max/model-training
1
0
Fork 0
mirror of https://git.gfz-potsdam.de/naaice/model-training.git synced 2025-12-13 10:28:22 +01:00
Code Issues Packages Projects Releases Wiki Activity

add batch normalization

Browse Source
This commit is contained in:
Hannes Signer 2025-02-27 17:35:32 +01:00
parent d62115b111
commit e9baf132b4
2 changed files with 123 additions and 214 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

317
src/POET_Training.ipynb
View File

@ -59,8 +59,8 @@
"name": "stderr",
"output_type": "stream",
"text": [
"2025-02-26 17:33:56.161717: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.\n",
"2025-02-26 17:33:56.182881: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\n",
"2025-02-27 16:53:02.063625: I tensorflow/core/util/port.cc:153] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.\n",
"2025-02-27 16:53:02.082519: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\n",
"To enable the following instructions: SSE4.1 SSE4.2 AVX AVX2 AVX_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.\n"
]
}
@ -158,7 +158,7 @@
},
{
"cell_type": "code",
"execution_count": null,
"execution_count": 14,
"metadata": {},
"outputs": [
{
@ -180,17 +180,17 @@
}
],
"source": [
"preprocess = preprocessing() #np.log1p, np.expm1\n",
"preprocess = preprocessing(np.log1p, np.expm1) #np.log1p, np.expm1\n",
"X, y = preprocess.cluster(df_design[species_columns], df_results[species_columns])\n",
"\n",
"# optional: perform log transformation\n",
"# X, y = preprocess.funcTranform(X, y)\n",
"X, y = preprocess.funcTranform(X, y)\n",
"\n",
"X_train, X_test, y_train, y_test = preprocess.split(X, y, ratio=0.2)\n",
"X_train_origin = X_train.copy()\n",
"X_train, y_train = preprocess.balancer(X_train, y_train, strategy=\"off\")\n",
"X_train, y_train = preprocess.class_selection(X_train, y_train, class_label=0)\n",
"preprocess.scale_fit(X_train, y_train, scaling=\"global\", type=\"standard\")\n",
"preprocess.scale_fit(X_train, y_train, scaling=\"global\", type=\"minmax\")\n",
"X_train, X_test, y_train, y_test = preprocess.scale_transform(\n",
" X_train, X_test, y_train, y_test\n",
")\n",
@ -206,7 +206,7 @@
},
{
"cell_type": "code",
"execution_count": 18,
"execution_count": 5,
"metadata": {},
"outputs": [
{
@ -257,7 +257,7 @@
},
{
"cell_type": "code",
"execution_count": 19,
"execution_count": 17,
"metadata": {},
"outputs": [],
"source": [
@ -275,7 +275,7 @@
},
{
"cell_type": "code",
"execution_count": 20,
"execution_count": 18,
"metadata": {},
"outputs": [],
"source": [
@ -303,7 +303,7 @@
},
{
"cell_type": "code",
"execution_count": 21,
"execution_count": 19,
"metadata": {},
"outputs": [],
"source": [
@ -321,7 +321,7 @@
"h3 = 0.5099528144902471\n",
"\n",
"\n",
"scaler_type = \"standard\"\n",
"scaler_type = \"minmax\"\n",
"loss_variant = \"huber_mass_balance\"\n",
"delta = 1.7642791340966357\n",
"\n",
@ -339,219 +339,96 @@
},
{
"cell_type": "code",
"execution_count": 22,
"execution_count": null,
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Epoch 1/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 0.0159 - loss: 0.0082 - mass_balance: 0.0116 - val_huber: 2.1637e-04 - val_loss: 0.0023 - val_mass_balance: 0.0043\n",
"Epoch 2/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 13ms/step - huber: 2.7016e-04 - loss: 0.0014 - mass_balance: 0.0027 - val_huber: 1.7522e-04 - val_loss: 0.0013 - val_mass_balance: 0.0025\n",
"Epoch 3/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 2.0684e-04 - loss: 0.0012 - mass_balance: 0.0022 - val_huber: 3.5211e-04 - val_loss: 0.0022 - val_mass_balance: 0.0042\n",
"Epoch 4/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m13s\u001b[0m 15ms/step - huber: 2.1493e-04 - loss: 0.0011 - mass_balance: 0.0020 - val_huber: 2.2204e-04 - val_loss: 0.0010 - val_mass_balance: 0.0019\n",
"Epoch 5/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m14s\u001b[0m 15ms/step - huber: 1.3008e-04 - loss: 7.9274e-04 - mass_balance: 0.0015 - val_huber: 4.4501e-05 - val_loss: 3.6538e-04 - val_mass_balance: 6.8946e-04\n",
"Epoch 6/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 1.1257e-04 - loss: 6.6159e-04 - mass_balance: 0.0012 - val_huber: 1.0521e-04 - val_loss: 6.6072e-04 - val_mass_balance: 0.0012\n",
"Epoch 7/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 1.1593e-04 - loss: 6.9253e-04 - mass_balance: 0.0013 - val_huber: 4.6279e-05 - val_loss: 3.6958e-04 - val_mass_balance: 6.9983e-04\n",
