gan.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "COMET INFO: Experiment is live on comet.ml https://www.comet.ml/shir994/gan-test/1d786bb2f69b49df925df1292d61815f\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from comet_ml import Experiment\n",
    "\n",
    "hyper_params = {    \n",
    "    \"TASK\": 4,\n",
    "    \"batch_size\": 128, # initially was 64\n",
    "    \"NOISE_DIM\": 30,\n",
    "    \"num_epochs\": 300,\n",
    "    \"learning_rate\": 0.001\n",
    "}\n",
    "experiment = Experiment(project_name=\"gan-test\", workspace=\"shir994\")\n",
    "experiment.log_parameters(hyper_params)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "env: CUDA_VISIBLE_DEVICES=0\n",
      "env: LIBRARY_PATH=/usr/local/cuda/lib64\n"
     ]
    }
   ],
   "source": [
    "%env CUDA_VISIBLE_DEVICES=0\n",
    "%env LIBRARY_PATH=/usr/local/cuda/lib64"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from model import YModel, R\n",
    "from gan.gan import Generator, Discriminator, WSDiscriminator, GANLosses\n",
    "from gan.metrics import Metrics\n",
    "from utils import sample_noise, iterate_minibatches, generate_data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# GANS"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim\n",
    "import torch\n",
    "\n",
    "import numpy as np\n",
    "from tqdm import trange\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "device = torch.device(\"cuda\", 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'hyper_params' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-4-66a95636b066>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mTASK\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mhyper_params\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'TASK'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m: name 'hyper_params' is not defined"
     ]
    }
   ],
   "source": [
    "TASK = hyper_params['TASK']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'hyper_params' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-5-ed49d81a866c>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mgenerator\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mGenerator\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mhyper_params\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'NOISE_DIM'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mout_dim\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mto\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdevice\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      2\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mTASK\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m4\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m     \u001b[0mdiscriminator\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mWSDiscriminator\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0min_dim\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mto\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdevice\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'hyper_params' is not defined"
     ]
    }
   ],
   "source": [
    "generator = Generator(hyper_params['NOISE_DIM'], out_dim = 1).to(device)\n",
    "\n",
    "if TASK == 4:\n",
    "    discriminator = WSDiscriminator(in_dim=1).to(device)\n",
    "else:\n",
    "    discriminator = Discriminator(in_dim=1).to(device)\n",
    "\n",
    "g_optimizer = optim.Adam(generator.parameters(),     lr=hyper_params['learning_rate'], betas=(0.5, 0.999))\n",
    "d_optimizer = optim.Adam(discriminator.parameters(), lr=hyper_params['learning_rate'], betas=(0.5, 0.999))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_sampler = YModel()\n",
    "fixed_noise = torch.Tensor(sample_noise(10000, hyper_params['NOISE_DIM'])).to(device)\n",
    "data, inputs = generate_data(y_sampler, device, 10000, mu_range=(-30, 30))\n",
    "metric_calc = Metrics((-50, 50), 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(data.cpu().numpy(), bins=100);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "def draw_conditional_samples(mu_range, x_range):\n",
    "    f = plt.figure(figsize=(21,16))\n",
    "\n",
    "    for i in range(4):\n",
    "        x = torch.Tensor([float(x_range[i])] * 10000).to(device)\n",
    "        for j in range(4):\n",
    "            mu = torch.Tensor([float(mu_range[j])] * 10000).to(device)\n",
    "            plt.subplot(4,4, i*4 + j + 1)\n",
    "\n",
    "            y_sampler.make_condition_sample({'mu': mu, 'X':x})\n",
    "            data = y_sampler.condition_sample().detach().cpu().numpy()\n",
    "\n",
    "            plt.hist(data, bins=100, density=True, label='true');\n",
    "            plt.hist(generator(fixed_noise, torch.stack([mu,x],dim=1)).detach().cpu().numpy(),\n",
    "                     bins=100, color='g', density=True, alpha=0.5, label='gan');\n",
    "            plt.grid()\n",
    "            plt.legend()\n",
    "            if j == 0:\n",
