diff --git a/environment.yml b/environment.yml
index 9d1cc085..6f34864c 100644
--- a/environment.yml
+++ b/environment.yml
@@ -1,9 +1,9 @@
 name: graph_attribution
 channels:
 - rdkit
-- conda-forge
 dependencies:
 - pip
+- python=3.7
 - rdkit
 - numpy
 - ipython
@@ -20,3 +20,5 @@ dependencies:
         - tensorflow
         - git+https://github.com/deepmind/sonnet.git
         - graph_nets
+        - more-itertools
+        - ml-collections
diff --git a/notebooks/train_attribute_and_evaluate.ipynb b/notebooks/train_attribute_and_evaluate.ipynb
index 896e79c8..fe0b8fb4 100644
--- a/notebooks/train_attribute_and_evaluate.ipynb
+++ b/notebooks/train_attribute_and_evaluate.ipynb
@@ -59,7 +59,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 56,
+   "execution_count": 13,
    "metadata": {
     "ExecuteTime": {
      "end_time": "2020-11-19T20:01:09.511494Z",
@@ -75,7 +75,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "tensorflow           = 2.4.0-dev20200824\n",
+      "tensorflow           = 2.4.1\n",
       "sonnet               = 2.0.0\n",
       "graph_attribution    = 1.0.0b\n"
      ]
@@ -83,9 +83,10 @@
    ],
    "source": [
     "import os\n",
+    "import time\n",
     "import itertools\n",
     "import collections\n",
-    "import tqdm.auto as tqdm\n",
+    "from tqdm import tqdm\n",
     "\n",
     "from IPython.display import display\n",
     "\n",
@@ -121,7 +122,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 43,
+   "execution_count": 5,
    "metadata": {
     "ExecuteTime": {
      "end_time": "2020-11-19T19:58:14.416163Z",
@@ -162,7 +163,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 44,
+   "execution_count": 6,
    "metadata": {
     "ExecuteTime": {
      "end_time": "2020-11-19T19:58:18.979633Z",
@@ -191,7 +192,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 45,
+   "execution_count": 7,
    "metadata": {
     "ExecuteTime": {
      "end_time": "2020-11-19T19:58:20.566788Z",
@@ -238,7 +239,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 46,
+   "execution_count": 8,
    "metadata": {
     "ExecuteTime": {
      "end_time": "2020-11-19T19:58:20.635213Z",
@@ -264,7 +265,7 @@
        "task_type: logic7"
       ]
      },
-     "execution_count": 46,
+     "execution_count": 8,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -286,7 +287,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 47,
+   "execution_count": 9,
    "metadata": {
     "ExecuteTime": {
      "end_time": "2020-11-19T19:58:23.774392Z",
@@ -373,7 +374,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 58,
+   "execution_count": 14,
    "metadata": {
     "ExecuteTime": {
      "end_time": "2020-11-19T20:06:19.165556Z",
@@ -388,7 +389,639 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "100%|██████████| 300/300 [04:28<00:00,  1.12it/s, train=0.0344, test=0.099] \n"
+      "\n",
+      "  0%|          | 0/300 [00:10<?, ?it/s]\u001b[A\n",
+      "\n",
+      "  0%|          | 0/300 [00:03<?, ?it/s, train=0.4, test=0.338]\u001b[A\n",
+      "  0%|          | 1/300 [00:03<18:41,  3.75s/it, train=0.4, test=0.338]\u001b[A\n",
+      "  0%|          | 1/300 [00:04<18:41,  3.75s/it, train=0.312, test=0.329]\u001b[A\n",
+      "  1%|          | 2/300 [00:04<09:01,  1.82s/it, train=0.312, test=0.329]\u001b[A\n",
+      "  1%|          | 2/300 [00:04<09:01,  1.82s/it, train=0.308, test=0.326]\u001b[A\n",
+      "  1%|          | 3/300 [00:04<05:50,  1.18s/it, train=0.308, test=0.326]\u001b[A\n",
+      "  1%|          | 3/300 [00:05<05:50,  1.18s/it, train=0.304, test=0.32] \u001b[A\n",
+      "  1%|▏         | 4/300 [00:05<04:19,  1.14it/s, train=0.304, test=0.32]\u001b[A\n",
+      "  1%|▏         | 4/300 [00:05<04:19,  1.14it/s, train=0.299, test=0.312]\u001b[A\n",
+      "  2%|▏         | 5/300 [00:05<03:30,  1.40it/s, train=0.299, test=0.312]\u001b[A\n",
+      "  2%|▏         | 5/300 [00:05<03:30,  1.40it/s, train=0.291, test=0.301]\u001b[A\n",
+      "  2%|▏         | 6/300 [00:05<03:00,  1.63it/s, train=0.291, test=0.301]\u001b[A\n",
+      "  2%|▏         | 6/300 [00:06<03:00,  1.63it/s, train=0.279, test=0.29] \u001b[A\n",
+      "  2%|▏         | 7/300 [00:06<02:44,  1.79it/s, train=0.279, test=0.29]\u001b[A\n",
+      "  2%|▏         | 7/300 [00:06<02:44,  1.79it/s, train=0.265, test=0.269]\u001b[A\n",
+      "  3%|▎         | 8/300 [00:06<02:35,  1.88it/s, train=0.265, test=0.269]\u001b[A\n",
+      "  3%|▎         | 8/300 [00:07<02:35,  1.88it/s, train=0.247, test=0.259]\u001b[A\n",
+      "  3%|▎         | 9/300 [00:07<02:25,  2.01it/s, train=0.247, test=0.259]\u001b[A\n",
+      "  3%|▎         | 9/300 [00:07<02:25,  2.01it/s, train=0.239, test=0.236]\u001b[A\n",
+      "  3%|▎         | 10/300 [00:07<02:17,  2.11it/s, train=0.239, test=0.236]\u001b[A\n",
+      "  3%|▎         | 10/300 [00:08<02:17,  2.11it/s, train=0.222, test=0.222]\u001b[A\n",
+      "  4%|▎         | 11/300 [00:08<02:12,  2.19it/s, train=0.222, test=0.222]\u001b[A\n",
+      "  4%|▎         | 11/300 [00:08<02:12,  2.19it/s, train=0.201, test=0.207]\u001b[A\n",
+      "  4%|▍         | 12/300 [00:08<02:09,  2.23it/s, train=0.201, test=0.207]\u001b[A\n",
+      "  4%|▍         | 12/300 [00:08<02:09,  2.23it/s, train=0.185, test=0.187]\u001b[A\n",
+      "  4%|▍         | 13/300 [00:08<02:11,  2.18it/s, train=0.185, test=0.187]\u001b[A\n",
+      "  4%|▍         | 13/300 [00:09<02:11,  2.18it/s, train=0.173, test=0.18] \u001b[A\n",
+      "  5%|▍         | 14/300 [00:09<02:10,  2.19it/s, train=0.173, test=0.18]\u001b[A\n",
+      "  5%|▍         | 14/300 [00:09<02:10,  2.19it/s, train=0.159, test=0.164]\u001b[A\n",
+      "  5%|▌         | 15/300 [00:09<02:07,  2.23it/s, train=0.159, test=0.164]\u001b[A\n",
+      "  5%|▌         | 15/300 [00:10<02:07,  2.23it/s, train=0.151, test=0.159]\u001b[A\n",
+      "  5%|▌         | 16/300 [00:10<02:05,  2.27it/s, train=0.151, test=0.159]\u001b[A\n",
+      "  5%|▌         | 16/300 [00:10<02:05,  2.27it/s, train=0.145, test=0.155]\u001b[A\n",
+      "  6%|▌         | 17/300 [00:10<02:04,  2.28it/s, train=0.145, test=0.155]\u001b[A\n",
+      "  6%|▌         | 17/300 [00:11<02:04,  2.28it/s, train=0.142, test=0.153]\u001b[A\n",
+      "  6%|▌         | 18/300 [00:11<02:02,  2.29it/s, train=0.142, test=0.153]\u001b[A\n",
+      "  6%|▌         | 18/300 [00:11<02:02,  2.29it/s, train=0.139, test=0.148]\u001b[A\n",
