From 9205ebb0e86a5fd7bceb7a8d5c6cd6fcd481d3ae Mon Sep 17 00:00:00 2001
From: Michal Moskal <michal@moskal.me>
Date: Fri, 16 Oct 2020 15:25:58 +0200
Subject: [PATCH] Update for new ml4f

---
 examples/gesture/train_model.ipynb | 597 +++++++++++++++--------------
 1 file changed, 304 insertions(+), 293 deletions(-)

diff --git a/examples/gesture/train_model.ipynb b/examples/gesture/train_model.ipynb
index bc21916..3271576 100644
--- a/examples/gesture/train_model.ipynb
+++ b/examples/gesture/train_model.ipynb
@@ -11,7 +11,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 15,
+   "execution_count": 12,
    "metadata": {},
    "outputs": [
     {
@@ -19,41 +19,24 @@
      "output_type": "stream",
      "text": [
       "\u001b[K\u001b[?25h\u001b[37;40mnpm\u001b[0m \u001b[0m\u001b[30;43mWARN\u001b[0m \u001b[0m\u001b[35mdeprecated\u001b[0m core-js@1.2.7: core-js@<3 is no longer maintained and not recommended for usage due to the number of issues. Please, upgrade your dependencies to the actual version of core-js@3.\n",
-      "\u001b[K\u001b[?25h/usr/local/bin/ml4f -> /usr/local/lib/node_modules/ml4f/ml4fipt: \u001b[32;40mtiming\u001b[0m \u001b[35maction:finalize\u001b[0m\u001b[0m\u001b[K[K\n",
+      "\u001b[K\u001b[?25h/usr/local/bin/ml4f -> /usr/local/lib/node_modules/ml4f/ml4f\u001b[35mfinalize\u001b[0m /usr/local/lib/node_modules\u001b[0m\u001b[KK\u001b[KK\n",
       "/usr/local/bin/ts-node -> /usr/local/lib/node_modules/ts-node/dist/bin.js\n",
       "/usr/local/bin/ts-script -> /usr/local/lib/node_modules/ts-node/dist/bin-script-deprecated.js\n",
       "/usr/local/bin/ts-node-transpile-only -> /usr/local/lib/node_modules/ts-node/dist/bin-transpile.js\n",
       "/usr/local/bin/ts-node-script -> /usr/local/lib/node_modules/ts-node/dist/bin-script.js\n",
       "/usr/local/bin/tsc -> /usr/local/lib/node_modules/typescript/bin/tsc\n",
       "/usr/local/bin/tsserver -> /usr/local/lib/node_modules/typescript/bin/tsserver\n",
+      "+ ml4f@1.4.0\n",
       "+ ts-node@9.0.0\n",
       "+ typescript@4.0.3\n",
-      "+ ml4f@1.3.0\n",
-      "updated 3 packages in 8.02s\n",
-      "\u001b[K\u001b[?25h\u001b[37;40mnpm\u001b[0m \u001b[0m\u001b[30;43mWARN\u001b[0m \u001b[0m\u001b[35mdeprecated\u001b[0m popper.js@1.16.1: You can find the new Popper v2 at @popperjs/core, this package is dedicated to the legacy v1\n",
-      "\u001b[K\u001b[?25h\u001b[37;40mnpm\u001b[0m \u001b[0m\u001b[30;43mWARN\u001b[0m \u001b[0m\u001b[35mdeprecated\u001b[0m request@2.88.2: request has been deprecated, see https://github.com/request/request/issues/3142\n",
-      "\u001b[K\u001b[?25h\u001b[37;40mnpm\u001b[0m \u001b[0m\u001b[30;43mWARN\u001b[0m \u001b[0m\u001b[35mdeprecated\u001b[0m har-validator@5.1.5: this library is no longer supportedt\u001b[0m\u001b[K\n",
-      "\u001b[K\u001b[?25h\u001b[37;40mnpm\u001b[0m \u001b[0m\u001b[30;43mWARN\u001b[0m \u001b[0m\u001b[35mrm\u001b[0m not removing /Users/michal/src/jupyter-jacdac/node_modules/.bin/semver as it wasn't installed by /Users/michal/src/jupyter-jacdac/node_modules/semver\n",
-      "\u001b[K\u001b[?25h\u001b[37;40mnpm\u001b[0m \u001b[0m\u001b[30;43mWARN\u001b[0m \u001b[0m\u001b[35mrm\u001b[0m not removing /Users/michal/src/jupyter-jacdac/node_modules/.bin/marked as it wasn't installed by /Users/michal/src/jupyter-jacdac/node_modules/marked\n",
-      "\u001b[K\u001b[?25h\u001b[37;40mnpm\u001b[0m \u001b[0m\u001b[34;40mnotice\u001b[0m\u001b[35m\u001b[0m created a lockfile as package-lock.json. You should commit this file.[0m\u001b[K\n",
-      "\u001b[0m\u001b[37;40mnpm\u001b[0m \u001b[0m\u001b[30;43mWARN\u001b[0m\u001b[35m\u001b[0m @lumino/coreutils@1.5.3 requires a peer of crypto@1.0.1 but none is installed. You must install peer dependencies yourself.\n",
-      "\u001b[0m\n",
-      "+ @types/node@14.11.8\n",
-      "added 147 packages from 15 contributors, removed 363 packages, updated 434 packages and audited 1022 packages in 64.183s\n",
-      "\n",
-      "43 packages are looking for funding\n",
-      "  run `npm fund` for details\n",
-      "\n",
-      "found \u001b[92m0\u001b[0m vulnerabilities\n",
-      "\n",
-      "\u001b[K\u001b[?25h7m            \u001b[27m\u001b[90m░░░░░░\u001b[0m⸩ ⠇ postinstall: \u001b[7msill\u001b[0m \u001b[35minstall\u001b[0m printInstalled\u001b[0m\u001b[K"
+      "updated 3 packages in 6.501s\n"
      ]
     }
    ],
    "source": [
     "! npm install -g -u typescript ts-node ml4f\n",
     "#! pip3 install tensorflow==2.3 tensorboard=2.3\n",
-    "! npm install @types/node"
+    "#! npm install @types/node"
    ]
   },
   {
@@ -65,7 +48,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 1,
+   "execution_count": 13,
    "metadata": {},
    "outputs": [
     {
@@ -125,7 +108,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 2,
+   "execution_count": 29,
    "metadata": {
     "colab": {},
     "colab_type": "code",
@@ -139,7 +122,7 @@
     "\n",
     "for fn in glob.glob(\"./built/seg/*.csv\"):\n",
     "    data = pd.read_csv(fn)\n",
-    "    #data.plot(figsize=(20,7),title=fn)\n"
+    "    data.plot(figsize=(20,7),title=fn)\n"
    ]
   },
   {
@@ -305,15 +288,24 @@
     ]
    },
    "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "Reusing TensorBoard on port 6007 (pid 17933), started 5:06:41 ago. (Use '!kill 17933' to kill it.)"
