " data_loaders,\n", + " losses\n", + " )\n", + "\n", + " # compute average epoch losses i.e. losses per example\n", + " avg_epoch_losses_sup = map(lambda v: v / sup_num, epoch_losses_sup)\n", + "\n", + " # store the loss and validation/testing accuracies in the logfile\n", + " str_loss_sup = \" \".join(map(str, avg_epoch_losses_sup))\n", + "\n", + " str_print = \"{} epoch: avg loss {}\".format(i, \"{}\".format(str_loss_sup))\n", + "\n", + " validation_accuracy = get_accuracy(data_loaders[\"valid\"], sup_vae.classifier, batch_size)\n", + " str_print += \" validation accuracy {}\".format(validation_accuracy)\n", + "\n", + " # this test accuracy is only for logging, this is not used\n", + " # to make any decisions during training\n", + " test_accuracy = get_accuracy(data_loaders[\"test\"], sup_vae.classifier, batch_size)\n", + " str_print += \" test accuracy {}\".format(test_accuracy)\n", + "\n", + " # update the best validation accuracy and the corresponding\n", + " # testing accuracy and the state of the parent module (including the networks)\n", + " if best_valid_acc < validation_accuracy:\n", + " best_valid_acc = validation_accuracy\n", + " corresponding_test_acc = test_accuracy\n", + " \n", + " visualize(sup_vae, None, data_loaders[\"test\"])\n", + "\n", + " print_and_log(logger, str_print)\n", + "\n", + "final_test_accuracy = get_accuracy(data_loaders[\"test\"], sup_vae.classifier, batch_size)\n", + "print_and_log(logger, \"best validation accuracy {} corresponding testing accuracy {} \"\n", + " \"last testing accuracy {}\".format(best_valid_acc, corresponding_test_acc, final_test_accuracy))\n", + "\n", + "# close the logger file object if we opened it earlier\n", + "logger.close()" + ], + "execution_count": 36, + "outputs": [ + { + "output_type": "stream", + "text": [ + "0 epoch: avg loss 73.4194898127238 10.337283599853516 validation accuracy 0.0 test accuracy 0.0\n", + "1 epoch: avg loss 73.1135249282837 9.959152315266927 validation accuracy 0.0 test accuracy 0.0\n", + "2 epoch: avg loss 72.90056605784098 9.599737705485026 validation accuracy 0.0 test accuracy 0.0\n", + "3 epoch: avg loss 72.72961992340088 9.309077065022786 validation accuracy 0.0 test accuracy 0.0\n", + "4 epoch: avg loss 72.53395691884359 8.995561258951822 validation accuracy 0.0 test accuracy 0.0\n", + "5 epoch: avg loss 72.29241495920817 8.699117222086588 validation accuracy 0.0 test accuracy 0.0\n", + "6 epoch: avg loss 72.0966392674764 8.405718975830078 validation accuracy 0.0 test accuracy 0.0\n", + "7 epoch: avg loss 71.84623018544515 8.126094775390625 validation accuracy 0.0 test accuracy 0.0\n", + "8 epoch: avg loss 71.72414229482015 7.864214937337239 validation accuracy 0.0 test accuracy 0.0\n", + "9 epoch: avg loss 71.49476098785401 7.599710852050781 validation accuracy 0.0 test accuracy 0.0\n", + "10 epoch: avg loss 71.34665864410401 7.383735764567057 validation accuracy 0.0 test accuracy 0.0\n", + "11 epoch: avg loss 71.24757972157796 7.4970786661783855 validation accuracy 0.0 test accuracy 0.0\n", + "12 epoch: avg loss 71.0358139175415 7.385145209757487 validation accuracy 0.0 test accuracy 0.0\n", + "13 epoch: avg loss 70.79814100087484 6.938219022623698 validation accuracy 0.0 test accuracy 0.0\n", + "14 epoch: avg loss 70.690160206604 6.606387151082357 validation accuracy 0.0 