+      "  6%|▋         | 19/300 [00:11<02:02,  2.28it/s, train=0.139, test=0.148]\u001b[A\n",
+      "  6%|▋         | 19/300 [00:12<02:02,  2.28it/s, train=0.135, test=0.146]\u001b[A\n",
+      "  7%|▋         | 20/300 [00:12<02:04,  2.25it/s, train=0.135, test=0.146]\u001b[A\n",
+      "  7%|▋         | 20/300 [00:12<02:04,  2.25it/s, train=0.136, test=0.147]\u001b[A\n",
+      "  7%|▋         | 21/300 [00:12<02:05,  2.22it/s, train=0.136, test=0.147]\u001b[A\n",
+      "  7%|▋         | 21/300 [00:12<02:05,  2.22it/s, train=0.135, test=0.143]\u001b[A\n",
+      "  7%|▋         | 22/300 [00:12<02:03,  2.26it/s, train=0.135, test=0.143]\u001b[A\n",
+      "  7%|▋         | 22/300 [00:13<02:03,  2.26it/s, train=0.132, test=0.141]\u001b[A\n",
+      "  8%|▊         | 23/300 [00:13<02:01,  2.29it/s, train=0.132, test=0.141]\u001b[A\n",
+      "  8%|▊         | 23/300 [00:13<02:01,  2.29it/s, train=0.13, test=0.143] \u001b[A\n",
+      "  8%|▊         | 24/300 [00:13<02:00,  2.29it/s, train=0.13, test=0.143]\u001b[A\n",
+      "  8%|▊         | 24/300 [00:14<02:00,  2.29it/s, train=0.128, test=0.138]\u001b[A\n",
+      "  8%|▊         | 25/300 [00:14<01:58,  2.32it/s, train=0.128, test=0.138]\u001b[A\n",
+      "  8%|▊         | 25/300 [00:14<01:58,  2.32it/s, train=0.126, test=0.137]\u001b[A\n",
+      "  9%|▊         | 26/300 [00:14<01:58,  2.30it/s, train=0.126, test=0.137]\u001b[A\n",
+      "  9%|▊         | 26/300 [00:15<01:58,  2.30it/s, train=0.124, test=0.138]\u001b[A\n",
+      "  9%|▉         | 27/300 [00:15<01:56,  2.35it/s, train=0.124, test=0.138]\u001b[A\n",
+      "  9%|▉         | 27/300 [00:15<01:56,  2.35it/s, train=0.124, test=0.141]\u001b[A\n",
+      "  9%|▉         | 28/300 [00:15<01:57,  2.31it/s, train=0.124, test=0.141]\u001b[A\n",
+      "  9%|▉         | 28/300 [00:15<01:57,  2.31it/s, train=0.124, test=0.134]\u001b[A\n",
+      " 10%|▉         | 29/300 [00:15<01:55,  2.35it/s, train=0.124, test=0.134]\u001b[A\n",
+      " 10%|▉         | 29/300 [00:16<01:55,  2.35it/s, train=0.12, test=0.133] \u001b[A\n",
+      " 10%|█         | 30/300 [00:16<01:54,  2.37it/s, train=0.12, test=0.133]\u001b[A\n",
+      " 10%|█         | 30/300 [00:16<01:54,  2.37it/s, train=0.12, test=0.132]\u001b[A\n",
+      " 10%|█         | 31/300 [00:16<01:52,  2.39it/s, train=0.12, test=0.132]\u001b[A\n",
+      " 10%|█         | 31/300 [00:17<01:52,  2.39it/s, train=0.118, test=0.131]\u001b[A\n",
+      " 11%|█         | 32/300 [00:17<01:51,  2.41it/s, train=0.118, test=0.131]\u001b[A\n",
+      " 11%|█         | 32/300 [00:17<01:51,  2.41it/s, train=0.116, test=0.131]\u001b[A\n",
+      " 11%|█         | 33/300 [00:17<01:50,  2.41it/s, train=0.116, test=0.131]\u001b[A\n",
+      " 11%|█         | 33/300 [00:18<01:50,  2.41it/s, train=0.116, test=0.131]\u001b[A\n",
+      " 11%|█▏        | 34/300 [00:18<01:56,  2.28it/s, train=0.116, test=0.131]\u001b[A\n",
+      " 11%|█▏        | 34/300 [00:18<01:56,  2.28it/s, train=0.114, test=0.128]\u001b[A\n",
+      " 12%|█▏        | 35/300 [00:18<01:58,  2.23it/s, train=0.114, test=0.128]\u001b[A\n",
+      " 12%|█▏        | 35/300 [00:18<01:58,  2.23it/s, train=0.112, test=0.131]\u001b[A\n",
+      " 12%|█▏        | 36/300 [00:18<01:55,  2.28it/s, train=0.112, test=0.131]\u001b[A\n",
+      " 12%|█▏        | 36/300 [00:19<01:55,  2.28it/s, train=0.113, test=0.127]\u001b[A\n",
+      " 12%|█▏        | 37/300 [00:19<01:55,  2.27it/s, train=0.113, test=0.127]\u001b[A\n",
+      " 12%|█▏        | 37/300 [00:19<01:55,  2.27it/s, train=0.111, test=0.128]\u001b[A\n",
+      " 13%|█▎        | 38/300 [00:19<01:53,  2.31it/s, train=0.111, test=0.128]\u001b[A\n",
+      " 13%|█▎        | 38/300 [00:20<01:53,  2.31it/s, train=0.11, test=0.127] \u001b[A\n",
+      " 13%|█▎        | 39/300 [00:20<01:52,  2.32it/s, train=0.11, test=0.127]\u001b[A\n",
+      " 13%|█▎        | 39/300 [00:20<01:52,  2.32it/s, train=0.108, test=0.124]\u001b[A\n",
+      " 13%|█▎        | 40/300 [00:20<01:53,  2.29it/s, train=0.108, test=0.124]\u001b[A\n",
+      " 13%|█▎        | 40/300 [00:21<01:53,  2.29it/s, train=0.107, test=0.124]\u001b[A\n",
+      " 14%|█▎        | 41/300 [00:21<01:53,  2.29it/s, train=0.107, test=0.124]\u001b[A\n",
+      " 14%|█▎        | 41/300 [00:21<01:53,  2.29it/s, train=0.106, test=0.124]\u001b[A\n",
+      " 14%|█▍        | 42/300 [00:21<01:52,  2.29it/s, train=0.106, test=0.124]\u001b[A\n",
+      " 14%|█▍        | 42/300 [00:22<01:52,  2.29it/s, train=0.105, test=0.123]\u001b[A\n",
+      " 14%|█▍        | 43/300 [00:22<01:52,  2.28it/s, train=0.105, test=0.123]\u001b[A\n",
+      " 14%|█▍        | 43/300 [00:22<01:52,  2.28it/s, train=0.104, test=0.124]\u001b[A\n",
+      " 15%|█▍        | 44/300 [00:22<01:53,  2.25it/s, train=0.104, test=0.124]\u001b[A\n",
+      " 15%|█▍        | 44/300 [00:22<01:53,  2.25it/s, train=0.105, test=0.125]\u001b[A\n",
+      " 15%|█▌        | 45/300 [00:22<01:52,  2.26it/s, train=0.105, test=0.125]\u001b[A\n",
+      " 15%|█▌        | 45/300 [00:23<01:52,  2.26it/s, train=0.103, test=0.119]\u001b[A\n",
+      " 15%|█▌        | 46/300 [00:23<01:52,  2.25it/s, train=0.103, test=0.119]\u001b[A\n",
+      " 15%|█▌        | 46/300 [00:23<01:52,  2.25it/s, train=0.102, test=0.119]\u001b[A\n",
+      " 16%|█▌        | 47/300 [00:23<01:55,  2.19it/s, train=0.102, test=0.119]\u001b[A\n",
+      " 16%|█▌        | 47/300 [00:24<01:55,  2.19it/s, train=0.0999, test=0.121]\u001b[A\n",
+      " 16%|█▌        | 48/300 [00:24<01:55,  2.19it/s, train=0.0999, test=0.121]\u001b[A\n",
+      " 16%|█▌        | 48/300 [00:24<01:55,  2.19it/s, train=0.0993, test=0.119]\u001b[A\n",
+      " 16%|█▋        | 49/300 [00:24<01:53,  2.20it/s, train=0.0993, test=0.119]\u001b[A\n",
+      " 16%|█▋        | 49/300 [00:25<01:53,  2.20it/s, train=0.0977, test=0.119]\u001b[A\n",
+      " 17%|█▋        | 50/300 [00:25<01:52,  2.21it/s, train=0.0977, test=0.119]\u001b[A\n",
+      " 17%|█▋        | 50/300 [00:25<01:52,  2.21it/s, train=0.0986, test=0.121]\u001b[A\n",
+      " 17%|█▋        | 51/300 [00:25<01:55,  2.16it/s, train=0.0986, test=0.121]\u001b[A\n",
+      " 17%|█▋        | 51/300 [00:26<01:55,  2.16it/s, train=0.0966, test=0.117]\u001b[A\n",
+      " 17%|█▋        | 52/300 [00:26<01:58,  2.10it/s, train=0.0966, test=0.117]\u001b[A\n",