+      ]
+     },
+     "metadata": {},
+     "output_type": "display_data"
+    },
     {
      "data": {
       "text/html": [
        "\n",
-       "      <iframe id=\"tensorboard-frame-5d8f007ba8a6843a\" width=\"100%\" height=\"800\" frameborder=\"0\">\n",
+       "      <iframe id=\"tensorboard-frame-3a2ade9a1ca5b3c9\" width=\"100%\" height=\"800\" frameborder=\"0\">\n",
        "      </iframe>\n",
        "      <script>\n",
        "        (function() {\n",
-       "          const frame = document.getElementById(\"tensorboard-frame-5d8f007ba8a6843a\");\n",
+       "          const frame = document.getElementById(\"tensorboard-frame-3a2ade9a1ca5b3c9\");\n",
        "          const url = new URL(\"/\", window.location);\n",
        "          const port = 6007;\n",
        "          if (port) {\n",
@@ -336,409 +328,409 @@
      "output_type": "stream",
      "text": [
       "Epoch 1/200\n",
-      " 1/32 [..............................] - ETA: 0s - loss: 1.4304 - accuracy: 0.1875WARNING:tensorflow:From /usr/local/lib/python3.8/site-packages/tensorflow/python/ops/summary_ops_v2.py:1277: stop (from tensorflow.python.eager.profiler) is deprecated and will be removed after 2020-07-01.\n",
+      " 1/32 [..............................] - ETA: 0s - loss: 1.4062 - accuracy: 0.2188WARNING:tensorflow:From /usr/local/lib/python3.8/site-packages/tensorflow/python/ops/summary_ops_v2.py:1277: stop (from tensorflow.python.eager.profiler) is deprecated and will be removed after 2020-07-01.\n",
       "Instructions for updating:\n",
       "use `tf.profiler.experimental.stop` instead.\n",
-      "WARNING:tensorflow:Callbacks method `on_train_batch_end` is slow compared to the batch time (batch time: 0.0030s vs `on_train_batch_end` time: 0.0093s). Check your callbacks.\n",
-      "32/32 [==============================] - 0s 7ms/step - loss: 1.2922 - accuracy: 0.4280 - val_loss: 1.1693 - val_accuracy: 0.5420\n",
+      "WARNING:tensorflow:Callbacks method `on_train_batch_end` is slow compared to the batch time (batch time: 0.0026s vs `on_train_batch_end` time: 0.0106s). Check your callbacks.\n",
+      "32/32 [==============================] - 0s 8ms/step - loss: 1.2629 - accuracy: 0.4510 - val_loss: 1.1565 - val_accuracy: 0.5140\n",
       "Epoch 2/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 1.1647 - accuracy: 0.4920 - val_loss: 1.0640 - val_accuracy: 0.5720\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 1.1306 - accuracy: 0.5060 - val_loss: 1.0688 - val_accuracy: 0.5220\n",
       "Epoch 3/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 1.0447 - accuracy: 0.5440 - val_loss: 0.9028 - val_accuracy: 0.6300\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 1.0130 - accuracy: 0.5500 - val_loss: 0.9456 - val_accuracy: 0.5840\n",
       "Epoch 4/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.8862 - accuracy: 0.5910 - val_loss: 0.7608 - val_accuracy: 0.6540\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.8923 - accuracy: 0.5860 - val_loss: 0.8307 - val_accuracy: 0.6260\n",
       "Epoch 5/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.7533 - accuracy: 0.6410 - val_loss: 0.6451 - val_accuracy: 0.7160\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.7552 - accuracy: 0.6520 - val_loss: 0.7044 - val_accuracy: 0.6700\n",
       "Epoch 6/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.6596 - accuracy: 0.6960 - val_loss: 0.5695 - val_accuracy: 0.8060\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.6329 - accuracy: 0.7270 - val_loss: 0.6320 - val_accuracy: 0.7580\n",
       "Epoch 7/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.5860 - accuracy: 0.7660 - val_loss: 0.4858 - val_accuracy: 0.8360\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.5743 - accuracy: 0.7600 - val_loss: 0.5732 - val_accuracy: 0.7860\n",
       "Epoch 8/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.5476 - accuracy: 0.7610 - val_loss: 0.4503 - val_accuracy: 0.8460\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.5129 - accuracy: 0.7950 - val_loss: 0.5097 - val_accuracy: 0.8200\n",
       "Epoch 9/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.4848 - accuracy: 0.8080 - val_loss: 0.4280 - val_accuracy: 0.8620\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.4707 - accuracy: 0.8190 - val_loss: 0.4959 - val_accuracy: 0.8320\n",
       "Epoch 10/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.4646 - accuracy: 0.8260 - val_loss: 0.4141 - val_accuracy: 0.8600\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.4567 - accuracy: 0.8130 - val_loss: 0.4653 - val_accuracy: 0.8100\n",
       "Epoch 11/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.4534 - accuracy: 0.8230 - val_loss: 0.3634 - val_accuracy: 0.8840\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.4212 - accuracy: 0.8250 - val_loss: 0.4406 - val_accuracy: 0.8480\n",
       "Epoch 12/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.4172 - accuracy: 0.8400 - val_loss: 0.3961 - val_accuracy: 0.8620\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.4054 - accuracy: 0.8410 - val_loss: 0.4217 - val_accuracy: 0.8360\n",
       "Epoch 13/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.4066 - accuracy: 0.8400 - val_loss: 0.3393 - val_accuracy: 0.9020\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.4017 - accuracy: 0.8340 - val_loss: 0.4201 - val_accuracy: 0.8400\n",
       "Epoch 14/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.3810 - accuracy: 0.8490 - val_loss: 0.3274 - val_accuracy: 0.9060\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.3794 - accuracy: 0.8610 - val_loss: 0.3872 - val_accuracy: 0.8660\n",
       "Epoch 15/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.3702 - accuracy: 0.8520 - val_loss: 0.3178 - val_accuracy: 0.8980\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.3534 - accuracy: 0.8650 - val_loss: 0.3962 - val_accuracy: 0.8660\n",
       "Epoch 16/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.3478 - accuracy: 0.8680 - val_loss: 0.2962 - val_accuracy: 0.9140\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.3605 - accuracy: 0.8660 - val_loss: 0.3795 - val_accuracy: 0.8480\n",
       "Epoch 17/200\n",
-      "32/32 [==============================] - 0s 3ms/step - loss: 0.3263 - accuracy: 0.8730 - val_loss: 0.3047 - val_accuracy: 0.9000\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.3538 - accuracy: 0.8700 - val_loss: 0.3488 - val_accuracy: 0.8720\n",
       "Epoch 18/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.3406 - accuracy: 0.8730 - val_loss: 0.2821 - val_accuracy: 0.9220\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.3126 - accuracy: 0.8820 - val_loss: 0.3673 - val_accuracy: 0.8660\n",
       "Epoch 19/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.3038 - accuracy: 0.8870 - val_loss: 0.2794 - val_accuracy: 0.9120\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.3070 - accuracy: 0.8780 - val_loss: 0.3556 - val_accuracy: 0.8780\n",
       "Epoch 20/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.3181 - accuracy: 0.8800 - val_loss: 0.2758 - val_accuracy: 0.9080\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.2999 - accuracy: 0.8740 - val_loss: 0.3505 - val_accuracy: 0.8620\n",
       "Epoch 21/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.2935 - accuracy: 0.8820 - val_loss: 0.2669 - val_accuracy: 0.9080\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.2891 - accuracy: 0.8950 - val_loss: 0.3124 - val_accuracy: 0.8940\n",
       "Epoch 22/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.2799 - accuracy: 0.8980 - val_loss: 0.2586 - val_accuracy: 0.9180\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.2653 - accuracy: 0.8950 - val_loss: 0.3017 - val_accuracy: 0.8840\n",
       "Epoch 23/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.2745 - accuracy: 0.9080 - val_loss: 0.2733 - val_accuracy: 0.9180\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.2690 - accuracy: 0.8900 - val_loss: 0.3013 - val_accuracy: 0.8820\n",
       "Epoch 24/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.2844 - accuracy: 0.8990 - val_loss: 0.2567 - val_accuracy: 0.9200\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.2627 - accuracy: 0.8940 - val_loss: 0.3195 - val_accuracy: 0.8680\n",
       "Epoch 25/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.2681 - accuracy: 0.8990 - val_loss: 0.2260 - val_accuracy: 0.9340\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.2699 - accuracy: 0.8910 - val_loss: 0.2769 - val_accuracy: 0.8980\n",
       "Epoch 26/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.2601 - accuracy: 0.8980 - val_loss: 0.2271 - val_accuracy: 0.9260\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.2537 - accuracy: 0.9060 - val_loss: 0.3049 - val_accuracy: 0.8880\n",