test accuracy 0.0\n", + "15 epoch: avg loss 70.46839635569255 6.533454721069336 validation accuracy 0.0 test accuracy 0.0\n", + "16 epoch: avg loss 70.39087748362223 6.275628011067709 validation accuracy 0.0 test accuracy 0.0\n", + "17 epoch: avg loss 70.16816714019775 6.026749960327148 validation accuracy 0.0 test accuracy 0.0\n", + "18 epoch: avg loss 70.02727948557536 5.86120785929362 validation accuracy 0.0 test accuracy 0.0\n", + "19 epoch: avg loss 69.95993371734619 5.694548964436849 validation accuracy 0.0 test accuracy 0.0\n", + "20 epoch: avg loss 69.76063533681234 5.675880317179362 validation accuracy 0.0 test accuracy 0.0\n", + "21 epoch: avg loss 69.65645001780192 5.492274950154623 validation accuracy 0.0 test accuracy 0.0\n", + "22 epoch: avg loss 69.47397186838786 5.404019834391276 validation accuracy 0.0 test accuracy 0.0\n", + "23 epoch: avg loss 69.3778068456014 5.669653151448568 validation accuracy 0.0 test accuracy 0.0\n", + "24 epoch: avg loss 69.24922564442953 5.432655401611328 validation accuracy 0.0 test accuracy 0.0\n", + "25 epoch: avg loss 69.02797335764566 5.363767221069336 validation accuracy 0.0 test accuracy 0.0\n", + "26 epoch: avg loss 68.97380993804931 5.430132021077474 validation accuracy 0.0 test accuracy 0.0\n", + "27 epoch: avg loss 68.8614875096639 5.341963358561198 validation accuracy 0.0 test accuracy 0.0\n", + "28 epoch: avg loss 68.72352076975504 5.49090717569987 validation accuracy 0.0 test accuracy 0.0\n", + "29 epoch: avg loss 68.57598680165609 5.599548092651367 validation accuracy 0.0 test accuracy 0.0\n", + "30 epoch: avg loss 68.42424849599202 5.55614609375 validation accuracy 0.0 test accuracy 0.0\n", + "31 epoch: avg loss 68.29375264638264 5.479485826619466 validation accuracy 0.0 test accuracy 0.0\n", + "32 epoch: avg loss 68.1904503595988 5.2717307159423825 validation accuracy 0.0 test accuracy 0.0\n", + "33 epoch: avg loss 68.08769537913004 5.024212093098958 validation accuracy 0.0 test accuracy 0.0\n", + "34 epoch: avg loss 67.929296534729 4.744880869547526 validation accuracy 0.0 test accuracy 0.0\n", + "35 epoch: avg loss 67.83656690419515 4.545162293497722 validation accuracy 0.0 test accuracy 0.0\n", + "36 epoch: avg loss 67.77494875640869 4.303533932495117 validation accuracy 0.0 test accuracy 0.0\n", + "37 epoch: avg loss 67.67430809885661 4.098325904337565 validation accuracy 0.0 test accuracy 0.0\n", + "38 epoch: avg loss 67.48732435862223 3.956320953369141 validation accuracy 0.0 test accuracy 0.0\n", + "39 epoch: avg loss 67.43751215159098 3.842155121358236 validation accuracy 0.0 test accuracy 0.0\n", + "40 epoch: avg loss 67.3197662612915 3.7160155166625977 validation accuracy 0.0 test accuracy 0.0\n", + "41 epoch: avg loss 67.20157636871338 3.5679363571166993 validation accuracy 0.0 test accuracy 0.0\n", + "42 epoch: avg loss 67.11693908843993 3.5223314137776693 validation accuracy 0.0 test accuracy 0.0\n", + "43 epoch: avg loss 66.99766852366129 3.3062897201538086 validation accuracy 0.0 test accuracy 0.0\n", + "44 epoch: avg loss 66.94425606435139 3.274656864420573 validation accuracy 0.0 test accuracy 0.0\n", + "45 epoch: avg loss 66.87721158192953 3.2535403299967447 validation accuracy 0.0 test accuracy 0.0\n", + "46 epoch: avg loss 66.76631597646077 3.224982407124837 validation accuracy 0.0 