"Epoch 8/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 13ms/step - huber: 9.9661e-05 - loss: 5.9260e-04 - mass_balance: 0.0011 - val_huber: 7.7692e-05 - val_loss: 3.9290e-04 - val_mass_balance: 7.3339e-04\n",
"Epoch 9/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 13ms/step - huber: 7.3264e-05 - loss: 4.7501e-04 - mass_balance: 8.9288e-04 - val_huber: 9.2113e-05 - val_loss: 8.4957e-04 - val_mass_balance: 0.0016\n",
"Epoch 10/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 8.6489e-05 - loss: 5.8290e-04 - mass_balance: 0.0011 - val_huber: 3.1637e-05 - val_loss: 5.8915e-04 - val_mass_balance: 0.0011\n",
"Epoch 11/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m14s\u001b[0m 16ms/step - huber: 5.9110e-05 - loss: 4.3897e-04 - mass_balance: 8.2805e-04 - val_huber: 3.6716e-05 - val_loss: 3.1306e-04 - val_mass_balance: 5.9171e-04\n",
"Epoch 12/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m24s\u001b[0m 27ms/step - huber: 5.8969e-05 - loss: 4.6788e-04 - mass_balance: 8.8358e-04 - val_huber: 7.4076e-05 - val_loss: 7.7939e-04 - val_mass_balance: 0.0015\n",
"Epoch 13/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 23ms/step - huber: 5.9244e-05 - loss: 3.9023e-04 - mass_balance: 7.3465e-04 - val_huber: 5.9634e-05 - val_loss: 3.9370e-04 - val_mass_balance: 7.3867e-04\n",
"Epoch 14/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m18s\u001b[0m 20ms/step - huber: 5.8721e-05 - loss: 3.8836e-04 - mass_balance: 7.3155e-04 - val_huber: 3.3829e-05 - val_loss: 2.5489e-04 - val_mass_balance: 4.8263e-04\n",
"Epoch 15/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m13s\u001b[0m 14ms/step - huber: 4.8019e-05 - loss: 3.4845e-04 - mass_balance: 6.5770e-04 - val_huber: 3.0502e-05 - val_loss: 2.2406e-04 - val_mass_balance: 4.2018e-04\n",
"Epoch 16/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m13s\u001b[0m 14ms/step - huber: 4.8281e-05 - loss: 3.4026e-04 - mass_balance: 6.4149e-04 - val_huber: 5.5124e-05 - val_loss: 4.3510e-04 - val_mass_balance: 8.2050e-04\n",
"Epoch 17/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 3.8080e-05 - loss: 3.1863e-04 - mass_balance: 6.0295e-04 - val_huber: 3.5506e-05 - val_loss: 6.1022e-04 - val_mass_balance: 0.0012\n",
"Epoch 18/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 3.6879e-05 - loss: 2.8393e-04 - mass_balance: 5.3626e-04 - val_huber: 3.7765e-05 - val_loss: 4.1528e-04 - val_mass_balance: 7.8552e-04\n",
"Epoch 19/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m14s\u001b[0m 15ms/step - huber: 3.2695e-05 - loss: 2.8065e-04 - mass_balance: 5.3144e-04 - val_huber: 1.8064e-05 - val_loss: 1.9093e-04 - val_mass_balance: 3.6273e-04\n",
"Epoch 20/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m22s\u001b[0m 25ms/step - huber: 3.1337e-05 - loss: 2.5379e-04 - mass_balance: 4.7992e-04 - val_huber: 2.5578e-05 - val_loss: 2.9832e-04 - val_mass_balance: 5.6782e-04\n",
"Epoch 21/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 23ms/step - huber: 3.7086e-05 - loss: 2.7006e-04 - mass_balance: 5.0938e-04 - val_huber: 2.2522e-05 - val_loss: 2.2766e-04 - val_mass_balance: 4.3007e-04\n",
"Epoch 22/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m19s\u001b[0m 21ms/step - huber: 2.8414e-05 - loss: 2.2388e-04 - mass_balance: 4.2329e-04 - val_huber: 1.7697e-05 - val_loss: 1.7123e-04 - val_mass_balance: 3.2254e-04\n",
"Epoch 23/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 16ms/step - huber: 3.0425e-05 - loss: 2.3615e-04 - mass_balance: 4.4640e-04 - val_huber: 1.9530e-05 - val_loss: 1.9458e-04 - val_mass_balance: 3.6947e-04\n",
"Epoch 24/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 2.5037e-05 - loss: 1.9369e-04 - mass_balance: 3.6621e-04 - val_huber: 1.9878e-05 - val_loss: 2.9097e-04 - val_mass_balance: 5.5204e-04\n",
"Epoch 25/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 13ms/step - huber: 2.6450e-05 - loss: 2.2114e-04 - mass_balance: 4.1811e-04 - val_huber: 3.3484e-05 - val_loss: 5.1009e-04 - val_mass_balance: 9.7122e-04\n",
"Epoch 26/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 2.8268e-05 - loss: 1.9281e-04 - mass_balance: 3.6375e-04 - val_huber: 1.4985e-05 - val_loss: 1.1141e-04 - val_mass_balance: 2.0968e-04\n",
"Epoch 27/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 14ms/step - huber: 2.8420e-05 - loss: 1.9706e-04 - mass_balance: 3.7174e-04 - val_huber: 2.0493e-05 - val_loss: 1.5578e-04 - val_mass_balance: 2.9476e-04\n",