       "Epoch 27/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.2474 - accuracy: 0.9100 - val_loss: 0.2411 - val_accuracy: 0.9220\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.2608 - accuracy: 0.8980 - val_loss: 0.2879 - val_accuracy: 0.8960\n",
       "Epoch 28/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.2504 - accuracy: 0.9160 - val_loss: 0.2468 - val_accuracy: 0.9180\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.2285 - accuracy: 0.9110 - val_loss: 0.2841 - val_accuracy: 0.8920\n",
       "Epoch 29/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.2239 - accuracy: 0.9120 - val_loss: 0.2094 - val_accuracy: 0.9300\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.2339 - accuracy: 0.9110 - val_loss: 0.2789 - val_accuracy: 0.8780\n",
       "Epoch 30/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.2411 - accuracy: 0.9060 - val_loss: 0.2309 - val_accuracy: 0.9340\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.2241 - accuracy: 0.9070 - val_loss: 0.2416 - val_accuracy: 0.9180\n",
       "Epoch 31/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.2309 - accuracy: 0.9220 - val_loss: 0.2255 - val_accuracy: 0.9240\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.2103 - accuracy: 0.9150 - val_loss: 0.2496 - val_accuracy: 0.9240\n",
       "Epoch 32/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.2245 - accuracy: 0.9170 - val_loss: 0.2067 - val_accuracy: 0.9380\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.2257 - accuracy: 0.9100 - val_loss: 0.2352 - val_accuracy: 0.9140\n",
       "Epoch 33/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.2028 - accuracy: 0.9220 - val_loss: 0.2254 - val_accuracy: 0.9280\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.2159 - accuracy: 0.9150 - val_loss: 0.2366 - val_accuracy: 0.9220\n",
       "Epoch 34/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.2135 - accuracy: 0.9180 - val_loss: 0.2381 - val_accuracy: 0.9260\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.2154 - accuracy: 0.9170 - val_loss: 0.2364 - val_accuracy: 0.9120\n",
       "Epoch 35/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.2007 - accuracy: 0.9320 - val_loss: 0.2180 - val_accuracy: 0.9280\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1977 - accuracy: 0.9110 - val_loss: 0.2324 - val_accuracy: 0.9140\n",
       "Epoch 36/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1870 - accuracy: 0.9360 - val_loss: 0.1932 - val_accuracy: 0.9360\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1909 - accuracy: 0.9250 - val_loss: 0.2456 - val_accuracy: 0.9100\n",
       "Epoch 37/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1895 - accuracy: 0.9410 - val_loss: 0.1972 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1947 - accuracy: 0.9320 - val_loss: 0.2222 - val_accuracy: 0.9300\n",
       "Epoch 38/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1980 - accuracy: 0.9320 - val_loss: 0.1980 - val_accuracy: 0.9300\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.1886 - accuracy: 0.9300 - val_loss: 0.2261 - val_accuracy: 0.9200\n",
       "Epoch 39/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1992 - accuracy: 0.9330 - val_loss: 0.2070 - val_accuracy: 0.9320\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1822 - accuracy: 0.9320 - val_loss: 0.2272 - val_accuracy: 0.9280\n",
       "Epoch 40/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1946 - accuracy: 0.9350 - val_loss: 0.2072 - val_accuracy: 0.9360\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.1815 - accuracy: 0.9320 - val_loss: 0.2266 - val_accuracy: 0.9180\n",
       "Epoch 41/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1826 - accuracy: 0.9400 - val_loss: 0.2159 - val_accuracy: 0.9260\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.1736 - accuracy: 0.9240 - val_loss: 0.2316 - val_accuracy: 0.9160\n",
       "Epoch 42/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1752 - accuracy: 0.9340 - val_loss: 0.2014 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.1612 - accuracy: 0.9400 - val_loss: 0.2244 - val_accuracy: 0.9180\n",
       "Epoch 43/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1716 - accuracy: 0.9310 - val_loss: 0.2040 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.1766 - accuracy: 0.9270 - val_loss: 0.2074 - val_accuracy: 0.9260\n",
       "Epoch 44/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1746 - accuracy: 0.9390 - val_loss: 0.2067 - val_accuracy: 0.9360\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.1619 - accuracy: 0.9480 - val_loss: 0.2161 - val_accuracy: 0.9280\n",
       "Epoch 45/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1613 - accuracy: 0.9450 - val_loss: 0.2083 - val_accuracy: 0.9340\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.1579 - accuracy: 0.9350 - val_loss: 0.2232 - val_accuracy: 0.9220\n",
       "Epoch 46/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1637 - accuracy: 0.9410 - val_loss: 0.2146 - val_accuracy: 0.9360\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.1431 - accuracy: 0.9500 - val_loss: 0.2059 - val_accuracy: 0.9300\n",
       "Epoch 47/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1636 - accuracy: 0.9400 - val_loss: 0.2225 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1583 - accuracy: 0.9330 - val_loss: 0.2258 - val_accuracy: 0.9360\n",
       "Epoch 48/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1649 - accuracy: 0.9350 - val_loss: 0.2193 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.1856 - accuracy: 0.9270 - val_loss: 0.1949 - val_accuracy: 0.9400\n",
       "Epoch 49/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1640 - accuracy: 0.9400 - val_loss: 0.2194 - val_accuracy: 0.9260\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.1453 - accuracy: 0.9440 - val_loss: 0.1979 - val_accuracy: 0.9320\n",
       "Epoch 50/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1726 - accuracy: 0.9320 - val_loss: 0.2122 - val_accuracy: 0.9300\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1390 - accuracy: 0.9530 - val_loss: 0.1810 - val_accuracy: 0.9460\n",
       "Epoch 51/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1627 - accuracy: 0.9400 - val_loss: 0.1977 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1665 - accuracy: 0.9290 - val_loss: 0.1907 - val_accuracy: 0.9400\n",
       "Epoch 52/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1499 - accuracy: 0.9500 - val_loss: 0.1895 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1531 - accuracy: 0.9400 - val_loss: 0.1788 - val_accuracy: 0.9480\n",
       "Epoch 53/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1704 - accuracy: 0.9310 - val_loss: 0.2386 - val_accuracy: 0.9340\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1400 - accuracy: 0.9380 - val_loss: 0.1862 - val_accuracy: 0.9480\n",
       "Epoch 54/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1461 - accuracy: 0.9430 - val_loss: 0.2156 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1426 - accuracy: 0.9460 - val_loss: 0.2388 - val_accuracy: 0.9280\n",
       "Epoch 55/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1663 - accuracy: 0.9400 - val_loss: 0.2192 - val_accuracy: 0.9380\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1463 - accuracy: 0.9340 - val_loss: 0.1909 - val_accuracy: 0.9420\n",
       "Epoch 56/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1807 - accuracy: 0.9340 - val_loss: 0.2299 - val_accuracy: 0.9340\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1388 - accuracy: 0.9410 - val_loss: 0.1955 - val_accuracy: 0.9420\n",
       "Epoch 57/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1396 - accuracy: 0.9510 - val_loss: 0.2054 - val_accuracy: 0.9360\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1329 - accuracy: 0.9510 - val_loss: 0.1908 - val_accuracy: 0.9420\n",
       "Epoch 58/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1480 - accuracy: 0.9410 - val_loss: 0.2263 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1364 - accuracy: 0.9470 - val_loss: 0.2163 - val_accuracy: 0.9380\n",
       "Epoch 59/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1560 - accuracy: 0.9320 - val_loss: 0.1808 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1402 - accuracy: 0.9420 - val_loss: 0.1961 - val_accuracy: 0.9300\n",
       "Epoch 60/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1440 - accuracy: 0.9450 - val_loss: 0.1795 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1331 - accuracy: 0.9510 - val_loss: 0.1812 - val_accuracy: 0.9380\n",
       "Epoch 61/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1386 - accuracy: 0.9520 - val_loss: 0.1890 - val_accuracy: 0.9360\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1247 - accuracy: 0.9490 - val_loss: 0.1737 - val_accuracy: 0.9420\n",