    "                plt.ylabel(\"x={:.3f}\".format(x[0].item()), fontsize=19)\n",
    "            if i == 0:\n",
    "                plt.title(\"mu={:.3f}\".format(mu[0].item()), fontsize=19)            \n",
    "    return f\n",
    "\n",
    "def draw_mu_samples(mu_range):\n",
    "    f = plt.figure(figsize=(21,16))\n",
    "    for i in range(4):\n",
    "        for j in range(3):\n",
    "            plt.subplot(4,3, i*3 + j + 1)\n",
    "            mu = torch.tensor([float(mu_range[i*3 + j])] * 10000)#.to(device)\n",
    "            y_sampler.make_condition_sample({'mu': mu})\n",
    "\n",
    "            x = y_sampler.x_dist.sample(mu.shape).to(device)\n",
    "            mu = mu.to(device)\n",
    "\n",
    "            plt.hist(y_sampler.condition_sample().cpu().numpy(), bins=100, density=True, label='true');\n",
    "            plt.hist(generator(fixed_noise, torch.stack([mu,x],dim=1)).detach().cpu().numpy(),\n",
    "                     bins=100, color='g', density=True, alpha=0.5, label='gan');\n",
    "            plt.grid()\n",
    "            plt.legend()\n",
    "            plt.title(\"mu={:.3f} \".format(mu_range[i*3 + j])) \n",
    "    return f\n",
    "\n",
    "def calculate_validation_metrics(mu_range, epoch):\n",
    "    js = []\n",
    "    for _mu in mu_range:\n",
    "        mu = torch.tensor([float(_mu)] * 10000)\n",
    "        y_sampler.make_condition_sample({'mu': mu})\n",
    "\n",
    "        x = y_sampler.x_dist.sample(mu.shape).to(device)\n",
    "        mu = mu.to(device)\n",
    "\n",
    "        js.append(metric_calc.compute_JS(y_sampler.condition_sample().cpu(),\n",
    "                  generator(fixed_noise, torch.stack([mu,x],dim=1)).detach().cpu()).item())\n",
    "#                 average_mu_JS.append(np.mean(js))\n",
    "    train_data_js = metric_calc.compute_JS(data.cpu(), generator(fixed_noise, inputs).detach().cpu())\n",
    "#                 train_data_JS.append(train_data_js)\n",
    "\n",
    "    experiment.log_metric(\"average_mu_JS\", np.mean(js), step=epoch)\n",
    "    experiment.log_metric(\"train_data_JS\", train_data_js, step=epoch)    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [],
   "source": [
    "from IPython import display\n",
    "def run_training():\n",
    "    #plt.xlim(lims)\n",
    "    #plt.ylim(lims)\n",
    "\n",
    "    # ===========================\n",
    "    # IMPORTANT PARAMETER:\n",
    "    # Number of D updates per G update\n",
    "    # ===========================\n",
    "    k_d, k_g = 4, 1\n",
    "\n",
    "    dis_loss, gen_loss = [], []\n",
    "    train_data_JS = []\n",
    "    average_mu_JS = []\n",
    "    gan_losses = GANLosses(TASK, device)\n",
    "\n",
    "    try:\n",
    "        with experiment.train():\n",
    "            for epoch in range(hyper_params['num_epochs']):\n",
    "                dis_epoch_loss = []\n",
    "                gen_epoch_loss = []\n",
    "\n",
    "                for input_data, inputs_batch in iterate_minibatches(data, hyper_params['batch_size'], y=inputs):\n",
    "                    # Optimize D\n",
    "                    for _ in range(k_d):\n",
    "                        # Sample noise\n",
    "                        noise = torch.Tensor(sample_noise(len(input_data), hyper_params['NOISE_DIM'])).to(device)\n",
    "\n",
    "                        # Do an update\n",
    "                        inp_data = input_data.to(device)\n",
    "                        data_gen = generator(noise, inputs_batch)\n",
    "\n",
    "                        loss = gan_losses.d_loss(discriminator(data_gen, inputs_batch),\n",
    "                                                discriminator(inp_data, inputs_batch))\n",
    "                        if TASK == 4:\n",
    "                            grad_penalty = gan_losses.calc_gradient_penalty(discriminator,\n",
    "                                                                            data_gen.data,\n",
    "                                                                            inputs_batch.data,\n",
    "                                                                            inp_data.data)\n",
    "                            loss += grad_penalty\n",
    "\n",
    "                        d_optimizer.zero_grad()\n",
    "                        loss.backward()\n",
    "                        d_optimizer.step()\n",
    "\n",
    "                        if TASK == 3:                    \n",
    "                            for p in discriminator.parameters():\n",
    "                                p.data.clamp_(clamp_lower, clamp_upper)\n",
    "                    dis_epoch_loss.append(loss.item())\n",
    "\n",
    "                    # Optimize G\n",
    "                    for _ in range(k_g):\n",
    "                        # Sample noise\n",
    "                        noise = torch.Tensor(sample_noise(len(input_data), hyper_params['NOISE_DIM'])).to(device)\n",