"Epoch 28/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 22ms/step - huber: 2.1046e-05 - loss: 1.6831e-04 - mass_balance: 3.1860e-04 - val_huber: 2.5317e-05 - val_loss: 1.6482e-04 - val_mass_balance: 3.1080e-04\n",
"Epoch 29/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m19s\u001b[0m 22ms/step - huber: 2.2267e-05 - loss: 1.7674e-04 - mass_balance: 3.3441e-04 - val_huber: 9.3458e-06 - val_loss: 1.1187e-04 - val_mass_balance: 2.1313e-04\n",
"Epoch 30/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 22ms/step - huber: 1.9771e-05 - loss: 1.5978e-04 - mass_balance: 3.0248e-04 - val_huber: 1.1929e-05 - val_loss: 1.1887e-04 - val_mass_balance: 2.2595e-04\n",
"Epoch 31/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 17ms/step - huber: 2.2051e-05 - loss: 1.6235e-04 - mass_balance: 3.0656e-04 - val_huber: 1.7063e-05 - val_loss: 1.8251e-04 - val_mass_balance: 3.4744e-04\n",
"Epoch 32/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 14ms/step - huber: 1.8538e-05 - loss: 1.7150e-04 - mass_balance: 3.2539e-04 - val_huber: 1.3060e-05 - val_loss: 1.0948e-04 - val_mass_balance: 2.0754e-04\n",
"Epoch 33/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 13ms/step - huber: 2.3532e-05 - loss: 1.4467e-04 - mass_balance: 2.7262e-04 - val_huber: 7.6551e-06 - val_loss: 8.5257e-05 - val_mass_balance: 1.6184e-04\n",
"Epoch 34/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 1.7583e-05 - loss: 1.5367e-04 - mass_balance: 2.9137e-04 - val_huber: 8.2259e-06 - val_loss: 9.4663e-05 - val_mass_balance: 1.7979e-04\n",
"Epoch 35/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 17ms/step - huber: 1.8602e-05 - loss: 1.3293e-04 - mass_balance: 2.5101e-04 - val_huber: 8.5681e-06 - val_loss: 1.1647e-04 - val_mass_balance: 2.2283e-04\n",
"Epoch 36/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m22s\u001b[0m 25ms/step - huber: 1.5507e-05 - loss: 1.3234e-04 - mass_balance: 2.5097e-04 - val_huber: 1.2965e-05 - val_loss: 2.4758e-04 - val_mass_balance: 4.7178e-04\n",
"Epoch 37/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m21s\u001b[0m 24ms/step - huber: 2.7496e-05 - loss: 1.3054e-04 - mass_balance: 2.4429e-04 - val_huber: 1.4207e-05 - val_loss: 1.0057e-04 - val_mass_balance: 1.9007e-04\n",
"Epoch 38/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m19s\u001b[0m 22ms/step - huber: 1.7117e-05 - loss: 1.1822e-04 - mass_balance: 2.2331e-04 - val_huber: 8.3411e-06 - val_loss: 1.0676e-04 - val_mass_balance: 2.0202e-04\n",
"Epoch 39/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m14s\u001b[0m 16ms/step - huber: 1.9310e-05 - loss: 1.1455e-04 - mass_balance: 2.1555e-04 - val_huber: 9.0852e-06 - val_loss: 1.2996e-04 - val_mass_balance: 2.4633e-04\n",
"Epoch 40/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m14s\u001b[0m 15ms/step - huber: 2.3680e-05 - loss: 1.1200e-04 - mass_balance: 2.0969e-04 - val_huber: 7.7604e-06 - val_loss: 8.3377e-05 - val_mass_balance: 1.5779e-04\n",
"Epoch 41/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 1.3581e-05 - loss: 1.0753e-04 - mass_balance: 2.0348e-04 - val_huber: 6.4747e-06 - val_loss: 7.6034e-05 - val_mass_balance: 1.4468e-04\n",
"Epoch 42/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m14s\u001b[0m 16ms/step - huber: 1.4963e-05 - loss: 9.3135e-05 - mass_balance: 1.7541e-04 - val_huber: 5.8801e-06 - val_loss: 5.5339e-05 - val_mass_balance: 1.0464e-04\n",
"Epoch 43/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 23ms/step - huber: 1.4002e-05 - loss: 1.0034e-04 - mass_balance: 1.8957e-04 - val_huber: 6.0165e-06 - val_loss: 5.2093e-05 - val_mass_balance: 9.8668e-05\n",
"Epoch 44/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m19s\u001b[0m 21ms/step - huber: 1.3720e-05 - loss: 8.8099e-05 - mass_balance: 1.6604e-04 - val_huber: 6.4475e-06 - val_loss: 1.4406e-04 - val_mass_balance: 2.7444e-04\n",
"Epoch 45/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 22ms/step - huber: 2.0150e-05 - loss: 9.2090e-05 - mass_balance: 1.7219e-04 - val_huber: 4.7389e-06 - val_loss: 5.7199e-05 - val_mass_balance: 1.0862e-04\n",
"Epoch 46/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 17ms/step - huber: 1.1187e-05 - loss: 7.6198e-05 - mass_balance: 1.4389e-04 - val_huber: 5.7339e-06 - val_loss: 1.1307e-04 - val_mass_balance: 2.1630e-04\n",