       "Epoch 62/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1434 - accuracy: 0.9490 - val_loss: 0.2147 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1537 - accuracy: 0.9420 - val_loss: 0.1696 - val_accuracy: 0.9480\n",
       "Epoch 63/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1457 - accuracy: 0.9520 - val_loss: 0.2030 - val_accuracy: 0.9360\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1100 - accuracy: 0.9570 - val_loss: 0.1627 - val_accuracy: 0.9520\n",
       "Epoch 64/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1422 - accuracy: 0.9420 - val_loss: 0.2216 - val_accuracy: 0.9380\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1296 - accuracy: 0.9570 - val_loss: 0.1716 - val_accuracy: 0.9380\n",
       "Epoch 65/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1270 - accuracy: 0.9530 - val_loss: 0.2065 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1232 - accuracy: 0.9510 - val_loss: 0.1945 - val_accuracy: 0.9320\n",
       "Epoch 66/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1306 - accuracy: 0.9470 - val_loss: 0.2082 - val_accuracy: 0.9380\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1263 - accuracy: 0.9550 - val_loss: 0.1634 - val_accuracy: 0.9420\n",
       "Epoch 67/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1193 - accuracy: 0.9570 - val_loss: 0.2554 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1193 - accuracy: 0.9620 - val_loss: 0.1786 - val_accuracy: 0.9460\n",
       "Epoch 68/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1282 - accuracy: 0.9470 - val_loss: 0.2390 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1313 - accuracy: 0.9520 - val_loss: 0.1740 - val_accuracy: 0.9460\n",
       "Epoch 69/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1217 - accuracy: 0.9560 - val_loss: 0.2338 - val_accuracy: 0.9360\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1381 - accuracy: 0.9480 - val_loss: 0.1416 - val_accuracy: 0.9500\n",
       "Epoch 70/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1419 - accuracy: 0.9430 - val_loss: 0.2301 - val_accuracy: 0.9360\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1215 - accuracy: 0.9520 - val_loss: 0.1763 - val_accuracy: 0.9520\n",
       "Epoch 71/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1399 - accuracy: 0.9510 - val_loss: 0.2073 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1151 - accuracy: 0.9620 - val_loss: 0.1739 - val_accuracy: 0.9440\n",
       "Epoch 72/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1381 - accuracy: 0.9520 - val_loss: 0.2262 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.1127 - accuracy: 0.9560 - val_loss: 0.1543 - val_accuracy: 0.9540\n",
       "Epoch 73/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1279 - accuracy: 0.9560 - val_loss: 0.2129 - val_accuracy: 0.9300\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1036 - accuracy: 0.9570 - val_loss: 0.1556 - val_accuracy: 0.9540\n",
       "Epoch 74/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1240 - accuracy: 0.9490 - val_loss: 0.2141 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1224 - accuracy: 0.9610 - val_loss: 0.1873 - val_accuracy: 0.9420\n",
       "Epoch 75/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1207 - accuracy: 0.9540 - val_loss: 0.2288 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1138 - accuracy: 0.9570 - val_loss: 0.1491 - val_accuracy: 0.9560\n",
       "Epoch 76/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1138 - accuracy: 0.9530 - val_loss: 0.2255 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1286 - accuracy: 0.9500 - val_loss: 0.1501 - val_accuracy: 0.9560\n",
       "Epoch 77/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1191 - accuracy: 0.9510 - val_loss: 0.2191 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1081 - accuracy: 0.9670 - val_loss: 0.1552 - val_accuracy: 0.9560\n",
       "Epoch 78/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1017 - accuracy: 0.9700 - val_loss: 0.2645 - val_accuracy: 0.9360\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1077 - accuracy: 0.9560 - val_loss: 0.1570 - val_accuracy: 0.9520\n",
       "Epoch 79/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1359 - accuracy: 0.9460 - val_loss: 0.2675 - val_accuracy: 0.9360\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0998 - accuracy: 0.9580 - val_loss: 0.1731 - val_accuracy: 0.9460\n",
       "Epoch 80/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1153 - accuracy: 0.9560 - val_loss: 0.2319 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1047 - accuracy: 0.9650 - val_loss: 0.1731 - val_accuracy: 0.9420\n",
       "Epoch 81/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1097 - accuracy: 0.9600 - val_loss: 0.2731 - val_accuracy: 0.9340\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1094 - accuracy: 0.9580 - val_loss: 0.1619 - val_accuracy: 0.9560\n",
       "Epoch 82/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1097 - accuracy: 0.9570 - val_loss: 0.2272 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1023 - accuracy: 0.9610 - val_loss: 0.1446 - val_accuracy: 0.9580\n",
       "Epoch 83/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1172 - accuracy: 0.9550 - val_loss: 0.2092 - val_accuracy: 0.9340\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1030 - accuracy: 0.9610 - val_loss: 0.1587 - val_accuracy: 0.9500\n",
       "Epoch 84/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1103 - accuracy: 0.9610 - val_loss: 0.2309 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1048 - accuracy: 0.9650 - val_loss: 0.1574 - val_accuracy: 0.9500\n",
       "Epoch 85/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1154 - accuracy: 0.9520 - val_loss: 0.2622 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1002 - accuracy: 0.9610 - val_loss: 0.1545 - val_accuracy: 0.9500\n",
       "Epoch 86/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1200 - accuracy: 0.9530 - val_loss: 0.2371 - val_accuracy: 0.9500\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0954 - accuracy: 0.9630 - val_loss: 0.1382 - val_accuracy: 0.9540\n",
       "Epoch 87/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1055 - accuracy: 0.9700 - val_loss: 0.2105 - val_accuracy: 0.9500\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.1036 - accuracy: 0.9600 - val_loss: 0.1649 - val_accuracy: 0.9440\n",
       "Epoch 88/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1018 - accuracy: 0.9650 - val_loss: 0.2094 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1002 - accuracy: 0.9660 - val_loss: 0.1341 - val_accuracy: 0.9540\n",
       "Epoch 89/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1063 - accuracy: 0.9530 - val_loss: 0.2097 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0916 - accuracy: 0.9650 - val_loss: 0.1665 - val_accuracy: 0.9560\n",
       "Epoch 90/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0978 - accuracy: 0.9680 - val_loss: 0.2384 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1095 - accuracy: 0.9550 - val_loss: 0.1618 - val_accuracy: 0.9460\n",
       "Epoch 91/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1014 - accuracy: 0.9620 - val_loss: 0.2236 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0856 - accuracy: 0.9690 - val_loss: 0.1743 - val_accuracy: 0.9520\n",
       "Epoch 92/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0972 - accuracy: 0.9630 - val_loss: 0.2487 - val_accuracy: 0.9500\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0884 - accuracy: 0.9670 - val_loss: 0.1768 - val_accuracy: 0.9460\n",
       "Epoch 93/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1024 - accuracy: 0.9540 - val_loss: 0.2414 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.1046 - accuracy: 0.9620 - val_loss: 0.1447 - val_accuracy: 0.9520\n",
       "Epoch 94/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1052 - accuracy: 0.9590 - val_loss: 0.2092 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0902 - accuracy: 0.9670 - val_loss: 0.1609 - val_accuracy: 0.9560\n",
       "Epoch 95/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1050 - accuracy: 0.9610 - val_loss: 0.2243 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0933 - accuracy: 0.9610 - val_loss: 0.1553 - val_accuracy: 0.9440\n",
       "Epoch 96/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1019 - accuracy: 0.9610 - val_loss: 0.2950 - val_accuracy: 0.9320\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.1032 - accuracy: 0.9640 - val_loss: 0.1623 - val_accuracy: 0.9480\n",
       "Epoch 97/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0995 - accuracy: 0.9630 - val_loss: 0.2321 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0930 - accuracy: 0.9670 - val_loss: 0.1525 - val_accuracy: 0.9540\n",
