DATA03_training_sample_gen1_3d_ViT_members.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "3a414f2d-8b54-4659-ab2b-b9ff5b5a4259",
   "metadata": {},
   "source": [
    "# Generate training samples for 3d ViT\n",
    "\n",
    "This notebook geneerate training samples (batch size = 1) for 3d ViT.\n",
    "* Training and validation samples were generated.\n",
    "* Samples were saved as .npy files; their file names contain the datetime information."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "df2ef1ec-9b90-4574-8ed9-806573e0ba76",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import sys\n",
    "import time\n",
    "import h5py\n",
    "import numpy as np\n",
    "from glob import glob\n",
    "from datetime import datetime, timedelta\n",
    "\n",
    "# ------------------------------------------------------- #\n",
    "# Turn-off warnings\n",
    "import logging\n",
    "import warnings\n",
    "\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3' \n",
    "logging.getLogger(\"tensorflow\").setLevel(logging.ERROR)\n",
    "\n",
    "# ------------------------------------------------------- #\n",
    "# Turn-off tensoflow-specific warnings\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import layers\n",
    "\n",
    "tf.autograph.set_verbosity(0)\n",
    "tf.get_logger().setLevel('ERROR')\n",
    "\n",
    "# ------------------------------------------------------- #\n",
    "# Import customized modules and settings\n",
    "sys.path.insert(0, '/glade/u/home/ksha/GAN_proj/')\n",
    "sys.path.insert(0, '/glade/u/home/ksha/GAN_proj/libs/')\n",
    "\n",
    "from namelist import *\n",
    "import data_utils as du\n",
    "import model_utils as mu"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "ae6cd660-668e-4659-9aae-914824eb7a76",
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "9558a6ba-79e8-4b07-86f7-a461742f7a81",
   "metadata": {},
   "outputs": [],
   "source": [
    "def norm_precip(x):\n",
    "    return np.log(0.1*x+1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "979438cf-134b-4fa6-a237-0216e83fd9aa",
   "metadata": {},
   "source": [
    "## Import VQ-VAE encoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "c08741a5-b988-4a78-86f1-f18960decbf8",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Hyperparameters\n",
    "filter_nums = [64, 128] # number of convolution kernels per down-/upsampling layer \n",
    "latent_dim = 4 # number of latent feature channels\n",
    "activation = 'gelu' # activation function\n",
    "num_embeddings = 128 #128 # number of the VQ codes\n",
    "\n",
    "input_size = (224, 464, 1) # size of MRMS input\n",
    "latent_size = (14, 29, latent_dim) # size of compressed latent features\n",
    "\n",
    "drop_encode = True\n",
    "drop_decode = True\n",
    "\n",
    "# location for saving new weights\n",
    "model_name_encoder_load = model_dir+'models/VQ_VAE_encoder_stack1_tune0'\n",
    "model_name_decoder_load = model_dir+'models/VQ_VAE_decoder_stack1_tune0'\n",
    "\n",
    "model_encoder = mu.VQ_VAE_encoder(input_size, filter_nums, latent_dim, num_embeddings, activation, drop_encode)\n",
    "\n",
    "W_old = mu.dummy_loader(model_name_encoder_load)\n",
    "model_encoder.set_weights(W_old)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9293af4e-95b2-4337-bce6-511c005b072e",
   "metadata": {},
   "source": [
    "## Generate training samples\n",
    "\n",
    "* Locations: `$camp_dir+BATCH_ViT_members/`\n",
    "* VQ-VAE encoder is applied"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "d312e4d7-cc5e-42de-8e53-ee0a15c6196f",
   "metadata": {},
   "outputs": [],
   "source": [
    "BATCH_dir = camp_dir+'BATCH_ViT_members/'\n",
    "batch_name = 'ViT_{}_{}.npy' #.format(datetime)\n",
    "\n",
    "LEADs = np.arange(6, 168+6, 6) # forecast lead times\n",
    "N_leads = len(LEADs)\n",
    "\n",
    "grid_shape = (N_leads, 14, 29, 4)\n",
    "ccpa_shape = (224, 464)\n",
    "\n",
    "with h5py.File(save_dir+'CCPA_domain.hdf', 'r') as h5io:\n",
    "    land_mask_CCPA = h5io['land_mask_CCPA'][...]\n",
    "land_mask_CCPA = np.logical_not(land_mask_CCPA)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "a65a4a5a-4025-4b54-ae90-779a73affcf9",
   "metadata": {},
   "outputs": [],
   "source": [
    "year = 2002\n",
    "\n",
    "# ------------------------------------------------------- #\n",
    "# Datetime information\n",
    "\n",
    "base = datetime(year, 1, 1)\n",
    "\n",
    "if year % 4 == 0:\n",
    "    N_days = 366\n",
    "else:\n",
    "    N_days = 365\n",
    "\n",
    "date_list = [base + timedelta(days=d) for d in range(N_days)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "e300fcf0-d599-46b0-b815-829b85a2c250",
   "metadata": {},
   "outputs": [],
   "source": [
    "filename_gefs = camp_dir+'GFS_reforecast/GEFS_{}_y{}.hdf'\n",
    "filename_gefs_AVE = camp_dir+'GFS_reforecast/GEFS_AVE_y{}.hdf'\n",
    "filename_gefs_STD = camp_dir+'GFS_reforecast/GEFS_STD_y{}.hdf'\n",