test accuracy 0.0\n", + "47 epoch: avg loss 66.68057815093994 3.1721942459106445 validation accuracy 0.0 test accuracy 0.0\n", + "48 epoch: avg loss 66.53252608388264 2.9452966857910154 validation accuracy 0.0 test accuracy 0.0\n", + "49 epoch: avg loss 66.4851595067342 2.8821236943562827 validation accuracy 0.0 test accuracy 0.0\n", + "50 epoch: avg loss 66.37938383127849 3.09201468556722 validation accuracy 0.0 test accuracy 0.0\n", + "51 epoch: avg loss 66.31664015452067 3.177724250793457 validation accuracy 0.0 test accuracy 0.0\n", + "52 epoch: avg loss 66.19034068349202 2.9874297996520998 validation accuracy 0.0 test accuracy 0.0\n", + "53 epoch: avg loss 66.05708732248942 3.232627372741699 validation accuracy 0.0 test accuracy 0.0\n", + "54 epoch: avg loss 66.01276608225504 3.173233707173665 validation accuracy 0.0 test accuracy 0.0\n", + "55 epoch: avg loss 65.88708648427327 3.110121321105957 validation accuracy 0.0 test accuracy 0.0\n", + "56 epoch: avg loss 65.81027199045818 3.161735363260905 validation accuracy 0.0 test accuracy 0.0\n", + "57 epoch: avg loss 65.73718862457275 3.3055569646199543 validation accuracy 0.0 test accuracy 0.0\n", + "58 epoch: avg loss 65.66992208811442 3.2017634826660157 validation accuracy 0.0 test accuracy 0.0\n", + "59 epoch: avg loss 65.60472373199462 3.028648991394043 validation accuracy 0.0 test accuracy 0.0\n", + "60 epoch: avg loss 65.51507672678629 3.0213762514750164 validation accuracy 0.0 test accuracy 0.0\n", + "61 epoch: avg loss 65.38191290842693 3.0188735694885254 validation accuracy 0.0 test accuracy 0.0\n", + "62 epoch: avg loss 65.30616182607015 2.9055699577331544 validation accuracy 0.0 test accuracy 0.0\n", + "63 epoch: avg loss 65.21349840647379 2.852216015370687 validation accuracy 0.0 test accuracy 0.0\n", + "64 epoch: avg loss 65.18253641916911 2.813871074167887 validation accuracy 0.0 test accuracy 0.0\n", + "65 epoch: avg loss 65.12543276519776 2.7915143440246584 validation accuracy 0.0 test accuracy 0.0\n", + "66 epoch: avg loss 65.08151169586182 2.802122728474935 validation accuracy 0.0 test accuracy 0.0\n", + "67 epoch: avg loss 64.97802228342692 2.7456317456563313 validation accuracy 0.0 test accuracy 0.0\n", + "68 epoch: avg loss 64.95747393544515 2.843347436014811 validation accuracy 0.0 test accuracy 0.0\n", + "69 epoch: avg loss 64.95950721995035 2.7493250696818032 validation accuracy 0.0 test accuracy 0.0\n", + "70 epoch: avg loss 64.85206590321859 2.549251330820719 validation accuracy 0.0 test accuracy 0.0\n", + "71 epoch: avg loss 64.77169641265868 2.617051190185547 validation accuracy 0.0 test accuracy 0.0\n", + "72 epoch: avg loss 64.71160574696859 2.5421745646158853 validation accuracy 0.0 test accuracy 0.0\n", + "73 epoch: avg loss 64.58634377593994 2.465168037160238 validation accuracy 0.0 test accuracy 0.0\n", + "74 epoch: avg loss 64.57885850575765 2.361467325337728 validation accuracy 0.0 test accuracy 0.0\n", + "75 epoch: avg loss 64.42846132151286 2.4082817609151204 validation accuracy 0.0 test accuracy 0.0\n", + "76 epoch: avg loss 64.38538060048421 2.298162012990316 validation accuracy 0.0 test accuracy 0.0\n", + "77 epoch: avg loss 64.3141145767212 2.2881553919474285 validation accuracy 0.0 test accuracy 0.0\n", + "78 epoch: avg loss 64.24270374501546 2.2224127756754557 validation