"Epoch 47/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 1.1631e-05 - loss: 7.7719e-05 - mass_balance: 1.4674e-04 - val_huber: 5.1228e-06 - val_loss: 9.2185e-05 - val_mass_balance: 1.7662e-04\n",
"Epoch 48/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 8.6178e-06 - loss: 7.7370e-05 - mass_balance: 1.4671e-04 - val_huber: 4.4743e-06 - val_loss: 8.0649e-05 - val_mass_balance: 1.5478e-04\n",
"Epoch 49/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 1.2652e-05 - loss: 6.9353e-05 - mass_balance: 1.3021e-04 - val_huber: 4.3246e-06 - val_loss: 6.8228e-05 - val_mass_balance: 1.3036e-04\n",
"Epoch 50/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 18ms/step - huber: 1.8886e-05 - loss: 7.5731e-05 - mass_balance: 1.4089e-04 - val_huber: 4.6933e-06 - val_loss: 5.2718e-05 - val_mass_balance: 1.0036e-04\n",
"Epoch 51/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 22ms/step - huber: 8.6954e-06 - loss: 6.2883e-05 - mass_balance: 1.1879e-04 - val_huber: 5.7759e-06 - val_loss: 7.6331e-05 - val_mass_balance: 1.4537e-04\n",
"Epoch 52/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m19s\u001b[0m 21ms/step - huber: 9.6971e-06 - loss: 6.1626e-05 - mass_balance: 1.1608e-04 - val_huber: 3.5665e-06 - val_loss: 5.3019e-05 - val_mass_balance: 1.0099e-04\n",
"Epoch 53/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 22ms/step - huber: 8.4618e-06 - loss: 5.6582e-05 - mass_balance: 1.0672e-04 - val_huber: 4.0161e-06 - val_loss: 1.1739e-04 - val_mass_balance: 2.2385e-04\n",
"Epoch 54/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 17ms/step - huber: 8.6070e-06 - loss: 6.1078e-05 - mass_balance: 1.1538e-04 - val_huber: 3.9764e-06 - val_loss: 7.8183e-05 - val_mass_balance: 1.5002e-04\n",
"Epoch 55/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 13ms/step - huber: 8.3744e-06 - loss: 5.2668e-05 - mass_balance: 9.9188e-05 - val_huber: 4.8194e-06 - val_loss: 6.0945e-05 - val_mass_balance: 1.1645e-04\n",
"Epoch 56/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 13ms/step - huber: 8.0430e-06 - loss: 5.5005e-05 - mass_balance: 1.0381e-04 - val_huber: 4.2099e-06 - val_loss: 4.2308e-05 - val_mass_balance: 8.0290e-05\n",
"Epoch 57/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 9.8632e-06 - loss: 4.6309e-05 - mass_balance: 8.6417e-05 - val_huber: 3.7704e-06 - val_loss: 5.2539e-05 - val_mass_balance: 1.0076e-04\n",
"Epoch 58/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 17ms/step - huber: 1.1049e-05 - loss: 4.8602e-05 - mass_balance: 9.0530e-05 - val_huber: 4.5764e-06 - val_loss: 4.2829e-05 - val_mass_balance: 8.1037e-05\n",
"Epoch 59/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 23ms/step - huber: 7.8099e-06 - loss: 4.4020e-05 - mass_balance: 8.2638e-05 - val_huber: 3.7273e-06 - val_loss: 3.4894e-05 - val_mass_balance: 6.5669e-05\n",
"Epoch 60/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 23ms/step - huber: 7.3985e-06 - loss: 4.2755e-05 - mass_balance: 8.0336e-05 - val_huber: 3.5656e-06 - val_loss: 5.2735e-05 - val_mass_balance: 1.0131e-04\n",
"Epoch 61/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m19s\u001b[0m 22ms/step - huber: 5.7896e-06 - loss: 4.4404e-05 - mass_balance: 8.3940e-05 - val_huber: 3.3344e-06 - val_loss: 4.1619e-05 - val_mass_balance: 7.9748e-05\n",
"Epoch 62/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 18ms/step - huber: 7.6573e-06 - loss: 4.2019e-05 - mass_balance: 7.8827e-05 - val_huber: 2.9571e-06 - val_loss: 3.8318e-05 - val_mass_balance: 7.3062e-05\n",
"Epoch 63/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 13ms/step - huber: 7.0246e-06 - loss: 4.1996e-05 - mass_balance: 7.8925e-05 - val_huber: 3.2846e-06 - val_loss: 4.7290e-05 - val_mass_balance: 9.0100e-05\n",
"Epoch 64/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 7.1768e-06 - loss: 4.0670e-05 - mass_balance: 7.6273e-05 - val_huber: 3.2443e-06 - val_loss: 3.0751e-05 - val_mass_balance: 5.8160e-05\n",
"Epoch 65/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m13s\u001b[0m 14ms/step - huber: 8.2322e-06 - loss: 3.9525e-05 - mass_balance: 7.3858e-05 - val_huber: 3.6928e-06 - val_loss: 4.9247e-05 - val_mass_balance: 9.3333e-05\n",
"Epoch 66/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m18s\u001b[0m 21ms/step - huber: 7.3951e-06 - loss: 3.3935e-05 - mass_balance: 6.3241e-05 - val_huber: 2.6322e-06 - val_loss: 2.9614e-05 - val_mass_balance: 5.6361e-05\n",