    "\n",
    "filename_ccpa = camp_dir+'CCPA/CCPA_y{}.hdf'\n",
    "\n",
    "with h5py.File(filename_ccpa.format(year), 'r') as h5io:\n",
    "    CCPA_base = h5io['CCPA'][...]\n",
    "L_base = len(CCPA_base)\n",
    "\n",
    "# forecast lead times can exceed one year\n",
    "N_beyond = 10\n",
    "N_total = L_base + N_beyond\n",
    "with h5py.File(filename_ccpa.format(year+1), 'r') as h5io:\n",
    "    CCPA_extra = h5io['CCPA'][:N_beyond, ...]\n",
    "\n",
    "CCPA = np.concatenate((CCPA_base, CCPA_extra), axis=0)\n",
    "CCPA = norm_precip(CCPA)\n",
    "CCPA[:, :, land_mask_CCPA] = 0.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "2d644aff-2c3b-4287-a5f0-072d5288b8f9",
   "metadata": {},
   "outputs": [],
   "source": [
    "with h5py.File(filename_gefs_AVE.format(year), 'r') as h5io:\n",
    "    GEFS_AVE = h5io['GEFS_APCP'][...] \n",
    "\n",
    "with h5py.File(filename_gefs_STD.format(year), 'r') as h5io:\n",
    "    GEFS_STD = h5io['GEFS_std'][...]\n",
    "\n",
    "GEFS_AVE = norm_precip(GEFS_AVE)\n",
    "GEFS_STD = norm_precip(GEFS_STD)\n",
    "\n",
    "GEFS_AVE[:, :, land_mask_CCPA] = 0.0\n",
    "GEFS_STD[:, :, land_mask_CCPA] = 0.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "e1540247-d85d-44d0-ac39-2e4eadf9e609",
   "metadata": {},
   "outputs": [],
   "source": [
    "GEFS_embed = np.empty(grid_shape[:-1]+(12,))\n",
    "CCPA_embed = np.empty(grid_shape)\n",
    "\n",
    "ens_names = ['c00', 'p01', 'p02', 'p03', 'p04']\n",
    "\n",
    "for ens in ens_names:\n",
    "    with h5py.File(filename_gefs.format(ens, year), 'r') as h5io:\n",
    "        try:\n",
    "            GEFS = h5io['GEFS_APCP'][...]\n",
    "        except:\n",
    "            GEFS = h5io['APCP'][...]\n",
    "        \n",
    "    GEFS = norm_precip(GEFS)\n",
    "    GEFS[:, :, land_mask_CCPA] = 0.0\n",
    "\n",
    "    for d, dt in enumerate(date_list):\n",
    "        dt_str = datetime.strftime(dt, '%Y%m%d')\n",
    "        GEFS_embed[...] = np.nan\n",
    "        CCPA_embed[...] = np.nan\n",
    "        \n",
    "        for ilead, lead in enumerate(LEADs):\n",
    "            d_ = lead // 24; day = d + d_\n",
    "            ind_hour = lead % 24; ind_hour = int(ind_hour/6)\n",
    "            \n",
    "            ccpa_input = CCPA[day, ind_hour, ...][None, ..., None]\n",
    "            gefs_input = GEFS[d, ilead, ...][None, ..., None]\n",
    "            gefs_ave_input = GEFS_AVE[d, ilead, ...][None, ..., None]\n",
    "            gefs_std_input = GEFS_STD[d, ilead, ...][None, ..., None]\n",
    "    \n",
    "            if np.sum(np.isnan(ccpa_input)) + np.sum(np.isnan(gefs_input)) == 0:\n",
    "                CCPA_embed[ilead, ...] = model_encoder.predict(ccpa_input, verbose=0)[0, ...]\n",
    "                GEFS_embed[ilead, ..., 0:4] = model_encoder.predict(gefs_input, verbose=0)[0, ...]\n",
    "                GEFS_embed[ilead, ..., 4:8] = model_encoder.predict(gefs_ave_input, verbose=0)[0, ...]\n",
    "                GEFS_embed[ilead, ..., 8:12] = model_encoder.predict(gefs_std_input, verbose=0)[0, ...]\n",
    "    \n",
    "            else:\n",
    "                CCPA_embed[ilead, ...] = np.nan\n",
    "                GEFS_embed[ilead, ...] = np.nan\n",
    "    \n",
    "        if np.sum(np.isnan(CCPA_embed)) + np.sum(np.isnan(GEFS_embed)) == 0:        \n",
    "            data_save = {}\n",
    "            data_save['CCPA_embed'] = CCPA_embed\n",
    "            data_save['GEFS_embed'] = GEFS_embed\n",
    "            save_name_ = BATCH_dir+batch_name.format(dt_str, ens)\n",
    "            print(save_name_)\n",
    "            np.save(save_name_, data_save)\n",
    "            #raise"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "371faca8-6743-4f6c-8712-892478fe9be4",
   "metadata": {},
   "source": [
    "### Check batch quality"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "f265ce51-81a2-45d2-bb35-ffdbe724b34a",
   "metadata": {},
   "outputs": [],
   "source": [
    "data_save = np.load(camp_dir+'BATCH_ViT/ViT_20020105.npy', allow_pickle=True)[()]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "71d802d6-16c4-4927-87a6-69b79f541348",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.QuadMesh at 0x14ff2987ed40>"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.pcolormesh(data_save['GEFS_embed'][-5, ..., 8], vmin=-0.1, vmax=0.1, cmap=plt.cm.RdBu_r)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "53f34363-3560-40ff-96c0-8dd0f210ccda",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.QuadMesh at 0x1520a5f1f610>"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.pcolormesh(data_save['GEFS_embed'][-5, ..., 0], vmin=-0.1, vmax=0.1, cmap=plt.cm.RdBu_r)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "bf49046f-ad8e-4abd-ab64-fdf25209b2c3",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.10.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}