accuracy 0.0 test accuracy 0.0\n", + "79 epoch: avg loss 64.09710526682535 2.1379135854085285 validation accuracy 0.0 test accuracy 0.0\n", + "80 epoch: avg loss 64.07882555491129 2.261145238494873 validation accuracy 0.0 test accuracy 0.0\n", + "81 epoch: avg loss 64.05341328277588 2.1424091829935707 validation accuracy 0.0 test accuracy 0.0\n", + "82 epoch: avg loss 63.90139794260661 2.0546640805562335 validation accuracy 0.0 test accuracy 0.0\n", + "83 epoch: avg loss 63.886747120666506 2.0148360023498535 validation accuracy 0.0 test accuracy 0.0\n", + "84 epoch: avg loss 63.863132594299316 2.0213047768910726 validation accuracy 0.0 test accuracy 0.0\n", + "85 epoch: avg loss 63.72724614003499 1.9374477088928224 validation accuracy 0.0 test accuracy 0.0\n", + "86 epoch: avg loss 63.707502190653486 1.969299099858602 validation accuracy 0.0 test accuracy 0.0\n", + "87 epoch: avg loss 63.60284013824463 1.9574982669830323 validation accuracy 0.0 test accuracy 0.0\n", + "88 epoch: avg loss 63.60544111480713 2.1192375803629555 validation accuracy 0.0 test accuracy 0.0\n", + "89 epoch: avg loss 63.47691573232015 2.071177163950602 validation accuracy 0.0 test accuracy 0.0\n", + "90 epoch: avg loss 63.45331827952067 1.870301713816325 validation accuracy 0.0 test accuracy 0.0\n", + "91 epoch: avg loss 63.43649330393473 1.7918146341959635 validation accuracy 0.0 test accuracy 0.0\n", + "92 epoch: avg loss 63.37312573394775 1.8475859204610188 validation accuracy 0.0 test accuracy 0.0\n", + "93 epoch: avg loss 63.23665246734619 1.9314768585205078 validation accuracy 0.0 test accuracy 0.0\n", + "94 epoch: avg loss 63.206928472391766 1.8032865263621012 validation accuracy 0.0 test accuracy 0.0\n", + "95 epoch: avg loss 63.235852890523276 1.798199406305949 validation accuracy 0.0 test accuracy 0.0\n", + "96 epoch: avg loss 63.176236850484216 1.788236808013916 validation accuracy 0.0 test accuracy 0.0\n", + "97 epoch: avg loss 63.0884791112264 1.8939534503936768 validation accuracy 0.0 test accuracy 0.0\n", + "98 epoch: avg loss 63.16810416819254 1.7823999228159586 validation accuracy 0.0 test accuracy 0.0\n", + "99 epoch: avg loss 63.05663078358968 1.8215582843780518 validation accuracy 0.0 test accuracy 0.0\n", + "best validation accuracy 0.0 corresponding testing accuracy 0.0 last testing accuracy 0.0\n" + ], + "name": "stdout" + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "ei0yHxTNMEjf", + "colab_type": "code", + "colab": {} + }, + "source": [ + "def reconstruct_image_w_label(xs, ys, vae):\n", + " # backward\n", + " xs = xs.cuda()\n", + " ys = ys.cuda()\n", + " sim_z_loc, sim_z_scale = vae.encoder_z([xs, ys])\n", + " zs = dist.Normal(sim_z_loc, sim_z_scale).to_event(1).sample()\n", + " # forward\n", + " zs = zs.cuda()\n", + " ys = ys.cuda()\n", + " loc = vae.decoder([zs, ys])\n", + " return dist.Bernoulli(loc).to_event(1).sample()\n", + "\n", + "def reconstruct_image(xs, vae):\n", + " # backward\n", + " xs = xs.cuda()\n", + " sim_ys = vae.encoder_y(xs)\n", + " sim_z_loc, sim_z_scale = vae.encoder_z([xs, sim_ys])\n", + " zs = dist.Normal(sim_z_loc, sim_z_scale).to_event(1).sample()\n", + " # forward\n", + " zs = zs.cuda()\n", + " loc = vae.decoder([zs, ys.cuda()])\n", + " return dist.Bernoulli(loc).to_event(1).sample()\n", + "\n", + "def convert_back(x):\n", + " return