    "\n",
    "                        # Do an update\n",
    "                        data_gen = generator(noise, inputs_batch)\n",
    "                        loss = gan_losses.g_loss(discriminator(data_gen, inputs_batch))\n",
    "                        g_optimizer.zero_grad()\n",
    "                        loss.backward()\n",
    "                        g_optimizer.step()\n",
    "                    gen_epoch_loss.append(loss.item())\n",
    "                \n",
    "                experiment.log_metric(\"d_loss\", np.mean(dis_epoch_loss), step=epoch)\n",
    "                experiment.log_metric(\"g_loss\", np.mean(gen_epoch_loss), step=epoch)\n",
    "#                 dis_loss.append(np.mean(dis_epoch_loss))\n",
    "#                 gen_loss.append(np.mean(gen_epoch_loss))\n",
    "                \n",
    "                calculate_validation_metrics(list(range(-30, 30, 2)), epoch)\n",
    "                \n",
    "                mu_range = list(range(-10, 10, 4))\n",
    "                x_range = list(range(1, 13, 3))\n",
    "                f = draw_conditional_samples(mu_range, x_range)\n",
    "                experiment.log_figure(\"conditional_samples\", f)\n",
    "                \n",
    "                mu_range = list(range(-10, 13, 2))\n",
    "                f = draw_mu_samples(mu_range)\n",
    "                experiment.log_figure(\"mu_samples\", f)\n",
    "                \n",
    "        return dis_loss, gen_loss, train_data_JS, average_mu_JS\n",
    "    except KeyboardInterrupt:\n",
    "        pass"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [],
   "source": [
    "%reload_ext autoreload"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/79 [00:00<?, ?it/s]/mnt/shirobokov/conda/envs/pytorch/lib/python3.7/site-packages/torch/nn/functional.py:1320: UserWarning: nn.functional.tanh is deprecated. Use torch.tanh instead.\n",
      "  warnings.warn(\"nn.functional.tanh is deprecated. Use torch.tanh instead.\")\n",
      "100%|██████████| 79/79 [00:02<00:00, 32.74it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 31.59it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 31.61it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 31.34it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 30.69it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 31.23it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 31.47it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 31.04it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 30.37it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 31.26it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 32.02it/s]\n",
      "/mnt/shirobokov/conda/envs/pytorch/lib/python3.7/site-packages/matplotlib/pyplot.py:514: RuntimeWarning: More than 20 figures have been opened. Figures created through the pyplot interface (`matplotlib.pyplot.figure`) are retained until explicitly closed and may consume too much memory. (To control this warning, see the rcParam `figure.max_open_warning`).\n",
      "  max_open_warning, RuntimeWarning)\n",
      "100%|██████████| 79/79 [00:02<00:00, 31.12it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 30.39it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 30.32it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 30.66it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 30.37it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 31.69it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 29.70it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 30.66it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 31.02it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 30.47it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 30.17it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 30.20it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 31.58it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 32.35it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 31.86it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 31.64it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 33.50it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 30.70it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 31.59it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 30.73it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 30.38it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 31.17it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 29.85it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 29.64it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 29.46it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 30.33it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 29.96it/s]\n",
      "100%|██████████| 79/79 [00:02<00:00, 30.46it/s]\n",
      "COMET INFO: ----------------------------\n",
      "COMET INFO: Comet.ml Experiment Summary:\n",
      "COMET INFO:   Data:\n",