       "Epoch 98/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1061 - accuracy: 0.9570 - val_loss: 0.2559 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0846 - accuracy: 0.9710 - val_loss: 0.1552 - val_accuracy: 0.9560\n",
       "Epoch 99/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1039 - accuracy: 0.9620 - val_loss: 0.2509 - val_accuracy: 0.9380\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0924 - accuracy: 0.9730 - val_loss: 0.1450 - val_accuracy: 0.9560\n",
       "Epoch 100/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1049 - accuracy: 0.9610 - val_loss: 0.2688 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0834 - accuracy: 0.9690 - val_loss: 0.1532 - val_accuracy: 0.9580\n",
       "Epoch 101/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0927 - accuracy: 0.9680 - val_loss: 0.2364 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0835 - accuracy: 0.9710 - val_loss: 0.1839 - val_accuracy: 0.9500\n",
       "Epoch 102/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0847 - accuracy: 0.9700 - val_loss: 0.2312 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0912 - accuracy: 0.9630 - val_loss: 0.1401 - val_accuracy: 0.9540\n",
       "Epoch 103/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0912 - accuracy: 0.9690 - val_loss: 0.2327 - val_accuracy: 0.9500\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0783 - accuracy: 0.9670 - val_loss: 0.1509 - val_accuracy: 0.9540\n",
       "Epoch 104/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0960 - accuracy: 0.9650 - val_loss: 0.2847 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0864 - accuracy: 0.9680 - val_loss: 0.1857 - val_accuracy: 0.9480\n",
       "Epoch 105/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1006 - accuracy: 0.9620 - val_loss: 0.2211 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0861 - accuracy: 0.9700 - val_loss: 0.1864 - val_accuracy: 0.9520\n",
       "Epoch 106/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0961 - accuracy: 0.9630 - val_loss: 0.2745 - val_accuracy: 0.9500\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0845 - accuracy: 0.9710 - val_loss: 0.1670 - val_accuracy: 0.9580\n",
       "Epoch 107/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1070 - accuracy: 0.9560 - val_loss: 0.2594 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0827 - accuracy: 0.9690 - val_loss: 0.1561 - val_accuracy: 0.9540\n",
       "Epoch 108/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1254 - accuracy: 0.9590 - val_loss: 0.2340 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0894 - accuracy: 0.9670 - val_loss: 0.1435 - val_accuracy: 0.9620\n",
       "Epoch 109/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1024 - accuracy: 0.9650 - val_loss: 0.2181 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0869 - accuracy: 0.9670 - val_loss: 0.1437 - val_accuracy: 0.9620\n",
       "Epoch 110/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0868 - accuracy: 0.9720 - val_loss: 0.2271 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0901 - accuracy: 0.9710 - val_loss: 0.1522 - val_accuracy: 0.9540\n",
       "Epoch 111/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0975 - accuracy: 0.9640 - val_loss: 0.2516 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0909 - accuracy: 0.9660 - val_loss: 0.1534 - val_accuracy: 0.9560\n",
       "Epoch 112/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1016 - accuracy: 0.9650 - val_loss: 0.2885 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0906 - accuracy: 0.9640 - val_loss: 0.1945 - val_accuracy: 0.9400\n",
       "Epoch 113/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0916 - accuracy: 0.9710 - val_loss: 0.2423 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0859 - accuracy: 0.9650 - val_loss: 0.1379 - val_accuracy: 0.9660\n",
       "Epoch 114/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0841 - accuracy: 0.9690 - val_loss: 0.2581 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0724 - accuracy: 0.9750 - val_loss: 0.1440 - val_accuracy: 0.9580\n",
       "Epoch 115/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0882 - accuracy: 0.9630 - val_loss: 0.2669 - val_accuracy: 0.9500\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0895 - accuracy: 0.9640 - val_loss: 0.1358 - val_accuracy: 0.9540\n",
       "Epoch 116/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1016 - accuracy: 0.9660 - val_loss: 0.2654 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0900 - accuracy: 0.9670 - val_loss: 0.1455 - val_accuracy: 0.9600\n",
       "Epoch 117/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0862 - accuracy: 0.9700 - val_loss: 0.2387 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0775 - accuracy: 0.9790 - val_loss: 0.1643 - val_accuracy: 0.9500\n",
       "Epoch 118/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0773 - accuracy: 0.9720 - val_loss: 0.2819 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0740 - accuracy: 0.9740 - val_loss: 0.1576 - val_accuracy: 0.9560\n",
       "Epoch 119/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0809 - accuracy: 0.9700 - val_loss: 0.2418 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0753 - accuracy: 0.9760 - val_loss: 0.1365 - val_accuracy: 0.9600\n",
       "Epoch 120/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0876 - accuracy: 0.9650 - val_loss: 0.2794 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0773 - accuracy: 0.9710 - val_loss: 0.1265 - val_accuracy: 0.9620\n",
       "Epoch 121/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0789 - accuracy: 0.9670 - val_loss: 0.2851 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0591 - accuracy: 0.9800 - val_loss: 0.1474 - val_accuracy: 0.9580\n",
       "Epoch 122/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0885 - accuracy: 0.9660 - val_loss: 0.2515 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0687 - accuracy: 0.9810 - val_loss: 0.1552 - val_accuracy: 0.9560\n",
       "Epoch 123/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0753 - accuracy: 0.9750 - val_loss: 0.2810 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0763 - accuracy: 0.9710 - val_loss: 0.1436 - val_accuracy: 0.9640\n",
       "Epoch 124/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0856 - accuracy: 0.9670 - val_loss: 0.2588 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0723 - accuracy: 0.9750 - val_loss: 0.1541 - val_accuracy: 0.9620\n",
       "Epoch 125/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0816 - accuracy: 0.9680 - val_loss: 0.2927 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0763 - accuracy: 0.9750 - val_loss: 0.1231 - val_accuracy: 0.9580\n",
       "Epoch 126/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0706 - accuracy: 0.9760 - val_loss: 0.2977 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0732 - accuracy: 0.9720 - val_loss: 0.1436 - val_accuracy: 0.9540\n",
       "Epoch 127/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0725 - accuracy: 0.9750 - val_loss: 0.2728 - val_accuracy: 0.9380\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0739 - accuracy: 0.9760 - val_loss: 0.1506 - val_accuracy: 0.9620\n",
       "Epoch 128/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0763 - accuracy: 0.9750 - val_loss: 0.2548 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0785 - accuracy: 0.9720 - val_loss: 0.1586 - val_accuracy: 0.9580\n",
       "Epoch 129/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0759 - accuracy: 0.9710 - val_loss: 0.2826 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0763 - accuracy: 0.9730 - val_loss: 0.1500 - val_accuracy: 0.9560\n",
       "Epoch 130/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0879 - accuracy: 0.9630 - val_loss: 0.2986 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0782 - accuracy: 0.9710 - val_loss: 0.1663 - val_accuracy: 0.9580\n",
       "Epoch 131/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0799 - accuracy: 0.9710 - val_loss: 0.3120 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0718 - accuracy: 0.9730 - val_loss: 0.1700 - val_accuracy: 0.9540\n",
       "Epoch 132/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0847 - accuracy: 0.9690 - val_loss: 0.2735 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0817 - accuracy: 0.9670 - val_loss: 0.2157 - val_accuracy: 0.9500\n",
       "Epoch 133/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0771 - accuracy: 0.9770 - val_loss: 0.2694 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0763 - accuracy: 0.9730 - val_loss: 0.1486 - val_accuracy: 0.9480\n",
