Train seq2seq model in Keras without using one-hot encoded data

Train seq2seq model in Keras without using one-hot encoded output.ipynb

{
  "cells": [
    {
      "metadata": {
        "ExecuteTime": {
          "end_time": "2018-04-16T13:40:33.812957Z",
          "start_time": "2018-04-16T13:40:32.813032Z"
        },
        "trusted": true
      },
      "cell_type": "code",
      "source": "import numpy as np\nimport keras\nimport keras.backend as K\nfrom keras.layers import *\nfrom keras.models import *\nimport random\n\nmax_len = 30\nvocab_size = 3000\nX = np.random.randn(500, max_len)\nY = np.random.randn(500, max_len)\n\ninput_ = Input(X.shape[1:])\nhidden = Embedding(vocab_size, 2)(input_)\noutput = Dense(vocab_size, activation='softmax')(hidden)\nm = Model(input_, output)\n\ndef loss(y_true, y_pred):\n    y_true = K.one_hot(y_true, vocab_size)\n    return keras.losses.categorical_crossentropy(y_true, y_pred)\ny_tensor = K.placeholder((None, max_len), dtype='int32')\nm.compile(loss=loss, optimizer='adam', target_tensors=[y_tensor])\n\nm.fit(X, Y, batch_size=128, epochs=10, validation_split=0.1)",
      "execution_count": 4,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": "Train on 450 samples, validate on 50 samples\nEpoch 1/10\n450/450 [==============================] - 0s 412us/step - loss: 6.7245 - val_loss: 6.6977\nEpoch 2/10\n450/450 [==============================] - 0s 139us/step - loss: 6.7180 - val_loss: 6.6909\nEpoch 3/10\n450/450 [==============================] - 0s 139us/step - loss: 6.7111 - val_loss: 6.6839\nEpoch 4/10\n450/450 [==============================] - 0s 141us/step - loss: 6.7041 - val_loss: 6.6768\nEpoch 5/10\n450/450 [==============================] - 0s 139us/step - loss: 6.6969 - val_loss: 6.6696\nEpoch 6/10\n450/450 [==============================] - 0s 136us/step - loss: 6.6897 - val_loss: 6.6623\nEpoch 7/10\n450/450 [==============================] - 0s 136us/step - loss: 6.6823 - val_loss: 6.6549\nEpoch 8/10\n450/450 [==============================] - 0s 134us/step - loss: 6.6749 - val_loss: 6.6473\nEpoch 9/10\n450/450 [==============================] - 0s 133us/step - loss: 6.6673 - val_loss: 6.6396\nEpoch 10/10\n450/450 [==============================] - 0s 137us/step - loss: 6.6595 - val_loss: 6.6318\n"
        },
        {
          "data": {
            "text/plain": "<keras.callbacks.History at 0x7fd4a6a4fb70>"
          },
          "execution_count": 4,
          "metadata": {},
          "output_type": "execute_result"
        }
      ]
    }
  ],
  "metadata": {
    "hide_input": false,
    "kernelspec": {
      "name": "mermaid",
      "display_name": "mermaid",
      "language": "python"
    },
    "language_info": {
      "name": "python",
      "version": "3.6.3",
      "mimetype": "text/x-python",
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "pygments_lexer": "ipython3",
      "nbconvert_exporter": "python",
      "file_extension": ".py"
    },
    "toc": {
      "nav_menu": {},
      "number_sections": true,
      "sideBar": true,
      "skip_h1_title": false,
      "toc_cell": false,
      "toc_position": {},
      "toc_section_display": "block",
      "toc_window_display": false
    },
    "varInspector": {
      "window_display": false,
      "cols": {
        "lenName": 16,
        "lenType": 16,
        "lenVar": 40
      },
      "kernels_config": {
        "python": {
          "library": "var_list.py",
          "delete_cmd_prefix": "del ",
          "delete_cmd_postfix": "",
          "varRefreshCmd": "print(var_dic_list())"
        },
        "r": {
          "library": "var_list.r",
          "delete_cmd_prefix": "rm(",
          "delete_cmd_postfix": ") ",
          "varRefreshCmd": "cat(var_dic_list()) "
        }
      },
      "types_to_exclude": [
        "module",
        "function",
        "builtin_function_or_method",
        "instance",
        "_Feature"
      ]
    },
    "gist": {
      "id": "",
      "data": {
        "description": "Train seq2seq model in Keras without using one-hot encoded data",
        "public": true
      }
    }
  },
  "nbformat": 4,
  "nbformat_minor": 2
}