"Epoch 67/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m19s\u001b[0m 22ms/step - huber: 7.9427e-06 - loss: 3.1778e-05 - mass_balance: 5.8965e-05 - val_huber: 3.0384e-06 - val_loss: 3.5186e-05 - val_mass_balance: 6.6310e-05\n",
"Epoch 68/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 22ms/step - huber: 9.2122e-06 - loss: 3.3397e-05 - mass_balance: 6.1652e-05 - val_huber: 2.6337e-06 - val_loss: 3.0151e-05 - val_mass_balance: 5.7459e-05\n",
"Epoch 69/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m17s\u001b[0m 20ms/step - huber: 7.5230e-06 - loss: 3.1762e-05 - mass_balance: 5.9018e-05 - val_huber: 3.2956e-06 - val_loss: 7.7759e-05 - val_mass_balance: 1.4897e-04\n",
"Epoch 70/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 14ms/step - huber: 5.1066e-06 - loss: 3.0846e-05 - mass_balance: 5.7952e-05 - val_huber: 2.6449e-06 - val_loss: 2.7740e-05 - val_mass_balance: 5.2617e-05\n",
"Epoch 71/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m11s\u001b[0m 13ms/step - huber: 5.9822e-06 - loss: 2.8337e-05 - mass_balance: 5.2830e-05 - val_huber: 3.2052e-06 - val_loss: 2.3249e-05 - val_mass_balance: 4.3688e-05\n",
"Epoch 72/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 6.0954e-06 - loss: 2.8623e-05 - mass_balance: 5.3352e-05 - val_huber: 2.6696e-06 - val_loss: 3.7623e-05 - val_mass_balance: 7.1514e-05\n",
"Epoch 73/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m18s\u001b[0m 20ms/step - huber: 6.5221e-06 - loss: 2.8740e-05 - mass_balance: 5.3428e-05 - val_huber: 2.3854e-06 - val_loss: 2.0875e-05 - val_mass_balance: 3.9654e-05\n",
"Epoch 74/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 23ms/step - huber: 5.3176e-06 - loss: 2.7184e-05 - mass_balance: 5.0835e-05 - val_huber: 2.4816e-06 - val_loss: 3.5760e-05 - val_mass_balance: 6.7804e-05\n",
"Epoch 75/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m19s\u001b[0m 22ms/step - huber: 5.0432e-06 - loss: 2.4995e-05 - mass_balance: 4.6679e-05 - val_huber: 2.5025e-06 - val_loss: 1.9573e-05 - val_mass_balance: 3.6922e-05\n",
"Epoch 76/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m19s\u001b[0m 22ms/step - huber: 4.8276e-06 - loss: 2.2201e-05 - mass_balance: 4.1353e-05 - val_huber: 2.4585e-06 - val_loss: 1.7460e-05 - val_mass_balance: 3.2908e-05\n",
"Epoch 77/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 4.4530e-06 - loss: 2.2517e-05 - mass_balance: 4.2087e-05 - val_huber: 2.6383e-06 - val_loss: 2.1215e-05 - val_mass_balance: 4.0054e-05\n",
"Epoch 78/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 4.7436e-06 - loss: 2.0644e-05 - mass_balance: 3.8389e-05 - val_huber: 2.7656e-06 - val_loss: 1.8618e-05 - val_mass_balance: 3.4989e-05\n",
"Epoch 79/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 14ms/step - huber: 8.6071e-06 - loss: 2.1559e-05 - mass_balance: 3.8975e-05 - val_huber: 2.4785e-06 - val_loss: 1.9744e-05 - val_mass_balance: 3.7424e-05\n",
"Epoch 80/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 18ms/step - huber: 4.8273e-06 - loss: 2.0366e-05 - mass_balance: 3.7813e-05 - val_huber: 2.3683e-06 - val_loss: 2.3192e-05 - val_mass_balance: 4.3858e-05\n",
"Epoch 81/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 22ms/step - huber: 4.1429e-06 - loss: 1.9682e-05 - mass_balance: 3.6682e-05 - val_huber: 2.4579e-06 - val_loss: 2.3470e-05 - val_mass_balance: 4.4667e-05\n",
"Epoch 82/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m19s\u001b[0m 22ms/step - huber: 4.3346e-06 - loss: 1.8738e-05 - mass_balance: 3.4823e-05 - val_huber: 2.2863e-06 - val_loss: 1.6274e-05 - val_mass_balance: 3.0652e-05\n",
"Epoch 83/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 22ms/step - huber: 4.6412e-06 - loss: 1.7737e-05 - mass_balance: 3.2783e-05 - val_huber: 2.2080e-06 - val_loss: 1.5864e-05 - val_mass_balance: 2.9908e-05\n",
"Epoch 84/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 17ms/step - huber: 6.0757e-06 - loss: 1.9005e-05 - mass_balance: 3.4790e-05 - val_huber: 2.2955e-06 - val_loss: 1.9878e-05 - val_mass_balance: 3.7724e-05\n",
"Epoch 85/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 5.0746e-06 - loss: 1.8464e-05 - mass_balance: 3.4061e-05 - val_huber: 2.2392e-06 - val_loss: 1.8274e-05 - val_mass_balance: 3.4456e-05\n",
"Epoch 86/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 6.1811e-06 - loss: 1.7876e-05 - mass_balance: 3.2582e-05 - val_huber: 2.2513e-06 - val_loss: 1.8889e-05 - val_mass_balance: 3.5718e-05\n",
