train.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "train.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/Singular-Brain/DeepBioLCNet/blob/main/train.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# ! pip install git+https://github.com/BindsNET/bindsnet.git"
      ],
      "metadata": {
        "id": "a0QdT9WgYp2B"
      },
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# from google.colab import drive\n",
        "# drive.mount('/content/drive')"
      ],
      "metadata": {
        "id": "6yDwMm9oWEhF"
      },
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "execution_count": 4,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 607
        },
        "id": "mAsL72JjQzZk",
        "outputId": "fbc4bc29-da87-4e40-91aa-d0550d05d9d5"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Running on cuda\n",
            "Weights loaded ...\n",
            "Previous model not found! Training from the beginning...\n",
            "\n",
            "tensor([  0,   0,   0,   0,   0,   0,   0,   0, 256,   0])\n",
            "tensor([   0,    0,    0, 1428,    0,    0,    0,    0,    0,    0])\n",
            "tensor([   0,    0,    0,    0, 2195,    0,    0,    0,    0,    0])\n",
            "tensor([   0,    0,    0,    0, 1553,    0,    0,    0,    0,    0])\n",
            "tensor([   0,    0,    0,    0,    0,    0,    0,    0, 3228,    0])\n",
            "tensor([   0,    0,    0,    0,    0, 3159,    0,    0,    0,    0])\n",
            "Epoch: 1, Running accuracy: 16.67%\n",
            "tensor([   0,    0,    0,    0, 1771,    0,    0,    0,    0,    0])\n",
            "tensor([  0,   0,   0,   0,   0, 336,   0,   0,   0,   0])\n"
          ]
        },
        {
          "output_type": "error",
          "ename": "KeyboardInterrupt",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-4-c5374c02b6b0>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m    134\u001b[0m \u001b[0mnet\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mBioLCNet\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m**\u001b[0m\u001b[0mhparams\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mreward_fn\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mDynamicDopamineInjection\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgpu\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mgpu\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    135\u001b[0m \u001b[0mnet\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtesting\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mFalse\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 136\u001b[0;31m \u001b[0mnet\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfit\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdataloader\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mdataloader\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mval_loader\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mval_loader\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mreward_hparams\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mreward_hparams\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mtrain_hparams\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    137\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    138\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/content/biolcnet.py\u001b[0m in \u001b[0;36mfit\u001b[0;34m(self, dataloader, val_loader, reward_hparams, n_train, n_test, n_val, val_interval, label, hparams, online_validate, running_window_length, verbose)\u001b[0m\n\u001b[1;32m   1359\u001b[0m                         \u001b[0mtrue_label\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlabel\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1360\u001b[0m                         \u001b[0mdopaminergic_layers\u001b[0m\u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdopaminergic_layers\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1361\u001b[0;31m                         \u001b[0mclamp\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mclamp\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1362\u001b[0m                         )\n\u001b[1;32m   1363\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;32m/content/biolcnet.py\u001b[0m in \u001b[0;36mrun\u001b[0;34m(self, inputs, time, one_step, **kwargs)\u001b[0m\n\u001b[1;32m    925\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    926\u001b[0m                 \u001b[0;32mif\u001b[0m \u001b[0ml\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mcurrent_inputs\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 927\u001b[0;31m                     \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlayers\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0ml\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mcurrent_inputs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0ml\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[0m\u001b[1;32m    928\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[1;32m    929\u001b[0m                     \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlayers\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0ml\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mzeros\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlayers\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0ml\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0ms\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mshape\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/python3.7/dist-packages/bindsnet/network/nodes.py\u001b[0m in \u001b[0;36mforward\u001b[0;34m(self, x)\u001b[0m\n\u001b[1;32m    219\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0ms\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    220\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 221\u001b[0;31m         \u001b[0msuper\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mforward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\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    222\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    223\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0mreset_state_variables\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m->\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