+      " 17%|█▋        | 52/300 [00:26<01:58,  2.10it/s, train=0.0957, test=0.118]\u001b[A\n",
+      " 18%|█▊        | 53/300 [00:26<01:57,  2.11it/s, train=0.0957, test=0.118]\u001b[A\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      " 18%|█▊        | 53/300 [00:27<01:57,  2.11it/s, train=0.0958, test=0.116]\u001b[A\n",
+      " 18%|█▊        | 54/300 [00:27<01:53,  2.16it/s, train=0.0958, test=0.116]\u001b[A\n",
+      " 18%|█▊        | 54/300 [00:27<01:53,  2.16it/s, train=0.0933, test=0.112]\u001b[A\n",
+      " 18%|█▊        | 55/300 [00:27<01:58,  2.06it/s, train=0.0933, test=0.112]\u001b[A\n",
+      " 18%|█▊        | 55/300 [00:28<01:58,  2.06it/s, train=0.0924, test=0.114]\u001b[A\n",
+      " 19%|█▊        | 56/300 [00:28<01:58,  2.06it/s, train=0.0924, test=0.114]\u001b[A\n",
+      " 19%|█▊        | 56/300 [00:28<01:58,  2.06it/s, train=0.0931, test=0.112]\u001b[A\n",
+      " 19%|█▉        | 57/300 [00:28<01:56,  2.08it/s, train=0.0931, test=0.112]\u001b[A\n",
+      " 19%|█▉        | 57/300 [00:29<01:56,  2.08it/s, train=0.0923, test=0.109]\u001b[A\n",
+      " 19%|█▉        | 58/300 [00:29<01:53,  2.13it/s, train=0.0923, test=0.109]\u001b[A\n",
+      " 19%|█▉        | 58/300 [00:29<01:53,  2.13it/s, train=0.0928, test=0.116]\u001b[A\n",
+      " 20%|█▉        | 59/300 [00:29<01:50,  2.19it/s, train=0.0928, test=0.116]\u001b[A\n",
+      " 20%|█▉        | 59/300 [00:29<01:50,  2.19it/s, train=0.0921, test=0.11] \u001b[A\n",
+      " 20%|██        | 60/300 [00:29<01:48,  2.21it/s, train=0.0921, test=0.11]\u001b[A\n",
+      " 20%|██        | 60/300 [00:30<01:48,  2.21it/s, train=0.0905, test=0.11]\u001b[A\n",
+      " 20%|██        | 61/300 [00:30<01:47,  2.22it/s, train=0.0905, test=0.11]\u001b[A\n",
+      " 20%|██        | 61/300 [00:30<01:47,  2.22it/s, train=0.09, test=0.113] \u001b[A\n",
+      " 21%|██        | 62/300 [00:30<01:47,  2.22it/s, train=0.09, test=0.113]\u001b[A\n",
+      " 21%|██        | 62/300 [00:31<01:47,  2.22it/s, train=0.0905, test=0.11]\u001b[A\n",
+      " 21%|██        | 63/300 [00:31<01:46,  2.23it/s, train=0.0905, test=0.11]\u001b[A\n",
+      " 21%|██        | 63/300 [00:31<01:46,  2.23it/s, train=0.0877, test=0.108]\u001b[A\n",
+      " 21%|██▏       | 64/300 [00:31<01:44,  2.26it/s, train=0.0877, test=0.108]\u001b[A\n",
+      " 21%|██▏       | 64/300 [00:32<01:44,  2.26it/s, train=0.086, test=0.108] \u001b[A\n",
+      " 22%|██▏       | 65/300 [00:32<01:41,  2.31it/s, train=0.086, test=0.108]\u001b[A\n",
+      " 22%|██▏       | 65/300 [00:32<01:41,  2.31it/s, train=0.0861, test=0.11]\u001b[A\n",
+      " 22%|██▏       | 66/300 [00:32<01:41,  2.31it/s, train=0.0861, test=0.11]\u001b[A\n",
+      " 22%|██▏       | 66/300 [00:32<01:41,  2.31it/s, train=0.0865, test=0.106]\u001b[A\n",
+      " 22%|██▏       | 67/300 [00:32<01:42,  2.28it/s, train=0.0865, test=0.106]\u001b[A\n",
+      " 22%|██▏       | 67/300 [00:33<01:42,  2.28it/s, train=0.0844, test=0.108]\u001b[A\n",
+      " 23%|██▎       | 68/300 [00:33<01:44,  2.23it/s, train=0.0844, test=0.108]\u001b[A\n",
+      " 23%|██▎       | 68/300 [00:33<01:44,  2.23it/s, train=0.0839, test=0.11] \u001b[A\n",
+      " 23%|██▎       | 69/300 [00:33<01:43,  2.23it/s, train=0.0839, test=0.11]\u001b[A\n",
+      " 23%|██▎       | 69/300 [00:34<01:43,  2.23it/s, train=0.0972, test=0.126]\u001b[A\n",
+      " 23%|██▎       | 70/300 [00:34<01:42,  2.25it/s, train=0.0972, test=0.126]\u001b[A\n",
+      " 23%|██▎       | 70/300 [00:34<01:42,  2.25it/s, train=0.0953, test=0.109]\u001b[A\n",
+      " 24%|██▎       | 71/300 [00:34<01:42,  2.24it/s, train=0.0953, test=0.109]\u001b[A\n",
+      " 24%|██▎       | 71/300 [00:35<01:42,  2.24it/s, train=0.0872, test=0.107]\u001b[A\n",
+      " 24%|██▍       | 72/300 [00:35<01:42,  2.21it/s, train=0.0872, test=0.107]\u001b[A\n",
+      " 24%|██▍       | 72/300 [00:35<01:42,  2.21it/s, train=0.0863, test=0.103]\u001b[A\n",
+      " 24%|██▍       | 73/300 [00:35<01:40,  2.27it/s, train=0.0863, test=0.103]\u001b[A\n",
+      " 24%|██▍       | 73/300 [00:36<01:40,  2.27it/s, train=0.0855, test=0.104]\u001b[A\n",
+      " 25%|██▍       | 74/300 [00:36<01:37,  2.31it/s, train=0.0855, test=0.104]\u001b[A\n",
+      " 25%|██▍       | 74/300 [00:36<01:37,  2.31it/s, train=0.082, test=0.104] \u001b[A\n",
+      " 25%|██▌       | 75/300 [00:36<01:36,  2.33it/s, train=0.082, test=0.104]\u001b[A\n",
+      " 25%|██▌       | 75/300 [00:36<01:36,  2.33it/s, train=0.0808, test=0.102]\u001b[A\n",
+      " 25%|██▌       | 76/300 [00:36<01:35,  2.34it/s, train=0.0808, test=0.102]\u001b[A\n",
+      " 25%|██▌       | 76/300 [00:37<01:35,  2.34it/s, train=0.0791, test=0.105]\u001b[A\n",
+      " 26%|██▌       | 77/300 [00:37<01:35,  2.34it/s, train=0.0791, test=0.105]\u001b[A\n",
+      " 26%|██▌       | 77/300 [00:37<01:35,  2.34it/s, train=0.0793, test=0.102]\u001b[A\n",
+      " 26%|██▌       | 78/300 [00:37<01:35,  2.33it/s, train=0.0793, test=0.102]\u001b[A\n",
+      " 26%|██▌       | 78/300 [00:38<01:35,  2.33it/s, train=0.0793, test=0.102]\u001b[A\n",
+      " 26%|██▋       | 79/300 [00:38<01:35,  2.31it/s, train=0.0793, test=0.102]\u001b[A\n",
+      " 26%|██▋       | 79/300 [00:38<01:35,  2.31it/s, train=0.0785, test=0.102]\u001b[A\n",
+      " 27%|██▋       | 80/300 [00:38<01:34,  2.33it/s, train=0.0785, test=0.102]\u001b[A\n",
+      " 27%|██▋       | 80/300 [00:39<01:34,  2.33it/s, train=0.0792, test=0.106]\u001b[A\n",
+      " 27%|██▋       | 81/300 [00:39<01:34,  2.32it/s, train=0.0792, test=0.106]\u001b[A\n",
+      " 27%|██▋       | 81/300 [00:39<01:34,  2.32it/s, train=0.0805, test=0.107]\u001b[A\n",
+      " 27%|██▋       | 82/300 [00:39<01:33,  2.32it/s, train=0.0805, test=0.107]\u001b[A\n",
+      " 27%|██▋       | 82/300 [00:39<01:33,  2.32it/s, train=0.0779, test=0.103]\u001b[A\n",
+      " 28%|██▊       | 83/300 [00:39<01:34,  2.31it/s, train=0.0779, test=0.103]\u001b[A\n",
+      " 28%|██▊       | 83/300 [00:40<01:34,  2.31it/s, train=0.0765, test=0.101]\u001b[A\n",
+      " 28%|██▊       | 84/300 [00:40<01:33,  2.30it/s, train=0.0765, test=0.101]\u001b[A\n",
+      " 28%|██▊       | 84/300 [00:40<01:33,  2.30it/s, train=0.0741, test=0.102]\u001b[A\n",
+      " 28%|██▊       | 85/300 [00:40<01:33,  2.31it/s, train=0.0741, test=0.102]\u001b[A\n",