      "COMET INFO:     url: https://www.comet.ml/shir994/gan-test/45365d46f7d04305b618d3f95845f644\n",
      "COMET INFO:   Metrics:\n",
      "COMET INFO:               sys.gpu.0.free_memory: 11985420288\n",
      "COMET INFO:           sys.gpu.0.gpu_utilization: 0\n",
      "COMET INFO:              sys.gpu.0.total_memory: 11996954624\n",
      "COMET INFO:               sys.gpu.0.used_memory: 11534336\n",
      "COMET INFO:               sys.gpu.1.free_memory: 10895425536\n",
      "COMET INFO:           sys.gpu.1.gpu_utilization: 0\n",
      "COMET INFO:              sys.gpu.1.total_memory: 11996954624\n",
      "COMET INFO:               sys.gpu.1.used_memory: 1101529088\n",
      "COMET INFO:               sys.gpu.2.free_memory: 11985420288\n",
      "COMET INFO:           sys.gpu.2.gpu_utilization: 0\n",
      "COMET INFO:              sys.gpu.2.total_memory: 11996954624\n",
      "COMET INFO:               sys.gpu.2.used_memory: 11534336\n",
      "COMET INFO:               sys.gpu.3.free_memory: 11714494464\n",
      "COMET INFO:           sys.gpu.3.gpu_utilization: 0\n",
      "COMET INFO:              sys.gpu.3.total_memory: 11996954624\n",
      "COMET INFO:               sys.gpu.3.used_memory: 282460160\n",
      "COMET INFO:                 train_average_mu_JS: 0.019432197402541835\n",
      "COMET INFO:                        train_d_loss: -0.7650495500692839\n",
      "COMET INFO:                        train_g_loss: -45.57585267175602\n",
      "COMET INFO:                          train_loss: -0.43976402282714844\n",
      "COMET INFO:         train_sys.gpu.0.free_memory: 11985420288\n",
      "COMET INFO:     train_sys.gpu.0.gpu_utilization: 0\n",
      "COMET INFO:         train_sys.gpu.0.used_memory: 11534336\n",
      "COMET INFO:         train_sys.gpu.1.free_memory: 10895425536\n",
      "COMET INFO:     train_sys.gpu.1.gpu_utilization: 32\n",
      "COMET INFO:         train_sys.gpu.1.used_memory: 1101529088\n",
      "COMET INFO:         train_sys.gpu.2.free_memory: 11985420288\n",
      "COMET INFO:     train_sys.gpu.2.gpu_utilization: 0\n",
      "COMET INFO:         train_sys.gpu.2.used_memory: 11534336\n",
      "COMET INFO:         train_sys.gpu.3.free_memory: 11673993216\n",
      "COMET INFO:     train_sys.gpu.3.gpu_utilization: 0\n",
      "COMET INFO:         train_sys.gpu.3.used_memory: 322961408\n",
      "COMET INFO:                 train_train_data_JS: tensor(0.0051)\n",
      "COMET INFO:   Uploads:\n",
      "COMET INFO:      assets: 0\n",
      "COMET INFO:     figures: 76\n",
      "COMET INFO:      images: 0\n",
      "COMET INFO: ----------------------------\n",
      "COMET INFO: Uploading stats to Comet before program termination (may take several seconds)\n"
     ]
    }
   ],
   "source": [
    "losses = run_training()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#  Conditional GAN simple results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "mu_range = list(range(-10, 10, 4))\n",
    "x_range = list(range(1, 13, 3))\n",
    "f = draw_conditional_samples(mu_range, x_range)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Unconditional samples"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "mu_range = list(range(-10, 13, 2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "draw_mu_samples"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Test on mu outside of training domain"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'mu=-80.000, x=1.000')"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "input_mu = torch.full(size=(10000, ), fill_value=-80.).to(device)\n",
    "input_x = torch.full(size=(10000, ), fill_value=1.).to(device)\n",
    "\n",
    "y_sampler.make_condition_sample({'mu': input_mu, 'X':input_x})\n",
    "data = y_sampler.condition_sample().detach().cpu().numpy() \n",
    "            \n",
    "plt.hist(data, bins=100, normed=True, label='true');\n",
    "plt.hist(generator(fixed_noise, torch.stack([input_mu, input_x], dim=1)).detach().cpu().numpy(),\n",
    "     bins=100, color='g', density=True, alpha=0.5, label='gan');\n",
    "plt.grid()\n",
    "plt.legend()\n",
    "plt.title(\"mu={:.3f}, x={:.3f}\".format(input_mu[0].item(), input_x[0].item()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1512x1152 with 12 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(21,16))\n",
    "mu_range = list(range(22, 60, 2))\n",
    "for i in range(4):\n",
    "    for j in range(3):\n",
    "        plt.subplot(4,3, i*3 + j + 1)\n",
    "        input_mu = torch.full(size=(10000, ), fill_value=mu_range[i*3 + j])#.to(device)\n",
    "        input_x = y_sampler.x_dist.sample(input_mu.shape).to(device)\n",
    "\n",
    "        y_sampler.make_condition_sample({'mu': input_mu})\n",
    "        data = y_sampler.condition_sample().detach().cpu().numpy()\n",
    "\n",
    "        input_mu = input_mu.to(device)\n",
    "\n",
    "        plt.hist(data, bins=100, normed=True, label='true');\n",
