keras_LSTM_rhone_data.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "cebb903b",
   "metadata": {},
   "outputs": [
    {
     "ename": "OSError",
     "evalue": "Unable to open file (unable to open file: name = 'Rhone_data_continuous.h5', errno = 2, error message = 'No such file or directory', flags = 0, o_flags = 0)",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mOSError\u001b[0m                                   Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-1-166204757541>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      9\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     10\u001b[0m \u001b[0mfilename\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m\"Rhone_data_continuous.h5\"\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 11\u001b[0;31m \u001b[0mf\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mh5py\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mFile\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfilename\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'r'\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[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"Keys: %s\"\u001b[0m \u001b[0;34m%\u001b[0m \u001b[0mf\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mkeys\u001b[0m\u001b[0;34m(\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~/.local/lib/python3.9/site-packages/h5py/_hl/files.py\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, name, mode, driver, libver, userblock_size, swmr, rdcc_nslots, rdcc_nbytes, rdcc_w0, track_order, fs_strategy, fs_persist, fs_threshold, **kwds)\u001b[0m\n\u001b[1;32m    422\u001b[0m             \u001b[0;32mwith\u001b[0m \u001b[0mphil\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    423\u001b[0m                 \u001b[0mfapl\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmake_fapl\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdriver\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlibver\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrdcc_nslots\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrdcc_nbytes\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mrdcc_w0\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mkwds\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 424\u001b[0;31m                 fid = make_fid(name, mode, userblock_size,\n\u001b[0m\u001b[1;32m    425\u001b[0m                                fapl, fcpl=make_fcpl(track_order=track_order, fs_strategy=fs_strategy,\n\u001b[1;32m    426\u001b[0m                                fs_persist=fs_persist, fs_threshold=fs_threshold),\n",
      "\u001b[0;32m~/.local/lib/python3.9/site-packages/h5py/_hl/files.py\u001b[0m in \u001b[0;36mmake_fid\u001b[0;34m(name, mode, userblock_size, fapl, fcpl, swmr)\u001b[0m\n\u001b[1;32m    188\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mswmr\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0mswmr_support\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    189\u001b[0m             \u001b[0mflags\u001b[0m \u001b[0;34m|=\u001b[0m \u001b[0mh5f\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mACC_SWMR_READ\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 190\u001b[0;31m         \u001b[0mfid\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mh5f\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mopen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mname\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mflags\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfapl\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mfapl\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    191\u001b[0m     \u001b[0;32melif\u001b[0m \u001b[0mmode\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m'r+'\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    192\u001b[0m         \u001b[0mfid\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mh5f\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mopen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mname\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mh5f\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mACC_RDWR\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfapl\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mfapl\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32mh5py/_objects.pyx\u001b[0m in \u001b[0;36mh5py._objects.with_phil.wrapper\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mh5py/_objects.pyx\u001b[0m in \u001b[0;36mh5py._objects.with_phil.wrapper\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mh5py/h5f.pyx\u001b[0m in \u001b[0;36mh5py.h5f.open\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;31mOSError\u001b[0m: Unable to open file (unable to open file: name = 'Rhone_data_continuous.h5', errno = 2, error message = 'No such file or directory', flags = 0, o_flags = 0)"