+      " 28%|██▊       | 85/300 [00:41<01:33,  2.31it/s, train=0.0761, test=0.102]\u001b[A\n",
+      " 29%|██▊       | 86/300 [00:41<01:32,  2.30it/s, train=0.0761, test=0.102]\u001b[A\n",
+      " 29%|██▊       | 86/300 [00:41<01:32,  2.30it/s, train=0.0778, test=0.1]  \u001b[A\n",
+      " 29%|██▉       | 87/300 [00:41<01:31,  2.32it/s, train=0.0778, test=0.1]\u001b[A\n",
+      " 29%|██▉       | 87/300 [00:42<01:31,  2.32it/s, train=0.074, test=0.101]\u001b[A\n",
+      " 29%|██▉       | 88/300 [00:42<01:31,  2.32it/s, train=0.074, test=0.101]\u001b[A\n",
+      " 29%|██▉       | 88/300 [00:42<01:31,  2.32it/s, train=0.0731, test=0.0985]\u001b[A\n",
+      " 30%|██▉       | 89/300 [00:42<01:31,  2.30it/s, train=0.0731, test=0.0985]\u001b[A\n",
+      " 30%|██▉       | 89/300 [00:43<01:31,  2.30it/s, train=0.0716, test=0.0991]\u001b[A\n",
+      " 30%|███       | 90/300 [00:43<01:36,  2.19it/s, train=0.0716, test=0.0991]\u001b[A\n",
+      " 30%|███       | 90/300 [00:43<01:36,  2.19it/s, train=0.0706, test=0.0997]\u001b[A\n",
+      " 30%|███       | 91/300 [00:43<01:33,  2.23it/s, train=0.0706, test=0.0997]\u001b[A\n",
+      " 30%|███       | 91/300 [00:43<01:33,  2.23it/s, train=0.072, test=0.103]  \u001b[A\n",
+      " 31%|███       | 92/300 [00:43<01:31,  2.27it/s, train=0.072, test=0.103]\u001b[A\n",
+      " 31%|███       | 92/300 [00:44<01:31,  2.27it/s, train=0.0735, test=0.1] \u001b[A\n",
+      " 31%|███       | 93/300 [00:44<01:31,  2.27it/s, train=0.0735, test=0.1]\u001b[A\n",
+      " 31%|███       | 93/300 [00:44<01:31,  2.27it/s, train=0.0696, test=0.102]\u001b[A\n",
+      " 31%|███▏      | 94/300 [00:44<01:30,  2.27it/s, train=0.0696, test=0.102]\u001b[A\n",
+      " 31%|███▏      | 94/300 [00:45<01:30,  2.27it/s, train=0.0687, test=0.0982]\u001b[A\n",
+      " 32%|███▏      | 95/300 [00:45<01:31,  2.25it/s, train=0.0687, test=0.0982]\u001b[A\n",
+      " 32%|███▏      | 95/300 [00:45<01:31,  2.25it/s, train=0.0672, test=0.0993]\u001b[A\n",
+      " 32%|███▏      | 96/300 [00:45<01:28,  2.29it/s, train=0.0672, test=0.0993]\u001b[A\n",
+      " 32%|███▏      | 96/300 [00:46<01:28,  2.29it/s, train=0.0701, test=0.0984]\u001b[A\n",
+      " 32%|███▏      | 97/300 [00:46<01:27,  2.31it/s, train=0.0701, test=0.0984]\u001b[A\n",
+      " 32%|███▏      | 97/300 [00:46<01:27,  2.31it/s, train=0.0696, test=0.0935]\u001b[A\n",
+      " 33%|███▎      | 98/300 [00:46<01:30,  2.23it/s, train=0.0696, test=0.0935]\u001b[A\n",
+      " 33%|███▎      | 98/300 [00:46<01:30,  2.23it/s, train=0.0684, test=0.13]  \u001b[A\n",
+      " 33%|███▎      | 99/300 [00:46<01:29,  2.25it/s, train=0.0684, test=0.13]\u001b[A\n",
+      " 33%|███▎      | 99/300 [00:47<01:29,  2.25it/s, train=0.101, test=0.119]\u001b[A\n",
+      " 33%|███▎      | 100/300 [00:47<01:29,  2.24it/s, train=0.101, test=0.119]\u001b[A\n",
+      " 33%|███▎      | 100/300 [00:47<01:29,  2.24it/s, train=0.0822, test=0.104]\u001b[A\n",
+      " 34%|███▎      | 101/300 [00:47<01:27,  2.26it/s, train=0.0822, test=0.104]\u001b[A\n",
+      " 34%|███▎      | 101/300 [00:48<01:27,  2.26it/s, train=0.0772, test=0.0962]\u001b[A\n",
+      " 34%|███▍      | 102/300 [00:48<01:27,  2.27it/s, train=0.0772, test=0.0962]\u001b[A\n",
+      " 34%|███▍      | 102/300 [00:48<01:27,  2.27it/s, train=0.0729, test=0.0979]\u001b[A\n",
+      " 34%|███▍      | 103/300 [00:48<01:25,  2.30it/s, train=0.0729, test=0.0979]\u001b[A\n",
+      " 34%|███▍      | 103/300 [00:49<01:25,  2.30it/s, train=0.0695, test=0.0959]\u001b[A\n",
+      " 35%|███▍      | 104/300 [00:49<01:25,  2.30it/s, train=0.0695, test=0.0959]\u001b[A\n",
+      " 35%|███▍      | 104/300 [00:49<01:25,  2.30it/s, train=0.0679, test=0.0969]\u001b[A\n",
+      " 35%|███▌      | 105/300 [00:49<01:24,  2.31it/s, train=0.0679, test=0.0969]\u001b[A\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      " 35%|███▌      | 105/300 [00:50<01:24,  2.31it/s, train=0.0664, test=0.098] \u001b[A\n",
+      " 35%|███▌      | 106/300 [00:50<01:25,  2.27it/s, train=0.0664, test=0.098]\u001b[A\n",
+      " 35%|███▌      | 106/300 [00:50<01:25,  2.27it/s, train=0.0677, test=0.0977]\u001b[A\n",
+      " 36%|███▌      | 107/300 [00:50<01:28,  2.19it/s, train=0.0677, test=0.0977]\u001b[A\n",
+      " 36%|███▌      | 107/300 [00:51<01:28,  2.19it/s, train=0.065, test=0.096]  \u001b[A\n",
+      " 36%|███▌      | 108/300 [00:51<01:29,  2.14it/s, train=0.065, test=0.096]\u001b[A\n",
+      " 36%|███▌      | 108/300 [00:51<01:29,  2.14it/s, train=0.0639, test=0.0975]\u001b[A\n",
+      " 36%|███▋      | 109/300 [00:51<01:28,  2.15it/s, train=0.0639, test=0.0975]\u001b[A\n",
+      " 36%|███▋      | 109/300 [00:51<01:28,  2.15it/s, train=0.0635, test=0.0965]\u001b[A\n",
+      " 37%|███▋      | 110/300 [00:51<01:29,  2.13it/s, train=0.0635, test=0.0965]\u001b[A\n",
+      " 37%|███▋      | 110/300 [00:52<01:29,  2.13it/s, train=0.0633, test=0.092] \u001b[A\n",
+      " 37%|███▋      | 111/300 [00:52<01:27,  2.16it/s, train=0.0633, test=0.092]\u001b[A\n",
+      " 37%|███▋      | 111/300 [00:52<01:27,  2.16it/s, train=0.0615, test=0.094]\u001b[A\n",
+      " 37%|███▋      | 112/300 [00:52<01:24,  2.23it/s, train=0.0615, test=0.094]\u001b[A\n",
+      " 37%|███▋      | 112/300 [00:53<01:24,  2.23it/s, train=0.0603, test=0.0958]\u001b[A\n",
+      " 38%|███▊      | 113/300 [00:53<01:21,  2.29it/s, train=0.0603, test=0.0958]\u001b[A\n",
+      " 38%|███▊      | 113/300 [00:53<01:21,  2.29it/s, train=0.0619, test=0.1]   \u001b[A\n",
+      " 38%|███▊      | 114/300 [00:53<01:22,  2.24it/s, train=0.0619, test=0.1]\u001b[A\n",
+      " 38%|███▊      | 114/300 [00:54<01:22,  2.24it/s, train=0.0632, test=0.0956]\u001b[A\n",
+      " 38%|███▊      | 115/300 [00:54<01:23,  2.20it/s, train=0.0632, test=0.0956]\u001b[A\n",
+      " 38%|███▊      | 115/300 [00:54<01:23,  2.20it/s, train=0.0605, test=0.0944]\u001b[A\n",
+      " 39%|███▊      | 116/300 [00:54<01:24,  2.17it/s, train=0.0605, test=0.0944]\u001b[A\n",
+      " 39%|███▊      | 116/300 [00:55<01:24,  2.17it/s, train=0.059, test=0.0946] \u001b[A\n",
+      " 39%|███▉      | 117/300 [00:55<01:25,  2.14it/s, train=0.059, test=0.0946]\u001b[A\n",
+      " 39%|███▉      | 117/300 [00:55<01:25,  2.14it/s, train=0.0591, test=0.0934]\u001b[A\n",
+      " 39%|███▉      | 118/300 [00:55<01:25,  2.13it/s, train=0.0591, test=0.0934]\u001b[A\n",