"Epoch 87/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 3.5479e-06 - loss: 1.5930e-05 - mass_balance: 2.9635e-05 - val_huber: 2.2526e-06 - val_loss: 1.6175e-05 - val_mass_balance: 3.0466e-05\n",
"Epoch 88/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m16s\u001b[0m 18ms/step - huber: 7.3204e-06 - loss: 1.6691e-05 - mass_balance: 2.9917e-05 - val_huber: 2.1545e-06 - val_loss: 1.3727e-05 - val_mass_balance: 2.5782e-05\n",
"Epoch 89/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m21s\u001b[0m 23ms/step - huber: 3.5428e-06 - loss: 1.4906e-05 - mass_balance: 2.7661e-05 - val_huber: 2.1938e-06 - val_loss: 1.4968e-05 - val_mass_balance: 2.8242e-05\n",
"Epoch 90/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m19s\u001b[0m 22ms/step - huber: 4.8572e-06 - loss: 1.5157e-05 - mass_balance: 2.7730e-05 - val_huber: 2.1789e-06 - val_loss: 1.7663e-05 - val_mass_balance: 3.3284e-05\n",
"Epoch 91/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m19s\u001b[0m 22ms/step - huber: 5.7642e-06 - loss: 1.5536e-05 - mass_balance: 2.8178e-05 - val_huber: 2.1638e-06 - val_loss: 1.2848e-05 - val_mass_balance: 2.4079e-05\n",
"Epoch 92/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m18s\u001b[0m 21ms/step - huber: 5.3907e-06 - loss: 1.3979e-05 - mass_balance: 2.5303e-05 - val_huber: 2.1050e-06 - val_loss: 1.5735e-05 - val_mass_balance: 2.9686e-05\n",
"Epoch 93/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 5.6190e-06 - loss: 1.4178e-05 - mass_balance: 2.5596e-05 - val_huber: 2.1393e-06 - val_loss: 1.3981e-05 - val_mass_balance: 2.6364e-05\n",
"Epoch 94/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 6.4031e-06 - loss: 1.4058e-05 - mass_balance: 2.5123e-05 - val_huber: 2.1212e-06 - val_loss: 1.2008e-05 - val_mass_balance: 2.2475e-05\n",
"Epoch 95/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 4.3502e-06 - loss: 1.4156e-05 - mass_balance: 2.5949e-05 - val_huber: 2.0968e-06 - val_loss: 1.0900e-05 - val_mass_balance: 2.0334e-05\n",
"Epoch 96/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m13s\u001b[0m 15ms/step - huber: 3.4378e-06 - loss: 1.2550e-05 - mass_balance: 2.3135e-05 - val_huber: 2.1139e-06 - val_loss: 1.3909e-05 - val_mass_balance: 2.6132e-05\n",
"Epoch 97/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 22ms/step - huber: 4.0237e-06 - loss: 1.3306e-05 - mass_balance: 2.4419e-05 - val_huber: 2.1032e-06 - val_loss: 1.0320e-05 - val_mass_balance: 1.9246e-05\n",
"Epoch 98/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 22ms/step - huber: 4.7400e-06 - loss: 1.2574e-05 - mass_balance: 2.2770e-05 - val_huber: 2.1382e-06 - val_loss: 1.1723e-05 - val_mass_balance: 2.1962e-05\n",
"Epoch 99/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m21s\u001b[0m 24ms/step - huber: 5.2539e-06 - loss: 1.2562e-05 - mass_balance: 2.2590e-05 - val_huber: 2.1173e-06 - val_loss: 1.4093e-05 - val_mass_balance: 2.6535e-05\n",
"Epoch 100/100\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m21s\u001b[0m 24ms/step - huber: 4.6950e-06 - loss: 1.2635e-05 - mass_balance: 2.2899e-05 - val_huber: 2.0971e-06 - val_loss: 1.0790e-05 - val_mass_balance: 2.0159e-05\n",
"Training took 1548.7767686843872 seconds\n"
"Epoch 1/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m19s\u001b[0m 21ms/step - huber: 2.0302e-06 - loss: 9.1101e-05 - mass_balance: 1.7680e-04 - val_huber: 1.6816e-06 - val_loss: 8.5608e-05 - val_mass_balance: 1.6647e-04\n",
"Epoch 2/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 17ms/step - huber: 1.9164e-06 - loss: 9.1236e-05 - mass_balance: 1.7711e-04 - val_huber: 1.6744e-06 - val_loss: 8.8582e-05 - val_mass_balance: 1.7216e-04\n",
"Epoch 3/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 1.9028e-06 - loss: 8.8366e-05 - mass_balance: 1.7163e-04 - val_huber: 1.6736e-06 - val_loss: 8.9763e-05 - val_mass_balance: 1.7434e-04\n",
"Epoch 4/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m13s\u001b[0m 14ms/step - huber: 1.9853e-06 - loss: 9.0045e-05 - mass_balance: 1.7478e-04 - val_huber: 1.6635e-06 - val_loss: 9.0544e-05 - val_mass_balance: 1.7591e-04\n",
"Epoch 5/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m18s\u001b[0m 21ms/step - huber: 1.9494e-06 - loss: 8.7276e-05 - mass_balance: 1.6948e-04 - val_huber: 1.6741e-06 - val_loss: 8.6483e-05 - val_mass_balance: 1.6809e-04\n",