       "Epoch 134/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0682 - accuracy: 0.9750 - val_loss: 0.3082 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0762 - accuracy: 0.9700 - val_loss: 0.1318 - val_accuracy: 0.9620\n",
       "Epoch 135/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0806 - accuracy: 0.9680 - val_loss: 0.2539 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0645 - accuracy: 0.9800 - val_loss: 0.1329 - val_accuracy: 0.9580\n",
       "Epoch 136/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0849 - accuracy: 0.9630 - val_loss: 0.3257 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0639 - accuracy: 0.9760 - val_loss: 0.1366 - val_accuracy: 0.9560\n",
       "Epoch 137/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0833 - accuracy: 0.9640 - val_loss: 0.2552 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0712 - accuracy: 0.9740 - val_loss: 0.1550 - val_accuracy: 0.9540\n",
       "Epoch 138/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0960 - accuracy: 0.9680 - val_loss: 0.2204 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0659 - accuracy: 0.9760 - val_loss: 0.1357 - val_accuracy: 0.9560\n",
       "Epoch 139/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0839 - accuracy: 0.9680 - val_loss: 0.2936 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0648 - accuracy: 0.9780 - val_loss: 0.1707 - val_accuracy: 0.9560\n",
       "Epoch 140/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0903 - accuracy: 0.9680 - val_loss: 0.2940 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0793 - accuracy: 0.9720 - val_loss: 0.1829 - val_accuracy: 0.9580\n",
       "Epoch 141/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1049 - accuracy: 0.9580 - val_loss: 0.3010 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0615 - accuracy: 0.9780 - val_loss: 0.1450 - val_accuracy: 0.9560\n",
       "Epoch 142/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0876 - accuracy: 0.9600 - val_loss: 0.2868 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0610 - accuracy: 0.9760 - val_loss: 0.1523 - val_accuracy: 0.9620\n",
       "Epoch 143/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0758 - accuracy: 0.9700 - val_loss: 0.2639 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0667 - accuracy: 0.9720 - val_loss: 0.1497 - val_accuracy: 0.9640\n",
       "Epoch 144/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0780 - accuracy: 0.9670 - val_loss: 0.2530 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0564 - accuracy: 0.9810 - val_loss: 0.1453 - val_accuracy: 0.9600\n",
       "Epoch 145/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0775 - accuracy: 0.9690 - val_loss: 0.2850 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0646 - accuracy: 0.9790 - val_loss: 0.1548 - val_accuracy: 0.9620\n",
       "Epoch 146/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0742 - accuracy: 0.9740 - val_loss: 0.2949 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0708 - accuracy: 0.9730 - val_loss: 0.1813 - val_accuracy: 0.9460\n",
       "Epoch 147/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0695 - accuracy: 0.9740 - val_loss: 0.2926 - val_accuracy: 0.9500\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0674 - accuracy: 0.9770 - val_loss: 0.1538 - val_accuracy: 0.9560\n",
       "Epoch 148/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0837 - accuracy: 0.9760 - val_loss: 0.2493 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0651 - accuracy: 0.9720 - val_loss: 0.1534 - val_accuracy: 0.9560\n",
       "Epoch 149/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0720 - accuracy: 0.9710 - val_loss: 0.2587 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0771 - accuracy: 0.9680 - val_loss: 0.1621 - val_accuracy: 0.9580\n",
       "Epoch 150/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0842 - accuracy: 0.9670 - val_loss: 0.2406 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0614 - accuracy: 0.9720 - val_loss: 0.1489 - val_accuracy: 0.9520\n",
       "Epoch 151/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0727 - accuracy: 0.9720 - val_loss: 0.2340 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0515 - accuracy: 0.9790 - val_loss: 0.1685 - val_accuracy: 0.9620\n",
       "Epoch 152/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0665 - accuracy: 0.9800 - val_loss: 0.2706 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0691 - accuracy: 0.9770 - val_loss: 0.1426 - val_accuracy: 0.9660\n",
       "Epoch 153/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0659 - accuracy: 0.9750 - val_loss: 0.2869 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0515 - accuracy: 0.9820 - val_loss: 0.1480 - val_accuracy: 0.9620\n",
       "Epoch 154/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0787 - accuracy: 0.9780 - val_loss: 0.2415 - val_accuracy: 0.9500\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0574 - accuracy: 0.9780 - val_loss: 0.1584 - val_accuracy: 0.9640\n",
       "Epoch 155/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0763 - accuracy: 0.9740 - val_loss: 0.2859 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0719 - accuracy: 0.9810 - val_loss: 0.1618 - val_accuracy: 0.9540\n",
       "Epoch 156/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0677 - accuracy: 0.9760 - val_loss: 0.2918 - val_accuracy: 0.9520\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0591 - accuracy: 0.9820 - val_loss: 0.1414 - val_accuracy: 0.9660\n",
       "Epoch 157/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0709 - accuracy: 0.9730 - val_loss: 0.3029 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0589 - accuracy: 0.9800 - val_loss: 0.1659 - val_accuracy: 0.9600\n",
       "Epoch 158/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0758 - accuracy: 0.9730 - val_loss: 0.2892 - val_accuracy: 0.9380\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0532 - accuracy: 0.9800 - val_loss: 0.1416 - val_accuracy: 0.9540\n",
       "Epoch 159/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0868 - accuracy: 0.9700 - val_loss: 0.2268 - val_accuracy: 0.9520\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0648 - accuracy: 0.9770 - val_loss: 0.1404 - val_accuracy: 0.9660\n",
       "Epoch 160/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0734 - accuracy: 0.9740 - val_loss: 0.3152 - val_accuracy: 0.9400\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0511 - accuracy: 0.9830 - val_loss: 0.1544 - val_accuracy: 0.9560\n",
       "Epoch 161/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0704 - accuracy: 0.9790 - val_loss: 0.3017 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0540 - accuracy: 0.9820 - val_loss: 0.1353 - val_accuracy: 0.9620\n",
       "Epoch 162/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0656 - accuracy: 0.9760 - val_loss: 0.3216 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0565 - accuracy: 0.9790 - val_loss: 0.1345 - val_accuracy: 0.9600\n",
       "Epoch 163/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0697 - accuracy: 0.9770 - val_loss: 0.2788 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0697 - accuracy: 0.9770 - val_loss: 0.1415 - val_accuracy: 0.9580\n",
       "Epoch 164/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0636 - accuracy: 0.9720 - val_loss: 0.2572 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0644 - accuracy: 0.9750 - val_loss: 0.1551 - val_accuracy: 0.9580\n",
       "Epoch 165/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0615 - accuracy: 0.9790 - val_loss: 0.3042 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0577 - accuracy: 0.9760 - val_loss: 0.1303 - val_accuracy: 0.9560\n",
       "Epoch 166/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0644 - accuracy: 0.9790 - val_loss: 0.2545 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0582 - accuracy: 0.9790 - val_loss: 0.1312 - val_accuracy: 0.9560\n",
       "Epoch 167/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0529 - accuracy: 0.9790 - val_loss: 0.2543 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0499 - accuracy: 0.9850 - val_loss: 0.1152 - val_accuracy: 0.9600\n",
       "Epoch 168/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0741 - accuracy: 0.9700 - val_loss: 0.2757 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0596 - accuracy: 0.9790 - val_loss: 0.1523 - val_accuracy: 0.9560\n",
       "Epoch 169/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0742 - accuracy: 0.9720 - val_loss: 0.2751 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0574 - accuracy: 0.9760 - val_loss: 0.1664 - val_accuracy: 0.9500\n",
       "Epoch 170/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0727 - accuracy: 0.9760 - val_loss: 0.2855 - val_accuracy: 0.9500\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0580 - accuracy: 0.9820 - val_loss: 0.1298 - val_accuracy: 0.9560\n",