          ]
        }
      ],
      "source": [
        "import torch\n",
        "import os\n",
        "import numpy as np\n",
        "\n",
        "from biolcnet import BioLCNet\n",
        "from reward import DynamicDopamineInjection\n",
        "from dataset import ClassSelector, load_datasets\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "gpu = True\n",
        "seed = 2045 # The singularity is near!\n",
        "\n",
        "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
        "\n",
        "\n",
        "if gpu and torch.cuda.is_available():\n",
        "    torch.cuda.manual_seed_all(seed)\n",
        "else:\n",
        "    torch.manual_seed(seed)\n",
        "    device = \"cpu\"\n",
        "    if gpu:\n",
        "        gpu = False\n",
        "\n",
        "torch.set_num_threads(os.cpu_count() - 1)\n",
        "print(\"Running on\", device)\n",
        "\n",
        "\n",
        "n_neurons = 1000\n",
        "n_classes = 10\n",
        "neuron_per_class = int(n_neurons/n_classes)\n",
        "\n",
        "\n",
        "train_hparams = {\n",
        "    'n_train' : 10,\n",
        "    'n_test' : 3,\n",
        "    'n_val' : 3,\n",
        "    'val_interval' : 5,\n",
        "    'running_window_length': 250,\n",
        "}\n",
        "\n",
        "network_hparams = {\n",
        "    # net structure\n",
        "    'epochs': 1,\n",
        "    'crop_size': 20,\n",
        "    'neuron_per_c': neuron_per_class,\n",
        "    'in_channels':1,\n",
        "    'n_channels1': 16,\n",
        "    'n_channels2': 0,\n",
        "    'filter_size1': 5,\n",
        "    'filter_size2': 0,\n",
        "    'stride1': 2,\n",
        "    'stride2': 0,\n",
        "    'deep': False,\n",
        "    'n_neurons' : n_neurons,\n",
        "    'n_classes': n_classes,\n",
        "    \n",
        "    # time & Phase\n",
        "    'dt' : 1,\n",
        "    'pre_observation': True,\n",
        "    'has_decision_period': True,\n",
        "    'observation_period': 256,\n",
        "    'decision_period': 256,\n",
        "    'time': 256*3,\n",
        "    'online_rewarding': False,\n",
        "\n",
        "    # Nodes\n",
        "    'theta_plus': 0.05,\n",
        "    'tc_theta_decay': 1e6,\n",
        "    'tc_trace':20,\n",
        "    'trace_additive' : False,\n",
        "    \n",
        "    # Learning\n",
        "    'nu_LC': (0.0001,0.01),\n",
        "    'nu_Output':0.1,\n",
        "\n",
        "    # weights\n",
        "    'wmin': 0.0,\n",
        "    'wmax': 1.0,\n",
        "    \n",
        "    # Inhibition\n",
        "    'inh_type': 'between_layers',\n",
        "    'inh_factor_FC': 100,\n",
        "    'inh_LC': True,\n",
        "    'inh_factor_LC': 100,\n",
        "    \n",
        "    # Normalization\n",
        "    'norm_factor_LC': 0.25*5*5,\n",
        "    \n",
        "    # clamping\n",
        "    'clamp_intensity': None,\n",
        "\n",
        "    # Save\n",
        "    'save_path': 'BioLCNet_Small_filter_16_trained.pth',  # Specify for saving the model (Especially for pre-training the lc layer)\n",
        "    'load_path': None,\n",
        "    'LC_weights_path': 'BioLCNet_Small_filter_16.pth', # Specify for loading the pre-trained lc weights\n",
        "\n",
        "    # Plot:\n",
        "    'confusion_matrix' : False,\n",
        "    'lc_weights_vis': False,\n",
        "    'out_weights_vis': False,\n",
        "    'lc_convergence_vis': False,\n",
        "    'out_convergence_vis': False,\n",
        "}\n",
        "\n",
        "reward_hparams= {\n",
        "    'n_labels': n_classes,\n",
        "    'neuron_per_class': neuron_per_class,\n",
        "    'variant': 'scalar',\n",
        "    'tc_reward':0,\n",
        "    'dopamine_base': 0.0,\n",
        "    'reward_base': 1.,\n",
        "    'punishment_base': 1.,\n",
        "    'sub_variant': 'RPE',\n",
        "    'td_nu': 0.0005,  #RPE\n",
        "    'ema_window': 10, #RPE\n",
        "    }\n",
        "\n",
        "# Dataset Hyperparameters\n",
        "mask = None\n",
        "mask_test = None\n",
        "n_classes = 10\n",
        "\n",
        "data_hparams = { \n",
        "    'intensity': 128,\n",
        "    'time': 256*3,\n",
        "    'crop_size': 20,\n",
        "    'round_input': False,\n",
        "}\n",
        "\n",
        "dataloader, val_loader, test_loader = load_datasets(data_hparams, target_classes=None, mask=mask, mask_test=mask_test)\n",
        "\n",
        "\n",
        "hparams = {**reward_hparams, **network_hparams, **train_hparams, **data_hparams}\n",
        "net = BioLCNet(**hparams, reward_fn = DynamicDopamineInjection, gpu = gpu)\n",
        "net.testing = False\n",
        "net.fit(dataloader = dataloader, val_loader = val_loader, reward_hparams = reward_hparams, **train_hparams)\n",
        "\n",
        "\n",
        "\n",
        "### Test\n",
        "data_hparams = { \n",
        "    'intensity': 128,\n",
        "    'time': 256*2,\n",
        "    'crop_size': 20,\n",
        "    'round_input': False,\n",
        "}\n",
        "network_hparams['time'] = 256*2\n",
        "net.time = 256*2\n",
        "_, _, test_loader = load_datasets(data_hparams, target_classes=None, mask=mask, mask_test=mask_test)\n",
        "net.testing = True\n",
        "test_acc = net.evaluate(test_loader, train_hparams['n_test'],train_hparams['n_test'],250)\n",
        "print(\"Test accuracy: \" + \"{:.2f}\".format(test_acc) + \"%\")"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        ""
      ],
      "metadata": {
        "id": "sqp9w_cfYoMk"
      },
      "execution_count": 3,
      "outputs": []
    }
  ]
}