"Epoch 6/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 22ms/step - huber: 1.7979e-06 - loss: 8.7661e-05 - mass_balance: 1.7025e-04 - val_huber: 1.6537e-06 - val_loss: 8.6392e-05 - val_mass_balance: 1.6783e-04\n",
"Epoch 7/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 23ms/step - huber: 2.0609e-06 - loss: 8.4980e-05 - mass_balance: 1.6501e-04 - val_huber: 1.6581e-06 - val_loss: 8.4510e-05 - val_mass_balance: 1.6422e-04\n",
"Epoch 8/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m17s\u001b[0m 19ms/step - huber: 2.1306e-06 - loss: 8.5293e-05 - mass_balance: 1.6556e-04 - val_huber: 1.6407e-06 - val_loss: 8.3087e-05 - val_mass_balance: 1.6164e-04\n",
"Epoch 9/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 14ms/step - huber: 1.7904e-06 - loss: 8.4495e-05 - mass_balance: 1.6414e-04 - val_huber: 1.6346e-06 - val_loss: 8.2466e-05 - val_mass_balance: 1.6029e-04\n",
"Epoch 10/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 13ms/step - huber: 1.7857e-06 - loss: 8.3844e-05 - mass_balance: 1.6288e-04 - val_huber: 1.6314e-06 - val_loss: 8.4307e-05 - val_mass_balance: 1.6374e-04\n",
"Epoch 11/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m18s\u001b[0m 20ms/step - huber: 2.0917e-06 - loss: 8.3925e-05 - mass_balance: 1.6292e-04 - val_huber: 1.6253e-06 - val_loss: 8.5509e-05 - val_mass_balance: 1.6608e-04\n",
"Epoch 12/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 23ms/step - huber: 2.1661e-06 - loss: 8.2850e-05 - mass_balance: 1.6084e-04 - val_huber: 1.6364e-06 - val_loss: 8.2984e-05 - val_mass_balance: 1.6121e-04\n",
"Epoch 13/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m21s\u001b[0m 24ms/step - huber: 1.6631e-06 - loss: 8.3116e-05 - mass_balance: 1.6148e-04 - val_huber: 1.6294e-06 - val_loss: 8.0669e-05 - val_mass_balance: 1.5678e-04\n",
"Epoch 14/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m19s\u001b[0m 21ms/step - huber: 1.7664e-06 - loss: 8.1699e-05 - mass_balance: 1.5876e-04 - val_huber: 1.6272e-06 - val_loss: 8.0021e-05 - val_mass_balance: 1.5553e-04\n",
"Epoch 15/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 14ms/step - huber: 1.8568e-06 - loss: 8.0615e-05 - mass_balance: 1.5659e-04 - val_huber: 1.6175e-06 - val_loss: 8.0275e-05 - val_mass_balance: 1.5602e-04\n",
"Epoch 16/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 14ms/step - huber: 1.7830e-06 - loss: 8.0901e-05 - mass_balance: 1.5717e-04 - val_huber: 1.6209e-06 - val_loss: 7.8130e-05 - val_mass_balance: 1.5187e-04\n",
"Epoch 17/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 14ms/step - huber: 1.7357e-06 - loss: 7.9596e-05 - mass_balance: 1.5464e-04 - val_huber: 1.6223e-06 - val_loss: 8.7514e-05 - val_mass_balance: 1.7020e-04\n",
"Epoch 18/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m18s\u001b[0m 20ms/step - huber: 1.8609e-06 - loss: 7.9730e-05 - mass_balance: 1.5490e-04 - val_huber: 1.6178e-06 - val_loss: 7.8086e-05 - val_mass_balance: 1.5175e-04\n",
"Epoch 19/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m24s\u001b[0m 27ms/step - huber: 1.8063e-06 - loss: 7.8635e-05 - mass_balance: 1.5278e-04 - val_huber: 1.6157e-06 - val_loss: 7.7298e-05 - val_mass_balance: 1.5027e-04\n",
"Epoch 20/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m21s\u001b[0m 23ms/step - huber: 1.9740e-06 - loss: 7.8909e-05 - mass_balance: 1.5325e-04 - val_huber: 1.6102e-06 - val_loss: 7.8826e-05 - val_mass_balance: 1.5312e-04\n",
"Epoch 21/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 23ms/step - huber: 1.8796e-06 - loss: 7.7536e-05 - mass_balance: 1.5063e-04 - val_huber: 1.6175e-06 - val_loss: 7.9814e-05 - val_mass_balance: 1.5504e-04\n",
"Epoch 22/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 17ms/step - huber: 1.7153e-06 - loss: 7.7000e-05 - mass_balance: 1.4963e-04 - val_huber: 1.6125e-06 - val_loss: 7.5456e-05 - val_mass_balance: 1.4669e-04\n",
"Epoch 23/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m13s\u001b[0m 15ms/step - huber: 2.0066e-06 - loss: 7.7276e-05 - mass_balance: 1.5009e-04 - val_huber: 1.6136e-06 - val_loss: 7.6164e-05 - val_mass_balance: 1.4803e-04\n",
"Epoch 24/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m14s\u001b[0m 15ms/step - huber: 1.9103e-06 - loss: 7.6767e-05 - mass_balance: 1.4911e-04 - val_huber: 1.6051e-06 - val_loss: 7.6967e-05 - val_mass_balance: 1.4963e-04\n",
