[FIX] Deprecate the losses.numpy module (#1471)

Author: marcopeix

docs/losses.numpy.html.md

@@ -128,8 +128,7 @@
           {
             "group": "Train/Evaluation",
             "pages": [
-              "losses.pytorch.html",
-              "losses.numpy.html"
+              "losses.pytorch.html"
             ]
           },
           {

docs/mintlify/docs.json

@@ -2,15 +2,14 @@
 os.environ["PYTORCH_ENABLE_MPS_FALLBACK"] = "1"
 
 import argparse
-import pandas as pd
+import numpy as np
 
 from ray import tune
 
 from neuralforecast.auto import AutoNHITS
 from neuralforecast.core import NeuralForecast
 
 from neuralforecast.losses.pytorch import MAE, HuberLoss
-from neuralforecast.losses.numpy import mae, mse
 #from datasetsforecast.long_horizon import LongHorizon, LongHorizonInfo
 from datasetsforecast.long_horizon2 import LongHorizon2, LongHorizon2Info
 
@@ -99,8 +98,8 @@
     print('y_true.shape (n_series, n_windows, n_time_out):\t', y_true.shape)
     print('y_hat.shape  (n_series, n_windows, n_time_out):\t', y_hat.shape)
 
-    print('MSE: ', mse(y_hat, y_true))
-    print('MAE: ', mae(y_hat, y_true))
+    print('MSE: ', np.mean((y_hat - y_true) ** 2))
+    print('MAE: ', np.mean(np.abs(y_hat - y_true)))
 
     # Save Outputs
     if not os.path.exists(f'./data/{dataset}'):

experiments/long_horizon/run_nhits.py

@@ -47,7 +47,7 @@
    "outputs": [],
    "source": [
     "%%capture\n",
-    "!pip install neuralforecast datasetsforecast"
+    "!pip install neuralforecast datasetsforecast utilsforecast"
    ]
   },
   {
@@ -433,35 +433,6 @@
     "nf.models[0].results.get_best_result().config"
    ]
   },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [
-    {
-     "name": "stdout",
-     "output_type": "stream",
-     "text": [
-      "Parsed results\n",
-      "2. y_true.shape (n_series, n_windows, n_time_out):\t (7, 11425, 96)\n",
-      "2. y_hat.shape  (n_series, n_windows, n_time_out):\t (7, 11425, 96)\n"
-     ]
-    }
-   ],
-   "source": [
-    "y_true = Y_hat_df.y.values\n",
-    "y_hat = Y_hat_df['AutoNHITS'].values\n",
-    "\n",
-    "n_series = len(Y_df.unique_id.unique())\n",
-    "\n",
-    "y_true = y_true.reshape(n_series, -1, horizon)\n",
-    "y_hat = y_hat.reshape(n_series, -1, horizon)\n",
-    "\n",
-    "print('Parsed results')\n",
-    "print('2. y_true.shape (n_series, n_windows, n_time_out):\\t', y_true.shape)\n",
-    "print('2. y_hat.shape  (n_series, n_windows, n_time_out):\\t', y_hat.shape)"
-   ]
-  },
   {
    "cell_type": "code",
    "execution_count": null,
@@ -479,23 +450,25 @@
     }
    ],
    "source": [
-    "fig, axs = plt.subplots(nrows=3, ncols=1, figsize=(10, 11))\n",
-    "fig.tight_layout()\n",
-    "\n",
+    "from utilsforecast.plotting import plot_series                                                                                      \n",
+    "                                                                                                                                  \n",
     "series = ['HUFL','HULL','LUFL','LULL','MUFL','MULL','OT']\n",
-    "series_idx = 3\n",
-    "\n",
-    "for idx, w_idx in enumerate([200, 300, 400]):\n",
-    "  axs[idx].plot(y_true[series_idx, w_idx,:],label='True')\n",
-    "  axs[idx].plot(y_hat[series_idx, w_idx,:],label='Forecast')\n",
-    "  axs[idx].grid()\n",
-    "  axs[idx].set_ylabel(series[series_idx]+f' window {w_idx}', \n",
-    "                      fontsize=17)\n",
-    "  if idx==2:\n",
-    "    axs[idx].set_xlabel('Forecast Horizon', fontsize=17)\n",
-    "plt.legend()\n",
-    "plt.show()\n",
-    "plt.close()"
+    "series_id = series[3]  # 'LULL'\n",
+    "\n",
+    "series_cutoffs = Y_hat_df.loc[Y_hat_df['unique_id'] == \"LULL\", 'cutoff'].unique()                                                \n",
+    "                                                        \n",
+    "for w_idx in [200, 300, 400]:\n",
+    "    cutoff = series_cutoffs[w_idx]\n",
+    "    plot_df = Y_hat_df.loc[Y_hat_df['cutoff'] == cutoff, ['unique_id', 'ds', 'y', 'AutoNHITS']]\n",
+    "\n",
+    "    fig = plot_series(\n",
+    "        df=plot_df[['unique_id', 'ds', 'y']],\n",
+    "        forecasts_df=plot_df[['unique_id', 'ds', 'AutoNHITS']],\n",
+    "        ids=[series_id],\n",
+    "        models=['AutoNHITS'],\n",
+    "        max_insample_length=96\n",
+    "    )\n",
+    "    display(fig)"
    ]
   },
   {
@@ -523,10 +496,17 @@
     }
    ],
    "source": [
-    "from neuralforecast.losses.numpy import mae, mse\n",
+    "from utilsforecast.evaluation import evaluate\n",
+    "from utilsforecast.losses import mae, mse\n",
+    "\n",
+    "eval_df = evaluate(\n",
+    "    df=Y_hat_df.drop(columns=[\"cutoff\"]),\n",
+    "    metrics=[mae, mse],\n",
+    "    agg_fn=\"mean\"\n",
+    ")\n",
     "\n",
-    "print('MAE: ', mae(y_hat, y_true))\n",
-    "print('MSE: ', mse(y_hat, y_true))"
+    "print('MAE: ', eval_df.iloc[0][\"AutoNHITS\"])\n",
+    "print('MSE: ', eval_df.iloc[1][\"AutoNHITS\"])"
    ]
   },
   {