x.cpu().reshape(-1, 64, 64).numpy().astype(np.uint8)\n", + "\n", + "def show_images_grid(imgs_, num_images):\n", + " ncols = int(np.ceil(num_images**0.5))\n", + " nrows = int(np.ceil(num_images / ncols))\n", + " _, axes = plt.subplots(ncols, nrows, figsize=(nrows * 3, ncols * 3))\n", + " axes = axes.flatten()\n", + " for ax_i, ax in enumerate(axes):\n", + " if ax_i < num_images:\n", + " ax.imshow(imgs_[ax_i].reshape(64, 64), cmap='Greys_r', interpolation='nearest')\n", + " ax.set_xticks([])\n", + " ax.set_yticks([])\n", + " else:\n", + " ax.axis('off')\n", + "\n", + "def viz_images(imgs, n):\n", + " imgs_ = []\n", + " for i, x in enumerate(imgs):\n", + " if i > n:\n", + " break\n", + " imgs_.append(convert_back(x))\n", + " show_images_grid(np.array(imgs_), n)" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "markdown", + "metadata": { + "id": "zm-CLHjBMEjg", + "colab_type": "text" + }, + "source": [ + "Four original images " + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "g7trYLCqMEjh", + "colab_type": "code", + "colab": {} + }, + "source": [ + "sup_iter = iter(data_loaders[\"sup\"])\n", + "xs, ys = next(sup_iter)" + ], + "execution_count": 0, + "outputs": [] + }, + { + "cell_type": "code", + "metadata": { + "id": "vVQmHBd_MEji", + "colab_type": "code", + "outputId": "d480f485-19bc-4c12-c317-735bdafa8eb1", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 392 + } + }, + "source": [ + "print(\"With y and x:\")\n", + "xs_sim1 = reconstruct_image_w_label(xs, ys, sup_vae)\n", + "viz_images(xs_sim1, 4)" + ], + "execution_count": 38, + "outputs": [ + { + "output_type": "stream", + "text": [ + "With y and x:\n" + ], + "name": "stdout" + }, + { + "output_type": "display_data", + "data": { + "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWsAAAFmCAYAAACr2LumAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSUVORK5CYII=\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "fEvI9XGTMEjk", + "colab_type": "code", + "outputId": "b7db84ee-5957-4f0c-c35f-e9e7f42c2ab6", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 392 + } + }, + "source": [ + "print(\"With just x:\")\n", + "# backward\n", + "xs_sim2 = reconstruct_image(xs, sup_vae)\n", + "viz_images(xs_sim2, 4)" + ], + "execution_count": 39, + "outputs": [ + { + "output_type": "stream", + "text": [ + "With just x:\n" + ], + "name": "stdout" + }, + { + "output_type": "display_data", + "data": { + "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWsAAAFmCAYAAACr2LumAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nP3rLqigPu5+WAAAAAElFTkSuQmCC\n", + "text/plain": [ + "
" + ] + }, + "metadata": { + "tags": [] + } + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "H_OtIYKxdNp-", + "colab_type": "code", + "outputId": "edf1ceff-99c8-421c-86d9-b4a3766e64b5", + "colab": { + "base_uri": "https://localhost:8080/", + "height": 35 + } + }, + "source": [ + "xs_sim2.shape" + ], + "execution_count": 0, + "outputs": [ + { + "output_type": "execute_result", + "data": { + "text/plain": [ + "torch.Size([256, 4096])" + ] + }, + "metadata": { + "tags": [] + }, + "execution_count": 44 + } + ] + }, + { + "cell_type": "code", + "metadata": { + "id": "tCZoEdLh0rET", + "colab_type": "code", + "colab": {} + }, + "source": [ + "" + ], + "execution_count": 0, + "outputs": [] + } + ] +} \ No newline at end of file