+      " 39%|███▉      | 118/300 [00:56<01:25,  2.13it/s, train=0.0595, test=0.101] \u001b[A\n",
+      " 40%|███▉      | 119/300 [00:56<01:27,  2.07it/s, train=0.0595, test=0.101]\u001b[A\n",
+      " 40%|███▉      | 119/300 [00:56<01:27,  2.07it/s, train=0.058, test=0.0918]\u001b[A\n",
+      " 40%|████      | 120/300 [00:56<01:27,  2.05it/s, train=0.058, test=0.0918]\u001b[A\n",
+      " 40%|████      | 120/300 [00:57<01:27,  2.05it/s, train=0.058, test=0.0908]\u001b[A\n",
+      " 40%|████      | 121/300 [00:57<01:24,  2.13it/s, train=0.058, test=0.0908]\u001b[A\n",
+      " 40%|████      | 121/300 [00:57<01:24,  2.13it/s, train=0.0573, test=0.0917]\u001b[A\n",
+      " 41%|████      | 122/300 [00:57<01:21,  2.20it/s, train=0.0573, test=0.0917]\u001b[A\n",
+      " 41%|████      | 122/300 [00:57<01:21,  2.20it/s, train=0.057, test=0.0935] \u001b[A\n",
+      " 41%|████      | 123/300 [00:57<01:18,  2.25it/s, train=0.057, test=0.0935]\u001b[A\n",
+      " 41%|████      | 123/300 [00:58<01:18,  2.25it/s, train=0.0597, test=0.0916]\u001b[A\n",
+      " 41%|████▏     | 124/300 [00:58<01:16,  2.29it/s, train=0.0597, test=0.0916]\u001b[A\n",
+      " 41%|████▏     | 124/300 [00:58<01:16,  2.29it/s, train=0.0579, test=0.0923]\u001b[A\n",
+      " 42%|████▏     | 125/300 [00:58<01:15,  2.33it/s, train=0.0579, test=0.0923]\u001b[A\n",
+      " 42%|████▏     | 125/300 [00:59<01:15,  2.33it/s, train=0.0548, test=0.0904]\u001b[A\n",
+      " 42%|████▏     | 126/300 [00:59<01:13,  2.36it/s, train=0.0548, test=0.0904]\u001b[A\n",
+      " 42%|████▏     | 126/300 [00:59<01:13,  2.36it/s, train=0.0544, test=0.0952]\u001b[A\n",
+      " 42%|████▏     | 127/300 [00:59<01:12,  2.38it/s, train=0.0544, test=0.0952]\u001b[A\n",
+      " 42%|████▏     | 127/300 [00:59<01:12,  2.38it/s, train=0.056, test=0.0911] \u001b[A\n",
+      " 43%|████▎     | 128/300 [00:59<01:11,  2.39it/s, train=0.056, test=0.0911]\u001b[A\n",
+      " 43%|████▎     | 128/300 [01:00<01:11,  2.39it/s, train=0.0553, test=0.0886]\u001b[A\n",
+      " 43%|████▎     | 129/300 [01:00<01:11,  2.40it/s, train=0.0553, test=0.0886]\u001b[A\n",
+      " 43%|████▎     | 129/300 [01:00<01:11,  2.40it/s, train=0.0552, test=0.0937]\u001b[A\n",
+      " 43%|████▎     | 130/300 [01:00<01:10,  2.40it/s, train=0.0552, test=0.0937]\u001b[A\n",
+      " 43%|████▎     | 130/300 [01:01<01:10,  2.40it/s, train=0.0564, test=0.0913]\u001b[A\n",
+      " 44%|████▎     | 131/300 [01:01<01:12,  2.33it/s, train=0.0564, test=0.0913]\u001b[A\n",
+      " 44%|████▎     | 131/300 [01:01<01:12,  2.33it/s, train=0.0524, test=0.0975]\u001b[A\n",
+      " 44%|████▍     | 132/300 [01:01<01:13,  2.28it/s, train=0.0524, test=0.0975]\u001b[A\n",
+      " 44%|████▍     | 132/300 [01:02<01:13,  2.28it/s, train=0.0551, test=0.0933]\u001b[A\n",
+      " 44%|████▍     | 133/300 [01:02<01:12,  2.30it/s, train=0.0551, test=0.0933]\u001b[A\n",
+      " 44%|████▍     | 133/300 [01:02<01:12,  2.30it/s, train=0.0549, test=0.0956]\u001b[A\n",
+      " 45%|████▍     | 134/300 [01:02<01:11,  2.34it/s, train=0.0549, test=0.0956]\u001b[A\n",
+      " 45%|████▍     | 134/300 [01:02<01:11,  2.34it/s, train=0.0529, test=0.0908]\u001b[A\n",
+      " 45%|████▌     | 135/300 [01:02<01:09,  2.37it/s, train=0.0529, test=0.0908]\u001b[A\n",
+      " 45%|████▌     | 135/300 [01:03<01:09,  2.37it/s, train=0.0508, test=0.0964]\u001b[A\n",
+      " 45%|████▌     | 136/300 [01:03<01:08,  2.38it/s, train=0.0508, test=0.0964]\u001b[A\n",
+      " 45%|████▌     | 136/300 [01:03<01:08,  2.38it/s, train=0.0515, test=0.0943]\u001b[A\n",
+      " 46%|████▌     | 137/300 [01:03<01:10,  2.32it/s, train=0.0515, test=0.0943]\u001b[A\n",
+      " 46%|████▌     | 137/300 [01:04<01:10,  2.32it/s, train=0.0528, test=0.0994]\u001b[A\n",
+      " 46%|████▌     | 138/300 [01:04<01:11,  2.26it/s, train=0.0528, test=0.0994]\u001b[A\n",
+      " 46%|████▌     | 138/300 [01:04<01:11,  2.26it/s, train=0.0509, test=0.0997]\u001b[A\n",
+      " 46%|████▋     | 139/300 [01:04<01:13,  2.20it/s, train=0.0509, test=0.0997]\u001b[A\n",
+      " 46%|████▋     | 139/300 [01:05<01:13,  2.20it/s, train=0.0508, test=0.091] \u001b[A\n",
+      " 47%|████▋     | 140/300 [01:05<01:13,  2.16it/s, train=0.0508, test=0.091]\u001b[A\n",
+      " 47%|████▋     | 140/300 [01:05<01:13,  2.16it/s, train=0.0496, test=0.0936]\u001b[A\n",
+      " 47%|████▋     | 141/300 [01:05<01:14,  2.12it/s, train=0.0496, test=0.0936]\u001b[A\n",
+      " 47%|████▋     | 141/300 [01:06<01:14,  2.12it/s, train=0.0505, test=0.0928]\u001b[A\n",
+      " 47%|████▋     | 142/300 [01:06<01:12,  2.17it/s, train=0.0505, test=0.0928]\u001b[A\n",
+      " 47%|████▋     | 142/300 [01:06<01:12,  2.17it/s, train=0.0501, test=0.0915]\u001b[A\n",
+      " 48%|████▊     | 143/300 [01:06<01:12,  2.17it/s, train=0.0501, test=0.0915]\u001b[A\n",
+      " 48%|████▊     | 143/300 [01:07<01:12,  2.17it/s, train=0.0487, test=0.0897]\u001b[A\n",
+      " 48%|████▊     | 144/300 [01:07<01:10,  2.20it/s, train=0.0487, test=0.0897]\u001b[A\n",
+      " 48%|████▊     | 144/300 [01:07<01:10,  2.20it/s, train=0.048, test=0.0969] \u001b[A\n",
+      " 48%|████▊     | 145/300 [01:07<01:09,  2.23it/s, train=0.048, test=0.0969]\u001b[A\n",
+      " 48%|████▊     | 145/300 [01:07<01:09,  2.23it/s, train=0.0485, test=0.093]\u001b[A\n",
+      " 49%|████▊     | 146/300 [01:07<01:08,  2.26it/s, train=0.0485, test=0.093]\u001b[A\n",
+      " 49%|████▊     | 146/300 [01:08<01:08,  2.26it/s, train=0.0495, test=0.0898]\u001b[A\n",
+      " 49%|████▉     | 147/300 [01:08<01:07,  2.28it/s, train=0.0495, test=0.0898]\u001b[A\n",
+      " 49%|████▉     | 147/300 [01:08<01:07,  2.28it/s, train=0.0498, test=0.0931]\u001b[A\n",
+      " 49%|████▉     | 148/300 [01:08<01:06,  2.30it/s, train=0.0498, test=0.0931]\u001b[A\n",
+      " 49%|████▉     | 148/300 [01:09<01:06,  2.30it/s, train=0.0506, test=0.0922]\u001b[A\n",
+      " 50%|████▉     | 149/300 [01:09<01:05,  2.30it/s, train=0.0506, test=0.0922]\u001b[A\n",
+      " 50%|████▉     | 149/300 [01:09<01:05,  2.30it/s, train=0.0482, test=0.0947]\u001b[A\n",
+      " 50%|█████     | 150/300 [01:09<01:05,  2.31it/s, train=0.0482, test=0.0947]\u001b[A\n",
+      " 50%|█████     | 150/300 [01:10<01:05,  2.31it/s, train=0.0463, test=0.0901]\u001b[A\n",