nbs/docs/tutorials/longhorizon_nhits.ipynb

@@ -55,7 +55,7 @@
    "outputs": [],
    "source": [
     "%%capture\n",
-    "!pip install neuralforecast datasetsforecast"
+    "!pip install neuralforecast datasetsforecast utilsforecast"
    ]
   },
   {
@@ -557,7 +557,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "from neuralforecast.losses.numpy import mae"
+    "from utilsforecast.evaluation import evaluate\n",
+    "from utilsforecast.losses import mae"
    ]
   },
   {
@@ -576,13 +577,15 @@
     }
    ],
    "source": [
-    "mae_informer = mae(Y_hat_df['y'], Y_hat_df['Informer'])\n",
-    "mae_autoformer = mae(Y_hat_df['y'], Y_hat_df['Autoformer'])\n",
-    "mae_patchtst = mae(Y_hat_df['y'], Y_hat_df['PatchTST'])\n",
+    "eval_df = evaluate(\n",
+    "    df=Y_hat_df.drop(columns=[\"cutoff\"]),\n",
+    "    metrics=[mae],\n",
+    "    agg_fn=\"mean\"\n",
+    ")\n",
     "\n",
-    "print(f'Informer: {mae_informer:.3f}')\n",
-    "print(f'Autoformer: {mae_autoformer:.3f}')\n",
-    "print(f'PatchTST: {mae_patchtst:.3f}')"
+    "print('Informer: ', eval_df.iloc[0][\"Informer\"])\n",
+    "print('Autoformer: ', eval_df.iloc[0][\"Autoformer\"])\n",
+    "print('PatchTST: ', eval_df.iloc[0][\"PatchTST\"])"
    ]
   },
   {

nbs/docs/tutorials/longhorizon_transformers.ipynb

@@ -1,14 +1,13 @@
 import logging
 import warnings
 
+import numpy as np
 import optuna
 import pandas as pd
 import pytest
-import pytorch_lightning as pl
 from ray import tune
 
 from neuralforecast.common._base_auto import BaseAuto
-from neuralforecast.losses.numpy import mae
 from neuralforecast.losses.pytorch import MAE, MSE
 from neuralforecast.models.mlp import MLP
 from neuralforecast.tsdataset import TimeSeriesDataset
@@ -58,7 +57,7 @@ def test_ray_tune(setup_module):
     )
     auto.fit(dataset=dataset)
     y_hat = auto.predict(dataset=dataset)
-    assert mae(Y_test_df["y"].values, y_hat[:, 0]) < 200
+    assert np.mean(np.abs(Y_test_df["y"].values - y_hat[:, 0])) < 200
 
 
 def config_f(trial):
@@ -97,7 +96,7 @@ def test_optuna_tune(setup_module):
     auto2.fit(dataset=dataset)
     assert isinstance(auto2.results, optuna.Study)
     y_hat2 = auto2.predict(dataset=dataset)
-    assert mae(Y_test_df["y"].values, y_hat2[:, 0]) < 200
+    assert np.mean(np.abs(Y_test_df["y"].values - y_hat2[:, 0])) < 200
     Y_test_df["AutoMLP"] = y_hat2
 
     pd.concat([Y_train_df, Y_test_df]).drop("unique_id", axis=1).set_index("ds").plot()