"Epoch 25/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m17s\u001b[0m 19ms/step - huber: 1.7270e-06 - loss: 7.5980e-05 - mass_balance: 1.4763e-04 - val_huber: 1.6134e-06 - val_loss: 7.4603e-05 - val_mass_balance: 1.4503e-04\n",
"Epoch 26/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m21s\u001b[0m 24ms/step - huber: 1.9137e-06 - loss: 7.5943e-05 - mass_balance: 1.4750e-04 - val_huber: 1.6108e-06 - val_loss: 7.6177e-05 - val_mass_balance: 1.4806e-04\n",
"Epoch 27/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m20s\u001b[0m 22ms/step - huber: 2.1447e-06 - loss: 7.5024e-05 - mass_balance: 1.4564e-04 - val_huber: 1.6123e-06 - val_loss: 7.5187e-05 - val_mass_balance: 1.4617e-04\n",
"Epoch 28/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m19s\u001b[0m 22ms/step - huber: 2.0587e-06 - loss: 7.4814e-05 - mass_balance: 1.4526e-04 - val_huber: 1.6096e-06 - val_loss: 7.4630e-05 - val_mass_balance: 1.4500e-04\n",
"Epoch 29/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 14ms/step - huber: 1.8129e-06 - loss: 7.3951e-05 - mass_balance: 1.4368e-04 - val_huber: 1.6081e-06 - val_loss: 7.4177e-05 - val_mass_balance: 1.4413e-04\n",
"Epoch 30/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 14ms/step - huber: 2.0191e-06 - loss: 7.4011e-05 - mass_balance: 1.4372e-04 - val_huber: 1.6102e-06 - val_loss: 7.4224e-05 - val_mass_balance: 1.4424e-04\n",
"Epoch 31/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m12s\u001b[0m 14ms/step - huber: 1.7857e-06 - loss: 7.3585e-05 - mass_balance: 1.4296e-04 - val_huber: 1.6118e-06 - val_loss: 7.3740e-05 - val_mass_balance: 1.4335e-04\n",
"Epoch 32/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 17ms/step - huber: 1.9859e-06 - loss: 7.2959e-05 - mass_balance: 1.4168e-04 - val_huber: 1.6117e-06 - val_loss: 7.2533e-05 - val_mass_balance: 1.4098e-04\n",
"Epoch 33/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m22s\u001b[0m 25ms/step - huber: 1.9710e-06 - loss: 7.2820e-05 - mass_balance: 1.4141e-04 - val_huber: 1.6038e-06 - val_loss: 7.4178e-05 - val_mass_balance: 1.4417e-04\n",
"Epoch 34/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m24s\u001b[0m 27ms/step - huber: 1.8785e-06 - loss: 7.2509e-05 - mass_balance: 1.4084e-04 - val_huber: 1.6048e-06 - val_loss: 7.2866e-05 - val_mass_balance: 1.4163e-04\n",
"Epoch 35/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m22s\u001b[0m 24ms/step - huber: 2.0480e-06 - loss: 7.2435e-05 - mass_balance: 1.4063e-04 - val_huber: 1.6072e-06 - val_loss: 7.1991e-05 - val_mass_balance: 1.3992e-04\n",
"Epoch 36/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m19s\u001b[0m 21ms/step - huber: 2.0031e-06 - loss: 7.1794e-05 - mass_balance: 1.3939e-04 - val_huber: 1.6078e-06 - val_loss: 7.1452e-05 - val_mass_balance: 1.3888e-04\n",
"Epoch 37/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m15s\u001b[0m 16ms/step - huber: 1.9493e-06 - loss: 7.1358e-05 - mass_balance: 1.3857e-04 - val_huber: 1.6105e-06 - val_loss: 7.1551e-05 - val_mass_balance: 1.3904e-04\n",
"Epoch 38/50\n",
"\u001b[1m886/886\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m13s\u001b[0m 14ms/step - huber: 1.9237e-06 - loss: 7.1155e-05 - mass_balance: 1.3818e-04 - val_huber: 1.6065e-06 - val_loss: 7.1723e-05 - val_mass_balance: 1.3939e-04\n",
"Epoch 39/50\n",
"\u001b[1m556/886\u001b[0m \u001b[32m━━━━━━━━━━━━\u001b[0m\u001b[37m━━━━━━━━\u001b[0m \u001b[1m5s\u001b[0m 16ms/step - huber: 1.9354e-06 - loss: 7.1366e-05 - mass_balance: 1.3857e-04"
]
}
],
"source": [
"history = model_training(model, epochs=100)"
"history = model_training(model, epochs=50)"
]
},
{
@ -926,6 +803,18 @@
"source": [
"test_model(model, X_test, y_test)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"### Model Explanation"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": []
}
],
"metadata": {

20
src/preprocessing.py
View File

@ -58,6 +58,26 @@ def model_definition(architecture):
keras.layers.Dense(8, dtype=dtype),
]
)
elif architecture == "large_batch_normalization":
model = keras.Sequential([
keras.layers.Input(shape=(8,), dtype=dtype),
BatchNormalization(), # Normalisierung der Eingabedaten
Dense(512, dtype=dtype),
BatchNormalization(), # Nach der Dense-Schicht für stabilere Verteilungen
LeakyReLU(negative_slope=0.01),
Dense(1024, dtype=dtype),
BatchNormalization(),
LeakyReLU(negative_slope=0.01),
Dense(512, dtype=dtype),
BatchNormalization(),
LeakyReLU(negative_slope=0.01),
Dense(8, dtype=dtype),
])
elif architecture == "paper":
model = keras.Sequential(