+      " 50%|█████     | 151/300 [01:10<01:05,  2.28it/s, train=0.0463, test=0.0901]\u001b[A\n",
+      " 50%|█████     | 151/300 [01:10<01:05,  2.28it/s, train=0.0475, test=0.0918]\u001b[A\n",
+      " 51%|█████     | 152/300 [01:10<01:04,  2.29it/s, train=0.0475, test=0.0918]\u001b[A\n",
+      " 51%|█████     | 152/300 [01:11<01:04,  2.29it/s, train=0.0457, test=0.0922]\u001b[A\n",
+      " 51%|█████     | 153/300 [01:11<01:04,  2.29it/s, train=0.0457, test=0.0922]\u001b[A\n",
+      " 51%|█████     | 153/300 [01:11<01:04,  2.29it/s, train=0.0475, test=0.0919]\u001b[A\n",
+      " 51%|█████▏    | 154/300 [01:11<01:03,  2.31it/s, train=0.0475, test=0.0919]\u001b[A\n",
+      " 51%|█████▏    | 154/300 [01:11<01:03,  2.31it/s, train=0.0458, test=0.091] \u001b[A\n",
+      " 52%|█████▏    | 155/300 [01:11<01:02,  2.32it/s, train=0.0458, test=0.091]\u001b[A\n",
+      " 52%|█████▏    | 155/300 [01:12<01:02,  2.32it/s, train=0.0441, test=0.0909]\u001b[A\n",
+      " 52%|█████▏    | 156/300 [01:12<01:02,  2.32it/s, train=0.0441, test=0.0909]\u001b[A\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      " 52%|█████▏    | 156/300 [01:12<01:02,  2.32it/s, train=0.0449, test=0.0878]\u001b[A\n",
+      " 52%|█████▏    | 157/300 [01:12<01:02,  2.30it/s, train=0.0449, test=0.0878]\u001b[A\n",
+      " 52%|█████▏    | 157/300 [01:13<01:02,  2.30it/s, train=0.0453, test=0.0933]\u001b[A\n",
+      " 53%|█████▎    | 158/300 [01:13<01:01,  2.31it/s, train=0.0453, test=0.0933]\u001b[A\n",
+      " 53%|█████▎    | 158/300 [01:13<01:01,  2.31it/s, train=0.0452, test=0.0887]\u001b[A\n",
+      " 53%|█████▎    | 159/300 [01:13<01:02,  2.27it/s, train=0.0452, test=0.0887]\u001b[A\n",
+      " 53%|█████▎    | 159/300 [01:14<01:02,  2.27it/s, train=0.0444, test=0.0874]\u001b[A\n",
+      " 53%|█████▎    | 160/300 [01:14<01:01,  2.26it/s, train=0.0444, test=0.0874]\u001b[A\n",
+      " 53%|█████▎    | 160/300 [01:14<01:01,  2.26it/s, train=0.0441, test=0.0913]\u001b[A\n",
+      " 54%|█████▎    | 161/300 [01:14<01:01,  2.27it/s, train=0.0441, test=0.0913]\u001b[A\n",
+      " 54%|█████▎    | 161/300 [01:14<01:01,  2.27it/s, train=0.0453, test=0.0908]\u001b[A\n",
+      " 54%|█████▍    | 162/300 [01:14<01:01,  2.24it/s, train=0.0453, test=0.0908]\u001b[A\n",
+      " 54%|█████▍    | 162/300 [01:15<01:01,  2.24it/s, train=0.0466, test=0.0886]\u001b[A\n",
+      " 54%|█████▍    | 163/300 [01:15<01:01,  2.24it/s, train=0.0466, test=0.0886]\u001b[A\n",
+      " 54%|█████▍    | 163/300 [01:15<01:01,  2.24it/s, train=0.0468, test=0.0982]\u001b[A\n",
+      " 55%|█████▍    | 164/300 [01:15<01:01,  2.22it/s, train=0.0468, test=0.0982]\u001b[A\n",
+      " 55%|█████▍    | 164/300 [01:16<01:01,  2.22it/s, train=0.0478, test=0.0886]\u001b[A\n",
+      " 55%|█████▌    | 165/300 [01:16<01:00,  2.22it/s, train=0.0478, test=0.0886]\u001b[A\n",
+      " 55%|█████▌    | 165/300 [01:16<01:00,  2.22it/s, train=0.0476, test=0.0919]\u001b[A\n",
+      " 55%|█████▌    | 166/300 [01:16<01:00,  2.23it/s, train=0.0476, test=0.0919]\u001b[A\n",
+      " 55%|█████▌    | 166/300 [01:17<01:00,  2.23it/s, train=0.0458, test=0.088] \u001b[A\n",
+      " 56%|█████▌    | 167/300 [01:17<00:58,  2.26it/s, train=0.0458, test=0.088]\u001b[A\n",
+      " 56%|█████▌    | 167/300 [01:17<00:58,  2.26it/s, train=0.0461, test=0.0941]\u001b[A\n",
+      " 56%|█████▌    | 168/300 [01:17<00:58,  2.27it/s, train=0.0461, test=0.0941]\u001b[A\n",
+      " 56%|█████▌    | 168/300 [01:18<00:58,  2.27it/s, train=0.0442, test=0.0933]\u001b[A\n",
+      " 56%|█████▋    | 169/300 [01:18<00:57,  2.28it/s, train=0.0442, test=0.0933]\u001b[A\n",
+      " 56%|█████▋    | 169/300 [01:18<00:57,  2.28it/s, train=0.0432, test=0.0896]\u001b[A\n",
+      " 57%|█████▋    | 170/300 [01:18<00:58,  2.23it/s, train=0.0432, test=0.0896]\u001b[A\n",
+      " 57%|█████▋    | 170/300 [01:18<00:58,  2.23it/s, train=0.0423, test=0.0891]\u001b[A\n",
+      " 57%|█████▋    | 171/300 [01:18<00:57,  2.23it/s, train=0.0423, test=0.0891]\u001b[A\n",
+      " 57%|█████▋    | 171/300 [01:19<00:57,  2.23it/s, train=0.0414, test=0.091] \u001b[A\n",
+      " 57%|█████▋    | 172/300 [01:19<00:57,  2.23it/s, train=0.0414, test=0.091]\u001b[A\n",
+      " 57%|█████▋    | 172/300 [01:19<00:57,  2.23it/s, train=0.0411, test=0.0878]\u001b[A\n",
+      " 58%|█████▊    | 173/300 [01:19<00:56,  2.25it/s, train=0.0411, test=0.0878]\u001b[A\n",
+      " 58%|█████▊    | 173/300 [01:20<00:56,  2.25it/s, train=0.0411, test=0.0872]\u001b[A\n",
+      " 58%|█████▊    | 174/300 [01:20<00:55,  2.25it/s, train=0.0411, test=0.0872]\u001b[A\n",
+      " 58%|█████▊    | 174/300 [01:20<00:55,  2.25it/s, train=0.0408, test=0.0834]\u001b[A\n",
+      " 58%|█████▊    | 175/300 [01:20<00:55,  2.27it/s, train=0.0408, test=0.0834]\u001b[A\n",
+      " 58%|█████▊    | 175/300 [01:21<00:55,  2.27it/s, train=0.0398, test=0.0862]\u001b[A\n",
+      " 59%|█████▊    | 176/300 [01:21<00:54,  2.27it/s, train=0.0398, test=0.0862]\u001b[A\n",
+      " 59%|█████▊    | 176/300 [01:21<00:54,  2.27it/s, train=0.0414, test=0.0819]\u001b[A\n",
+      " 59%|█████▉    | 177/300 [01:21<00:53,  2.30it/s, train=0.0414, test=0.0819]\u001b[A\n",
+      " 59%|█████▉    | 177/300 [01:22<00:53,  2.30it/s, train=0.0418, test=0.087] \u001b[A\n",
+      " 59%|█████▉    | 178/300 [01:22<00:53,  2.27it/s, train=0.0418, test=0.087]\u001b[A\n",
+      " 59%|█████▉    | 178/300 [01:22<00:53,  2.27it/s, train=0.0425, test=0.087]\u001b[A\n",
+      " 60%|█████▉    | 179/300 [01:22<00:53,  2.28it/s, train=0.0425, test=0.087]\u001b[A\n",
+      " 60%|█████▉    | 179/300 [01:22<00:53,  2.28it/s, train=0.0397, test=0.0871]\u001b[A\n",
+      " 60%|██████    | 180/300 [01:22<00:52,  2.28it/s, train=0.0397, test=0.0871]\u001b[A\n",
+      " 60%|██████    | 180/300 [01:23<00:52,  2.28it/s, train=0.039, test=0.0869] \u001b[A\n",
+      " 60%|██████    | 181/300 [01:23<00:51,  2.30it/s, train=0.039, test=0.0869]\u001b[A\n",
+      " 60%|██████    | 181/300 [01:23<00:51,  2.30it/s, train=0.0377, test=0.0875]\u001b[A\n",
+      " 61%|██████    | 182/300 [01:23<00:50,  2.33it/s, train=0.0377, test=0.0875]\u001b[A\n",
+      " 61%|██████    | 182/300 [01:24<00:50,  2.33it/s, train=0.039, test=0.0898] \u001b[A\n",
+      " 61%|██████    | 183/300 [01:24<00:50,  2.33it/s, train=0.039, test=0.0898]\u001b[A\n",