    "        plt.hist(generator(fixed_noise, torch.stack([input_mu, input_x], dim=1)).detach().cpu().numpy(),\n",
    "             bins=100, color='g', density=True, alpha=0.5, label='gan');\n",
    "        plt.grid()\n",
    "        plt.legend()\n",
    "        plt.title(\"mu={:.3f}\".format(input_mu[0].item()))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Implementing simple marginalisation ?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "mu = torch.full((10000,), 0.).to(device)\n",
    "x = y_sampler.x_dist.sample([mu.shape[0]]).to(device)\n",
    "plt.hist(generator(fixed_noise, torch.stack([mu,x], dim=1)).detach().cpu().numpy(), bins=100);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "#torch.save(generator, \"new_gan.pt\")\n",
    "#generator = torch.load(\"new_gan.pt\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "from copy import deepcopy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "class InputOptimisation(nn.Module):\n",
    "    def __init__(self, generator_model):\n",
    "        super().__init__()\n",
    "        self.gen = deepcopy(generator_model)\n",
    "        for param in self.gen.parameters():\n",
    "            param.requires_grad = False\n",
    "        self.gen.eval()\n",
    "        \n",
    "#     def generate_random_samples(data, n_bins=300):\n",
    "#         hist, bins = np.histogram(data, bins=n_bins);\n",
    "\n",
    "#         bin_midpoints = bins[:-1] + np.diff(bins)/2\n",
    "#         cdf = np.cumsum(hist)\n",
    "#         cdf = cdf / cdf[-1]\n",
    "\n",
    "#         samples = np.random.rand(100000)\n",
    "#         value_bins = np.searchsorted(cdf, samples)\n",
    "#         random_from_cdf = bin_midpoints[value_bins]        \n",
    "        \n",
    "        \n",
    "    def forward(self, noise, inputs):\n",
    "#         y_given_mu = []\n",
    "#         for val in noise:\n",
    "#             x = y_sampler.x_dist.sample([mu.shape[0]]).to(device)\n",
    "#             local_noise = val.reshape(1,-1).repeat(mu.shape[0],1).to(device)\n",
    "#             y = self.gen(local_noise, torch.stack([mu,x],dim=1)).mean()\n",
    "#             y_given_mu.append(y)\n",
    "\n",
    "#         return torch.stack(y_given_mu)\n",
    "        return self.gen(noise, inputs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "io_model = InputOptimisation(generator)\n",
    "\n",
    "lr = 0.1\n",
    "# io_optimizer = optim.Adam(io_model.parameters(), lr=lr)\n",
    "\n",
    "sample_size = 1000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "mu_optim = torch.tensor(50.)\n",
    "\n",
    "mu_optim = mu_optim.repeat(sample_size).to(device)\n",
    "mu_optim.requires_grad=True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "# %%time\n",
    "# noise = torch.Tensor(sample_noise(sample_size)).to(device)\n",
    "# io_model(noise, mu_optim)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 20.5 s, sys: 792 ms, total: 21.3 s\n",
      "Wall time: 21.8 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "n_iter = 10000\n",
    "losses = []\n",
    "m_vals = []\n",
    "for _iter in range(n_iter):    \n",
    "    noise = torch.Tensor(sample_noise(sample_size)).to(device)\n",
    "    x = y_sampler.x_dist.sample([mu_optim.shape[0]]).to(device)\n",
    "    # Do an update\n",
    "    data_gen = io_model(noise, torch.stack([mu_optim, x], dim=1))\n",
    "    loss = R(data_gen)\n",
    "    losses.append(loss.item())\n",
    "    io_model.zero_grad()\n",
    "    loss.backward()\n",
    "    with torch.no_grad():\n",
    "        mu_optim -= lr * mu_optim.grad.mean()\n",
    "        mu_optim.grad.zero_()\n",
    "    m_vals.append(mu_optim[0].item())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "def movingaverage(interval, window_size):\n",
    "    window = np.ones(int(window_size))/float(window_size)\n",
    "    return np.convolve(interval, window, 'same')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7f2f9bbcff28>]"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(12,6))\n",
    "plt.plot(losses);\n",
    "plt.grid()\n",
    "plt.ylabel(\"Loss\", fontsize=19)\n",
    "plt.xlabel(\"iter\", fontsize=19)\n",
    "plt.plot((movingaverage(losses, 50)), c='r')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 0, 'iter')"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(12,6))\n",
    "plt.plot(m_vals);\n",
    "plt.grid()\n",
    "plt.ylabel(\"$\\mu$\", fontsize=19)\n",
    "plt.xlabel(\"iter\", fontsize=19)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(582.8477172851562, tensor(-9.3814, device='cuda:0', grad_fn=<SelectBackward>))"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.min(losses), mu_optim[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x7f2fa1d1d0f0>"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(m_vals, losses)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.2"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}