       "Epoch 171/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0537 - accuracy: 0.9830 - val_loss: 0.3245 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0571 - accuracy: 0.9780 - val_loss: 0.1265 - val_accuracy: 0.9580\n",
       "Epoch 172/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0587 - accuracy: 0.9780 - val_loss: 0.2885 - val_accuracy: 0.9500\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0702 - accuracy: 0.9730 - val_loss: 0.1356 - val_accuracy: 0.9680\n",
       "Epoch 173/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0596 - accuracy: 0.9790 - val_loss: 0.2881 - val_accuracy: 0.9500\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0554 - accuracy: 0.9770 - val_loss: 0.1362 - val_accuracy: 0.9660\n",
       "Epoch 174/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0646 - accuracy: 0.9780 - val_loss: 0.3350 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0453 - accuracy: 0.9850 - val_loss: 0.1303 - val_accuracy: 0.9660\n",
       "Epoch 175/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.1166 - accuracy: 0.9720 - val_loss: 0.2887 - val_accuracy: 0.9500\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0485 - accuracy: 0.9840 - val_loss: 0.1358 - val_accuracy: 0.9640\n",
       "Epoch 176/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0715 - accuracy: 0.9730 - val_loss: 0.2311 - val_accuracy: 0.9520\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0476 - accuracy: 0.9800 - val_loss: 0.1425 - val_accuracy: 0.9580\n",
       "Epoch 177/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0557 - accuracy: 0.9780 - val_loss: 0.3196 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0490 - accuracy: 0.9780 - val_loss: 0.1385 - val_accuracy: 0.9640\n",
       "Epoch 178/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0575 - accuracy: 0.9830 - val_loss: 0.3027 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0752 - accuracy: 0.9710 - val_loss: 0.1669 - val_accuracy: 0.9440\n",
       "Epoch 179/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0606 - accuracy: 0.9770 - val_loss: 0.3084 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0619 - accuracy: 0.9760 - val_loss: 0.1400 - val_accuracy: 0.9620\n",
       "Epoch 180/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0760 - accuracy: 0.9750 - val_loss: 0.3056 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0571 - accuracy: 0.9760 - val_loss: 0.1345 - val_accuracy: 0.9620\n",
       "Epoch 181/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0612 - accuracy: 0.9790 - val_loss: 0.3167 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0559 - accuracy: 0.9770 - val_loss: 0.1763 - val_accuracy: 0.9600\n",
       "Epoch 182/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0576 - accuracy: 0.9800 - val_loss: 0.2263 - val_accuracy: 0.9520\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0525 - accuracy: 0.9810 - val_loss: 0.1630 - val_accuracy: 0.9620\n",
       "Epoch 183/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0543 - accuracy: 0.9800 - val_loss: 0.3170 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0562 - accuracy: 0.9860 - val_loss: 0.1318 - val_accuracy: 0.9540\n",
       "Epoch 184/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0544 - accuracy: 0.9760 - val_loss: 0.2598 - val_accuracy: 0.9500\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0422 - accuracy: 0.9870 - val_loss: 0.1480 - val_accuracy: 0.9500\n",
       "Epoch 185/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0604 - accuracy: 0.9810 - val_loss: 0.2624 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0485 - accuracy: 0.9840 - val_loss: 0.1447 - val_accuracy: 0.9620\n",
       "Epoch 186/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0594 - accuracy: 0.9770 - val_loss: 0.2637 - val_accuracy: 0.9420\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0545 - accuracy: 0.9850 - val_loss: 0.1613 - val_accuracy: 0.9500\n",
       "Epoch 187/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0595 - accuracy: 0.9760 - val_loss: 0.2523 - val_accuracy: 0.9560\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0694 - accuracy: 0.9740 - val_loss: 0.1139 - val_accuracy: 0.9640\n",
       "Epoch 188/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0627 - accuracy: 0.9750 - val_loss: 0.2721 - val_accuracy: 0.9500\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0475 - accuracy: 0.9810 - val_loss: 0.1370 - val_accuracy: 0.9680\n",
       "Epoch 189/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0508 - accuracy: 0.9820 - val_loss: 0.2766 - val_accuracy: 0.9520\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0637 - accuracy: 0.9760 - val_loss: 0.1849 - val_accuracy: 0.9460\n",
       "Epoch 190/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0614 - accuracy: 0.9730 - val_loss: 0.2828 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0618 - accuracy: 0.9770 - val_loss: 0.1760 - val_accuracy: 0.9480\n",
       "Epoch 191/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0613 - accuracy: 0.9780 - val_loss: 0.3344 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0493 - accuracy: 0.9840 - val_loss: 0.2082 - val_accuracy: 0.9560\n",
       "Epoch 192/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0597 - accuracy: 0.9770 - val_loss: 0.2588 - val_accuracy: 0.9460\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0702 - accuracy: 0.9720 - val_loss: 0.1435 - val_accuracy: 0.9700\n",
       "Epoch 193/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0596 - accuracy: 0.9800 - val_loss: 0.2628 - val_accuracy: 0.9520\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0468 - accuracy: 0.9820 - val_loss: 0.1708 - val_accuracy: 0.9600\n",
       "Epoch 194/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0611 - accuracy: 0.9750 - val_loss: 0.2894 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0532 - accuracy: 0.9790 - val_loss: 0.1625 - val_accuracy: 0.9660\n",
       "Epoch 195/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0535 - accuracy: 0.9840 - val_loss: 0.2822 - val_accuracy: 0.9480\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0425 - accuracy: 0.9820 - val_loss: 0.1651 - val_accuracy: 0.9600\n",
       "Epoch 196/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0498 - accuracy: 0.9780 - val_loss: 0.2576 - val_accuracy: 0.9540\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0485 - accuracy: 0.9790 - val_loss: 0.1534 - val_accuracy: 0.9680\n",
       "Epoch 197/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0562 - accuracy: 0.9790 - val_loss: 0.2912 - val_accuracy: 0.9440\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0661 - accuracy: 0.9760 - val_loss: 0.2143 - val_accuracy: 0.9500\n",
       "Epoch 198/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0534 - accuracy: 0.9770 - val_loss: 0.2598 - val_accuracy: 0.9500\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0714 - accuracy: 0.9760 - val_loss: 0.1700 - val_accuracy: 0.9580\n",
       "Epoch 199/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0464 - accuracy: 0.9810 - val_loss: 0.2861 - val_accuracy: 0.9500\n",
+      "32/32 [==============================] - 0s 2ms/step - loss: 0.0912 - accuracy: 0.9740 - val_loss: 0.1616 - val_accuracy: 0.9560\n",
       "Epoch 200/200\n",
-      "32/32 [==============================] - 0s 2ms/step - loss: 0.0522 - accuracy: 0.9800 - val_loss: 0.3037 - val_accuracy: 0.9520\n"
+      "32/32 [==============================] - 0s 3ms/step - loss: 0.0502 - accuracy: 0.9830 - val_loss: 0.1710 - val_accuracy: 0.9620\n"
      ]
     }
    ],
@@ -760,7 +752,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 14,
+   "execution_count": 23,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -777,7 +769,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 11,
+   "execution_count": 26,
    "metadata": {
     "colab": {
      "base_uri": "https://localhost:8080/",
@@ -792,40 +784,36 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "INFO:tensorflow:Assets written to: /var/folders/43/4pdy4jr56rx76bghv9vbsx140000gn/T/tmpg7k__9oe/assets\n"
+      "INFO:tensorflow:Assets written to: /var/folders/43/4pdy4jr56rx76bghv9vbsx140000gn/T/tmpy78vg8pq/assets\n"
      ]
     },
     {
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "INFO:tensorflow:Assets written to: /var/folders/43/4pdy4jr56rx76bghv9vbsx140000gn/T/tmpg7k__9oe/assets\n"
+      "INFO:tensorflow:Assets written to: /var/folders/43/4pdy4jr56rx76bghv9vbsx140000gn/T/tmpy78vg8pq/assets\n"
      ]
     },
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "INFO:tensorflow:Assets written to: /var/folders/43/4pdy4jr56rx76bghv9vbsx140000gn/T/tmp8rz78j_u/assets\n"