+      " 61%|██████    | 183/300 [01:24<00:50,  2.33it/s, train=0.0404, test=0.0856]\u001b[A\n",
+      " 61%|██████▏   | 184/300 [01:24<00:50,  2.32it/s, train=0.0404, test=0.0856]\u001b[A\n",
+      " 61%|██████▏   | 184/300 [01:25<00:50,  2.32it/s, train=0.0409, test=0.0916]\u001b[A\n",
+      " 62%|██████▏   | 185/300 [01:25<00:52,  2.21it/s, train=0.0409, test=0.0916]\u001b[A\n",
+      " 62%|██████▏   | 185/300 [01:25<00:52,  2.21it/s, train=0.0413, test=0.0903]\u001b[A\n",
+      " 62%|██████▏   | 186/300 [01:25<00:51,  2.20it/s, train=0.0413, test=0.0903]\u001b[A\n",
+      " 62%|██████▏   | 186/300 [01:26<00:51,  2.20it/s, train=0.0403, test=0.0848]\u001b[A\n",
+      " 62%|██████▏   | 187/300 [01:26<00:50,  2.24it/s, train=0.0403, test=0.0848]\u001b[A\n",
+      " 62%|██████▏   | 187/300 [01:26<00:50,  2.24it/s, train=0.0393, test=0.0912]\u001b[A\n",
+      " 63%|██████▎   | 188/300 [01:26<00:50,  2.21it/s, train=0.0393, test=0.0912]\u001b[A\n",
+      " 63%|██████▎   | 188/300 [01:26<00:50,  2.21it/s, train=0.0371, test=0.0828]\u001b[A\n",
+      " 63%|██████▎   | 189/300 [01:26<00:49,  2.24it/s, train=0.0371, test=0.0828]\u001b[A\n",
+      " 63%|██████▎   | 189/300 [01:27<00:49,  2.24it/s, train=0.0368, test=0.0874]\u001b[A\n",
+      " 63%|██████▎   | 190/300 [01:27<00:48,  2.26it/s, train=0.0368, test=0.0874]\u001b[A\n",
+      " 63%|██████▎   | 190/300 [01:27<00:48,  2.26it/s, train=0.0363, test=0.0855]\u001b[A\n",
+      " 64%|██████▎   | 191/300 [01:27<00:48,  2.27it/s, train=0.0363, test=0.0855]\u001b[A\n",
+      " 64%|██████▎   | 191/300 [01:28<00:48,  2.27it/s, train=0.0362, test=0.0864]\u001b[A\n",
+      " 64%|██████▍   | 192/300 [01:28<00:47,  2.28it/s, train=0.0362, test=0.0864]\u001b[A\n",
+      " 64%|██████▍   | 192/300 [01:28<00:47,  2.28it/s, train=0.0357, test=0.0872]\u001b[A\n",
+      " 64%|██████▍   | 193/300 [01:28<00:46,  2.31it/s, train=0.0357, test=0.0872]\u001b[A\n",
+      " 64%|██████▍   | 193/300 [01:29<00:46,  2.31it/s, train=0.0372, test=0.0967]\u001b[A\n",
+      " 65%|██████▍   | 194/300 [01:29<00:45,  2.30it/s, train=0.0372, test=0.0967]\u001b[A\n",
+      " 65%|██████▍   | 194/300 [01:29<00:45,  2.30it/s, train=0.0392, test=0.0902]\u001b[A\n",
+      " 65%|██████▌   | 195/300 [01:29<00:46,  2.26it/s, train=0.0392, test=0.0902]\u001b[A\n",
+      " 65%|██████▌   | 195/300 [01:30<00:46,  2.26it/s, train=0.0355, test=0.0861]\u001b[A\n",
+      " 65%|██████▌   | 196/300 [01:30<00:46,  2.24it/s, train=0.0355, test=0.0861]\u001b[A\n",
+      " 65%|██████▌   | 196/300 [01:30<00:46,  2.24it/s, train=0.0343, test=0.0816]\u001b[A\n",
+      " 66%|██████▌   | 197/300 [01:30<00:46,  2.22it/s, train=0.0343, test=0.0816]\u001b[A\n",
+      " 66%|██████▌   | 197/300 [01:30<00:46,  2.22it/s, train=0.033, test=0.0859] \u001b[A\n",
+      " 66%|██████▌   | 198/300 [01:30<00:48,  2.12it/s, train=0.033, test=0.0859]\u001b[A\n",
+      " 66%|██████▌   | 198/300 [01:31<00:48,  2.12it/s, train=0.0325, test=0.0843]\u001b[A\n",
+      " 66%|██████▋   | 199/300 [01:31<00:49,  2.04it/s, train=0.0325, test=0.0843]\u001b[A\n",
+      " 66%|██████▋   | 199/300 [01:31<00:49,  2.04it/s, train=0.0321, test=0.0834]\u001b[A\n",
+      " 67%|██████▋   | 200/300 [01:31<00:47,  2.09it/s, train=0.0321, test=0.0834]\u001b[A\n",
+      " 67%|██████▋   | 200/300 [01:32<00:47,  2.09it/s, train=0.0332, test=0.0822]\u001b[A\n",
+      " 67%|██████▋   | 201/300 [01:32<00:46,  2.14it/s, train=0.0332, test=0.0822]\u001b[A\n",
+      " 67%|██████▋   | 201/300 [01:32<00:46,  2.14it/s, train=0.0336, test=0.0877]\u001b[A\n",
+      " 67%|██████▋   | 202/300 [01:32<00:44,  2.18it/s, train=0.0336, test=0.0877]\u001b[A\n",
+      " 67%|██████▋   | 202/300 [01:33<00:44,  2.18it/s, train=0.0373, test=0.0883]\u001b[A\n",
+      " 68%|██████▊   | 203/300 [01:33<00:44,  2.20it/s, train=0.0373, test=0.0883]\u001b[A\n",
+      " 68%|██████▊   | 203/300 [01:33<00:44,  2.20it/s, train=0.0362, test=0.0933]\u001b[A\n",
+      " 68%|██████▊   | 204/300 [01:33<00:43,  2.21it/s, train=0.0362, test=0.0933]\u001b[A\n",
+      " 68%|██████▊   | 204/300 [01:34<00:43,  2.21it/s, train=0.0346, test=0.0806]\u001b[A\n",
+      " 68%|██████▊   | 205/300 [01:34<00:42,  2.21it/s, train=0.0346, test=0.0806]\u001b[A\n",
+      " 68%|██████▊   | 205/300 [01:34<00:42,  2.21it/s, train=0.0345, test=0.0835]\u001b[A\n",
+      " 69%|██████▊   | 206/300 [01:34<00:41,  2.24it/s, train=0.0345, test=0.0835]\u001b[A\n",
+      " 69%|██████▊   | 206/300 [01:35<00:41,  2.24it/s, train=0.0333, test=0.0928]\u001b[A\n",
+      " 69%|██████▉   | 207/300 [01:35<00:41,  2.27it/s, train=0.0333, test=0.0928]\u001b[A\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      " 69%|██████▉   | 207/300 [01:35<00:41,  2.27it/s, train=0.0362, test=0.0813]\u001b[A\n",
+      " 69%|██████▉   | 208/300 [01:35<00:40,  2.27it/s, train=0.0362, test=0.0813]\u001b[A\n",
+      " 69%|██████▉   | 208/300 [01:35<00:40,  2.27it/s, train=0.0369, test=0.08]  \u001b[A\n",
+      " 70%|██████▉   | 209/300 [01:35<00:40,  2.26it/s, train=0.0369, test=0.08]\u001b[A\n",
+      " 70%|██████▉   | 209/300 [01:36<00:40,  2.26it/s, train=0.0369, test=0.0851]\u001b[A\n",
+      " 70%|███████   | 210/300 [01:36<00:40,  2.25it/s, train=0.0369, test=0.0851]\u001b[A\n",
+      " 70%|███████   | 210/300 [01:36<00:40,  2.25it/s, train=0.0331, test=0.0825]\u001b[A\n",
+      " 70%|███████   | 211/300 [01:36<00:39,  2.26it/s, train=0.0331, test=0.0825]\u001b[A\n",
+      " 70%|███████   | 211/300 [01:37<00:39,  2.26it/s, train=0.0321, test=0.083] \u001b[A\n",
+      " 71%|███████   | 212/300 [01:37<00:38,  2.27it/s, train=0.0321, test=0.083]\u001b[A\n",
+      " 71%|███████   | 212/300 [01:37<00:38,  2.27it/s, train=0.0309, test=0.0803]\u001b[A\n",
+      " 71%|███████   | 213/300 [01:37<00:38,  2.28it/s, train=0.0309, test=0.0803]\u001b[A\n",
+      " 71%|███████   | 213/300 [01:38<00:38,  2.28it/s, train=0.0301, test=0.0839]\u001b[A\n",
+      " 71%|███████▏  | 214/300 [01:38<00:37,  2.27it/s, train=0.0301, test=0.0839]\u001b[A\n",
+      " 71%|███████▏  | 214/300 [01:38<00:37,  2.27it/s, train=0.0291, test=0.085] \u001b[A\n",
+      " 72%|███████▏  | 215/300 [01:38<00:37,  2.26it/s, train=0.0291, test=0.085]\u001b[A\n",
+      " 72%|███████▏  | 215/300 [01:39<00:37,  2.26it/s, train=0.0285, test=0.0823]\u001b[A\n",