     ]
    }
   ],
   "source": [
    "import h5py\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import layers\n",
    "import seaborn as sns\n",
    "\n",
    "filename = \"/data/fast0/datasets/Rhone_data_continuous.h5\"\n",
    "f = h5py.File(filename, 'r')\n",
    "print(\"Keys: %s\" % f.keys())\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "id": "15f0b9e6",
   "metadata": {},
   "outputs": [],
   "source": [
    "# LSTM / training parameters\n",
    "\n",
    "split_fraction = 0.715\n",
    "step = 2 # Sample rate (i.e., keep one in every six values)\n",
    "look_back = 720 # How far to look back in time\n",
    "learning_rate = 0.001\n",
    "batch_size = 256\n",
    "epochs = 10\n",
    "sequence_length = int(look_back / step)\n",
    "train_split = int(split_fraction * int(len(f['Discharge'])))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "id": "2d9924f0",
   "metadata": {},
   "outputs": [],
   "source": [
    "# dataset_all = keras.preprocessing.timeseries_dataset_from_array(\n",
    "#     f['DAS Data'],\n",
    "#     f['Discharge'],\n",
    "#     sequence_length=sequence_length,\n",
    "#     sampling_rate=step,\n",
    "#     batch_size=batch_size\n",
    "# )\n",
    "\n",
    "x = np.transpose([chan[750] for chan in f['DAS Data']])\n",
    "y = f['Discharge']\n",
    "\n",
    "# Normalize by hand because LSTM normalization layers are complicated!\n",
    "\n",
    "x_mean = np.mean( x[:train_split], axis=0 )\n",
    "x_std  = np.std(  x[:train_split], axis=0 )\n",
    "\n",
    "y_mean = np.mean( y[:train_split] )\n",
    "y_std  = np.std(  y[:train_split] )\n",
    "\n",
    "x_train = (x[0:train_split-1,] - x_mean)/x_std\n",
    "#x_train = (x[0:train_split-1,:] - x_mean)/x_std\n",
    "y_train = (y[0:train_split-1] - y_mean)/y_std\n",
    "\n",
    "x_val = (x[train_split:] - x_mean)/x_std\n",
    "y_val = (y[train_split:] - y_mean)/y_std\n",
    "\n",
    "x_train = x_train.reshape(len(x_train), 1)\n",
    "x_val  = x_val.reshape(len(x_val), 1)\n",
    "y_train = y_train.reshape(len(y_train), 1)\n",
    "y_val = y_val.reshape(len(y_val), 1)\n",
    "\n",
    "dataset_train = keras.preprocessing.timeseries_dataset_from_array(\n",
    "    x_train,y_train,\n",
    "    sequence_length=sequence_length,\n",
    "    sampling_rate=step,\n",
    "    batch_size=batch_size\n",
    ")\n",
    "\n",
    "dataset_val = keras.preprocessing.timeseries_dataset_from_array(\n",
    "    x_val,y_val,\n",
    "    sequence_length=sequence_length,\n",
    "    sampling_rate=step,\n",
    "    batch_size=batch_size\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "id": "ce66604e",
   "metadata": {},
   "outputs": [],
   "source": [
    "# def get_dataset_partitions_tf(ds, ds_size, train_split=0.8, val_split=0.1, test_split=0.1, shuffle=True, shuffle_size=10000):\n",
    "#     assert (train_split + test_split + val_split) == 1\n",
    "# #https://towardsdatascience.com/how-to-split-a-tensorflow-dataset-into-train-validation-and-test-sets-526c8dd29438    \n",
    "#     if shuffle:\n",
    "#         # Specify seed to always have the same split distribution between runs\n",
    "#         ds = ds.shuffle(shuffle_size, seed=12)\n",
    "    \n",
    "#     train_size = int(train_split * ds_size)\n",
    "#     val_size = int(val_split * ds_size)\n",
    "    \n",
    "#     train_ds = ds.take(train_size)    \n",
    "#     val_ds = ds.skip(train_size).take(val_size)\n",
    "#     test_ds = ds.skip(train_size).skip(val_size)\n",
    "    \n",
    "#     return train_ds, val_ds, test_ds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "id": "20ef66cc",
   "metadata": {},
   "outputs": [],
   "source": [
    "# dataset_train, dataset_val, dataset_test = get_dataset_partitions_tf(dataset_all, len(f))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 137,