+      "INFO:tensorflow:Assets written to: /var/folders/43/4pdy4jr56rx76bghv9vbsx140000gn/T/tmpvq5_xed6/assets\n"
      ]
     },
     {
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "INFO:tensorflow:Assets written to: /var/folders/43/4pdy4jr56rx76bghv9vbsx140000gn/T/tmp8rz78j_u/assets\n"
+      "INFO:tensorflow:Assets written to: /var/folders/43/4pdy4jr56rx76bghv9vbsx140000gn/T/tmpvq5_xed6/assets\n"
      ]
     },
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "RUN tensorflowjs_converter --input_format keras --output_format tfjs_layers_model built/model.h5 built/ml4f/converted.tfjs\n",
+      "RUN tensorflowjs_converter --input_format keras --output_format tfjs_layers_model built/model.h5 built/converted.tfjs\n",
       "RUN OK\n",
-      "\n",
-      "============================\n",
-      "Hi there 👋. Looks like you are running TensorFlow.js in Node.js. To speed things up dramatically, install our node backend, which binds to TensorFlow C++, by running npm i @tensorflow/tfjs-node, or npm i @tensorflow/tfjs-node-gpu if you have CUDA. Then call require('@tensorflow/tfjs-node'); (-gpu suffix for CUDA) at the start of your program. Visit https://github.com/tensorflow/tfjs-node for more details.\n",
-      "============================\n",
       "Validating partial models...\n",
       "testing Conv2D: [50,3,1] => [47,1,16]...\n",
       "also with no activation...\n",
@@ -837,15 +825,23 @@
       "testing Dense: [336] => [4]...\n",
       "also with no activation...\n",
       "Compiling full model...\n",
-      "write built/ml4f/model.asm (91809 bytes)\n",
-      "write built/ml4f/model.js (18056 bytes)\n",
-      "write built/ml4f/model.bin (13052 bytes)\n",
+      "write built/model-float32.asm (91742 bytes)\n",
+      "write built/model-float32.js (18056 bytes)\n",
+      "write built/model-float32.ml4f (13052 bytes)\n",
+      "write built/model-float32.txt (235 bytes)\n",
+      "\n",
+      "\n",
+      "*** built/model-float32.ml4f\n",
+      "\n",
+      "Accuracy: 0.9600\n",
+      "  231    1    0    1\n",
+      "    3   91    1    0\n",
+      "    5    0   78    0\n",
+      "    7    1    1   80\n",
+      "\n",
       "model: 12.75k; code: 2.46k (19.3%); arena: 4.38k; test 0.00k\n",
       "total cycles: 225149 (2.680ms at 84MHz)\n",
       "\n",
-      "============================\n",
-      "Hi there 👋. Looks like you are running TensorFlow.js in Node.js. To speed things up dramatically, install our node backend, which binds to TensorFlow C++, by running npm i @tensorflow/tfjs-node, or npm i @tensorflow/tfjs-node-gpu if you have CUDA. Then call require('@tensorflow/tfjs-node'); (-gpu suffix for CUDA) at the start of your program. Visit https://github.com/tensorflow/tfjs-node for more details.\n",
-      "============================\n",
       "Validating partial models...\n",
       "testing Conv2D: [50,3,1] => [47,1,16]...\n",
       "also with no activation...\n",
@@ -857,11 +853,23 @@
       "testing Dense: [336] => [4]...\n",
       "also with no activation...\n",
       "Compiling full model...\n",
-      "write built/ml4f-f16/model.asm (97312 bytes)\n",
-      "write built/ml4f-f16/model.js (17076 bytes)\n",
-      "write built/ml4f-f16/model.bin (8112 bytes)\n",
+      "write built/model-float16.asm (97245 bytes)\n",
+      "write built/model-float16.js (17076 bytes)\n",
+      "write built/model-float16.ml4f (8112 bytes)\n",
+      "write built/model-float16.txt (234 bytes)\n",
+      "\n",
+      "\n",
+      "*** built/model-float16.ml4f\n",
+      "\n",
+      "Accuracy: 0.9600\n",
+      "  231    1    0    1\n",
+      "    3   91    1    0\n",
+      "    5    0   78    0\n",
+      "    7    1    1   80\n",
+      "\n",
       "model: 7.92k; code: 2.51k (31.7%); arena: 4.38k; test 0.00k\n",
-      "total cycles: 227036 (2.703ms at 84MHz)\n"
+      "total cycles: 227036 (2.703ms at 84MHz)\n",
+      "\n"
      ]
     }
    ],
@@ -891,9 +899,9 @@
     "# Save the model to disk\n",
     "open(\"built/model-int8.tflite\", \"wb\").write(model_tflite)\n",
     "\n",
-    "! ml4f --output built/ml4f built/model.h5\n",
+    "! ml4f           --basename model-float32 --eval built/test.json built/model.h5\n",
     "# to speed things up, use the tfjs model from previous step\n",
-    "! ml4f --float16 --output built/ml4f-f16 built/ml4f/converted.tfjs\n"
+    "! ml4f --float16 --basename model-float16 --eval built/test.json built/converted.tfjs\n"
    ]
   },
   {
@@ -905,22 +913,22 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 12,
+   "execution_count": 27,
    "metadata": {},
    "outputs": [
     {
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "-rw-r--r-- 1 michal staff  8112 Oct 15 13:04 built/ml4f-f16/model.bin\n",
-      "-rw-r--r-- 1 michal staff 13052 Oct 15 13:04 built/ml4f/model.bin\n",
-      "-rw-r--r-- 1 michal staff 13420 Oct 15 13:04 built/model-float32.tflite\n",
-      "-rw-r--r-- 1 michal staff  8480 Oct 15 13:04 built/model-int8.tflite\n"
+      "-rw-r--r-- 1 michal staff  8112 Oct 16 15:20 built/model-float16.ml4f\n",
+      "-rw-r--r-- 1 michal staff 13052 Oct 16 15:20 built/model-float32.ml4f\n",
+      "-rw-r--r-- 1 michal staff 13420 Oct 16 15:20 built/model-float32.tflite\n",
+      "-rw-r--r-- 1 michal staff  8480 Oct 16 15:20 built/model-int8.tflite\n"
      ]
     }
    ],
    "source": [
-    "! ls -l built/*tflite built/ml4f*/*.bin"
+    "! ls -l built/*.tflite built/*.ml4f"
    ]
   },
   {
@@ -935,7 +943,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 17,
+   "execution_count": 28,
    "metadata": {
     "colab": {
      "base_uri": "https://localhost:8080/",
@@ -953,50 +961,54 @@
       "\n",
       "*** built/model.h5\n",
       "\n",
-      "Accuracy: 0.964000\n",
+      "Accuracy: 0.960000\n",
       "tf.Tensor(\n",
-      "[[244   6   1   3]\n",
-      " [  3  66   1   0]\n",
-      " [  1   0 100   0]\n",
-      " [  3   0   0  72]], shape=(4, 4), dtype=int32)\n",
+      "[[231   3   5   7]\n",
+      " [  1  91   0   1]\n",
+      " [  0   1  78   1]\n",
+      " [  1   0   0  80]], shape=(4, 4), dtype=int32)\n",
       "\n",
       "*** built/model-float32.tflite\n",
       "\n",
-      "Accuracy: 0.964000\n",
+      "Accuracy: 0.960000\n",
       "tf.Tensor(\n",
-      "[[244   6   1   3]\n",
-      " [  3  66   1   0]\n",
-      " [  1   0 100   0]\n",
-      " [  3   0   0  72]], shape=(4, 4), dtype=int32)\n",
+      "[[231   3   5   7]\n",
+      " [  1  91   0   1]\n",
+      " [  0   1  78   1]\n",
+      " [  1   0   0  80]], shape=(4, 4), dtype=int32)\n",
       "\n",
       "*** built/model-int8.tflite\n",
       "\n",
-      "Accuracy: 0.960000\n",
+      "Accuracy: 0.966000\n",
       "tf.Tensor(\n",
-      "[[246   5   1   2]\n",
-      " [  4  66   0   0]\n",
-      " [  2   1  98   0]\n",
-      " [  3   1   1  70]], shape=(4, 4), dtype=int32)\n",
+      "[[233   5   3   5]\n",
+      " [  1  92   0   0]\n",
+      " [  0   1  79   0]\n",
+      " [  1   1   0  79]], shape=(4, 4), dtype=int32)\n",
       "\n",
-      "*** built/ml4f/model.bin\n",
-      "Accuracy: 0.9640\n",
-      "  244    3    1    3\n",
-      "    6   66    0    0\n",
-      "    1    1  100    0\n",
-      "    3    0    0   72\n",
+      "*** built/model-float16.ml4f\n",
+      "\n",
+      "Accuracy: 0.9600\n",
+      "  231    1    0    1\n",
+      "    3   91    1    0\n",
+      "    5    0   78    0\n",
+      "    7    1    1   80\n",
+      "\n",
+      "model: 7.92k; code: 2.51k (31.7%); arena: 4.38k; test 0.00k\n",
+      "total cycles: 227036 (2.703ms at 84MHz)\n",
+      "\n",
+      "\n",
+      "*** built/model-float32.ml4f\n",
+      "\n",
+      "Accuracy: 0.9600\n",
+      "  231    1    0    1\n",
+      "    3   91    1    0\n",
+      "    5    0   78    0\n",
+      "    7    1    1   80\n",
       "\n",
       "model: 12.75k; code: 2.46k (19.3%); arena: 4.38k; test 0.00k\n",
       "total cycles: 225149 (2.680ms at 84MHz)\n",
-      "\n",
-      "*** built/ml4f-f16/model.bin\n",
-      "Accuracy: 0.9680\n",
-      "  246    3    1    3\n",
-      "    5   66    0    0\n",
-      "    1    1  100    0\n",
-      "    2    0    0   72\n",
-      "\n",
-      "model: 7.92k; code: 2.51k (31.7%); arena: 4.38k; test 0.00k\n",
-      "total cycles: 227036 (2.703ms at 84MHz)\n"
+      "\n"
      ]
     }
    ],
@@ -1042,9 +1054,8 @@
     "eval_tflite(\"built/model-float32.tflite\")\n",
     "eval_tflite(\"built/model-int8.tflite\")\n",
     "\n",
-    "# eval ml4f models\n",
-    "! ml4f --output built/ml4f built/ml4f/converted.tfjs -n --eval built/test.json\n",
-    "! ml4f --float16 --output built/ml4f-f16 built/ml4f/converted.tfjs -n --eval built/test.json"
+    "# show eval of ml4f models\n",
+    "! cat built/model-*.txt"
    ]
   },
   {