   "id": "fc8657a1",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Input shape: (256, 360, 1)\n",
      "Target shape: (256, 1)\n"
     ]
    }
   ],
   "source": [
    "for batch in dataset_train.take(1):\n",
    "    inputs, targets = batch\n",
    "\n",
    "print(\"Input shape:\", inputs.numpy().shape)\n",
    "print(\"Target shape:\", targets.numpy().shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "id": "2bf2391b",
   "metadata": {},
   "outputs": [],
   "source": [
    "model = keras.Sequential([\n",
    "    layers.LSTM(32),\n",
    "    layers.Dense(1)\n",
    "])\n",
    "\n",
    "\n",
    "model.compile(optimizer=keras.optimizers.Adam(learning_rate=learning_rate), loss=\"mse\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 139,
   "id": "caea1e3c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "173/173 [==============================] - 37s 209ms/step - loss: 0.9526 - val_loss: 4.0687\n",
      "\n",
      "Epoch 00001: loss improved from inf to 0.95262, saving model to model_checkpoint.h5\n",
      "Epoch 2/10\n",
      "173/173 [==============================] - 36s 207ms/step - loss: 0.9515 - val_loss: 4.2017\n",
      "\n",
      "Epoch 00002: loss improved from 0.95262 to 0.95154, saving model to model_checkpoint.h5\n",
      "Epoch 3/10\n",
      "173/173 [==============================] - 35s 201ms/step - loss: 0.9376 - val_loss: 4.2118\n",
      "\n",
      "Epoch 00003: loss improved from 0.95154 to 0.93760, saving model to model_checkpoint.h5\n",
      "Epoch 4/10\n",
      "173/173 [==============================] - 33s 191ms/step - loss: 0.9317 - val_loss: 4.2315\n",
      "\n",
      "Epoch 00004: loss improved from 0.93760 to 0.93173, saving model to model_checkpoint.h5\n",
      "Epoch 5/10\n",
      "173/173 [==============================] - 33s 188ms/step - loss: 0.9276 - val_loss: 4.2339\n",
      "\n",
      "Epoch 00005: loss improved from 0.93173 to 0.92760, saving model to model_checkpoint.h5\n",
      "Epoch 6/10\n",
      "173/173 [==============================] - 32s 186ms/step - loss: 0.9252 - val_loss: 4.2393\n",
      "\n",
      "Epoch 00006: loss improved from 0.92760 to 0.92524, saving model to model_checkpoint.h5\n",
      "Epoch 7/10\n",
      "173/173 [==============================] - 33s 190ms/step - loss: 0.9236 - val_loss: 4.2525\n",
      "\n",
      "Epoch 00007: loss improved from 0.92524 to 0.92357, saving model to model_checkpoint.h5\n",
      "Epoch 8/10\n",
      "173/173 [==============================] - 34s 198ms/step - loss: 0.9216 - val_loss: 4.2565\n",
      "\n",
      "Epoch 00008: loss improved from 0.92357 to 0.92161, saving model to model_checkpoint.h5\n",
      "Epoch 9/10\n",
      "173/173 [==============================] - 33s 192ms/step - loss: 0.9205 - val_loss: 4.2630\n",
      "\n",
      "Epoch 00009: loss improved from 0.92161 to 0.92051, saving model to model_checkpoint.h5\n",
      "Epoch 10/10\n",
      "173/173 [==============================] - 36s 207ms/step - loss: 0.9192 - val_loss: 4.2584\n",
      "\n",
      "Epoch 00010: loss improved from 0.92051 to 0.91916, saving model to model_checkpoint.h5\n"
     ]
    }
   ],
   "source": [
    "path_checkpoint = \"model_checkpoint.h5\"\n",
    "es_callback = keras.callbacks.EarlyStopping(monitor=\"loss\", min_delta=0, patience=5)\n",
    "\n",
    "modelckpt_callback = keras.callbacks.ModelCheckpoint(\n",
    "    monitor=\"loss\",\n",
    "    filepath=path_checkpoint,\n",
    "    verbose=1,\n",
    "    save_weights_only=True,\n",
    "    save_best_only=True,\n",
    ")\n",
    "\n",
    "history = model.fit(\n",
    "    dataset_train,\n",
    "    epochs=epochs,\n",
    "    validation_data=dataset_val,\n",
    "    callbacks=[es_callback, modelckpt_callback],\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "id": "071b0242",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0, 0.5, 'Model Discharge')"
      ]
     },
     "execution_count": 146,
     "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.subplots()\n",
    "for xx, yy in dataset_val.take(len(dataset_val)):\n",
    "    plt.plot(yy*y_std+y_mean,model.predict(xx)*y_std+y_mean,'ok')\n",
    "plt.plot((5,20),(5,20))\n",
    "plt.xlabel('True Discharge')\n",
    "plt.ylabel('Model Discharge')"
   ]
  }
 ],
 "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": "3.9.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}