+      " 72%|███████▏  | 216/300 [01:39<00:37,  2.27it/s, train=0.0285, test=0.0823]\u001b[A\n",
+      " 72%|███████▏  | 216/300 [01:39<00:37,  2.27it/s, train=0.0285, test=0.0842]\u001b[A\n",
+      " 72%|███████▏  | 217/300 [01:39<00:36,  2.27it/s, train=0.0285, test=0.0842]\u001b[A\n",
+      " 72%|███████▏  | 217/300 [01:39<00:36,  2.27it/s, train=0.0283, test=0.0828]\u001b[A\n",
+      " 73%|███████▎  | 218/300 [01:39<00:36,  2.27it/s, train=0.0283, test=0.0828]\u001b[A\n",
+      " 73%|███████▎  | 218/300 [01:40<00:36,  2.27it/s, train=0.0287, test=0.084] \u001b[A\n",
+      " 73%|███████▎  | 219/300 [01:40<00:36,  2.23it/s, train=0.0287, test=0.084]\u001b[A\n",
+      " 73%|███████▎  | 219/300 [01:40<00:36,  2.23it/s, train=0.0295, test=0.0821]\u001b[A\n",
+      " 73%|███████▎  | 220/300 [01:40<00:35,  2.24it/s, train=0.0295, test=0.0821]\u001b[A\n",
+      " 73%|███████▎  | 220/300 [01:41<00:35,  2.24it/s, train=0.0278, test=0.0824]\u001b[A\n",
+      " 74%|███████▎  | 221/300 [01:41<00:35,  2.24it/s, train=0.0278, test=0.0824]\u001b[A\n",
+      " 74%|███████▎  | 221/300 [01:41<00:35,  2.24it/s, train=0.0278, test=0.0802]\u001b[A\n",
+      " 74%|███████▍  | 222/300 [01:41<00:34,  2.26it/s, train=0.0278, test=0.0802]\u001b[A\n",
+      " 74%|███████▍  | 222/300 [01:42<00:34,  2.26it/s, train=0.0285, test=0.0855]\u001b[A\n",
+      " 74%|███████▍  | 223/300 [01:42<00:34,  2.26it/s, train=0.0285, test=0.0855]\u001b[A\n",
+      " 74%|███████▍  | 223/300 [01:42<00:34,  2.26it/s, train=0.0286, test=0.0812]\u001b[A\n",
+      " 75%|███████▍  | 224/300 [01:42<00:34,  2.23it/s, train=0.0286, test=0.0812]\u001b[A\n",
+      " 75%|███████▍  | 224/300 [01:43<00:34,  2.23it/s, train=0.0276, test=0.0751]\u001b[A\n",
+      " 75%|███████▌  | 225/300 [01:43<00:33,  2.24it/s, train=0.0276, test=0.0751]\u001b[A\n",
+      " 75%|███████▌  | 225/300 [01:43<00:33,  2.24it/s, train=0.0268, test=0.0829]\u001b[A\n",
+      " 75%|███████▌  | 226/300 [01:43<00:32,  2.24it/s, train=0.0268, test=0.0829]\u001b[A\n",
+      " 75%|███████▌  | 226/300 [01:43<00:32,  2.24it/s, train=0.0275, test=0.0846]\u001b[A\n",
+      " 76%|███████▌  | 227/300 [01:43<00:32,  2.26it/s, train=0.0275, test=0.0846]\u001b[A\n",
+      " 76%|███████▌  | 227/300 [01:44<00:32,  2.26it/s, train=0.0283, test=0.0888]\u001b[A\n",
+      " 76%|███████▌  | 228/300 [01:44<00:32,  2.25it/s, train=0.0283, test=0.0888]\u001b[A\n",
+      " 76%|███████▌  | 228/300 [01:44<00:32,  2.25it/s, train=0.028, test=0.0745] \u001b[A\n",
+      " 76%|███████▋  | 229/300 [01:44<00:31,  2.22it/s, train=0.028, test=0.0745]\u001b[A\n",
+      " 76%|███████▋  | 229/300 [01:45<00:31,  2.22it/s, train=0.0268, test=0.0845]\u001b[A\n",
+      " 77%|███████▋  | 230/300 [01:45<00:31,  2.22it/s, train=0.0268, test=0.0845]\u001b[A\n",
+      " 77%|███████▋  | 230/300 [01:45<00:31,  2.22it/s, train=0.0259, test=0.0741]\u001b[A\n",
+      " 77%|███████▋  | 231/300 [01:45<00:31,  2.22it/s, train=0.0259, test=0.0741]\u001b[A\n",
+      " 77%|███████▋  | 231/300 [01:46<00:31,  2.22it/s, train=0.0275, test=0.0838]\u001b[A\n",
+      " 77%|███████▋  | 232/300 [01:46<00:30,  2.23it/s, train=0.0275, test=0.0838]\u001b[A\n",
+      " 77%|███████▋  | 232/300 [01:46<00:30,  2.23it/s, train=0.0297, test=0.0793]\u001b[A\n",
+      " 78%|███████▊  | 233/300 [01:46<00:29,  2.24it/s, train=0.0297, test=0.0793]\u001b[A\n",
+      " 78%|███████▊  | 233/300 [01:47<00:29,  2.24it/s, train=0.0296, test=0.0839]\u001b[A\n",
+      " 78%|███████▊  | 234/300 [01:47<00:29,  2.22it/s, train=0.0296, test=0.0839]\u001b[A\n",
+      " 78%|███████▊  | 234/300 [01:47<00:29,  2.22it/s, train=0.0263, test=0.0764]\u001b[A\n",
+      " 78%|███████▊  | 235/300 [01:47<00:29,  2.19it/s, train=0.0263, test=0.0764]\u001b[A\n",
+      " 78%|███████▊  | 235/300 [01:47<00:29,  2.19it/s, train=0.0273, test=0.0797]\u001b[A\n",
+      " 79%|███████▊  | 236/300 [01:47<00:28,  2.22it/s, train=0.0273, test=0.0797]\u001b[A\n",
+      " 79%|███████▊  | 236/300 [01:48<00:28,  2.22it/s, train=0.0263, test=0.0793]\u001b[A\n",
+      " 79%|███████▉  | 237/300 [01:48<00:28,  2.18it/s, train=0.0263, test=0.0793]\u001b[A\n",
+      " 79%|███████▉  | 237/300 [01:48<00:28,  2.18it/s, train=0.0272, test=0.0786]\u001b[A\n",
+      " 79%|███████▉  | 238/300 [01:48<00:28,  2.18it/s, train=0.0272, test=0.0786]\u001b[A\n",
+      " 79%|███████▉  | 238/300 [01:49<00:28,  2.18it/s, train=0.0274, test=0.0805]\u001b[A\n",
+      " 80%|███████▉  | 239/300 [01:49<00:29,  2.10it/s, train=0.0274, test=0.0805]\u001b[A\n",
+      " 80%|███████▉  | 239/300 [01:49<00:29,  2.10it/s, train=0.0274, test=0.0839]\u001b[A\n",
+      " 80%|████████  | 240/300 [01:49<00:28,  2.11it/s, train=0.0274, test=0.0839]\u001b[A\n",
+      " 80%|████████  | 240/300 [01:50<00:28,  2.11it/s, train=0.0276, test=0.0805]\u001b[A\n",
+      " 80%|████████  | 241/300 [01:50<00:28,  2.07it/s, train=0.0276, test=0.0805]\u001b[A\n",
+      " 80%|████████  | 241/300 [01:50<00:28,  2.07it/s, train=0.0264, test=0.0827]\u001b[A\n",
+      " 81%|████████  | 242/300 [01:50<00:27,  2.10it/s, train=0.0264, test=0.0827]\u001b[A\n",
+      " 81%|████████  | 242/300 [01:51<00:27,  2.10it/s, train=0.0258, test=0.0795]\u001b[A\n",
+      " 81%|████████  | 243/300 [01:51<00:26,  2.13it/s, train=0.0258, test=0.0795]\u001b[A\n",
+      " 81%|████████  | 243/300 [01:51<00:26,  2.13it/s, train=0.0255, test=0.0799]\u001b[A\n",
+      " 81%|████████▏ | 244/300 [01:51<00:25,  2.19it/s, train=0.0255, test=0.0799]\u001b[A\n",
+      " 81%|████████▏ | 244/300 [01:52<00:25,  2.19it/s, train=0.0243, test=0.0808]\u001b[A\n",
+      " 82%|████████▏ | 245/300 [01:52<00:24,  2.22it/s, train=0.0243, test=0.0808]\u001b[A\n",
+      " 82%|████████▏ | 245/300 [01:52<00:24,  2.22it/s, train=0.0242, test=0.0842]\u001b[A\n",
+      " 82%|████████▏ | 246/300 [01:52<00:23,  2.25it/s, train=0.0242, test=0.0842]\u001b[A\n",
+      " 82%|████████▏ | 246/300 [01:53<00:23,  2.25it/s, train=0.0236, test=0.0782]\u001b[A\n",
+      " 82%|████████▏ | 247/300 [01:53<00:24,  2.19it/s, train=0.0236, test=0.0782]\u001b[A\n",
+      " 82%|████████▏ | 247/300 [01:53<00:24,  2.19it/s, train=0.0227, test=0.0866]\u001b[A\n",
+      " 83%|████████▎ | 248/300 [01:53<00:24,  2.16it/s, train=0.0227, test=0.0866]\u001b[A\n",
+      " 83%|████████▎ | 248/300 [01:54<00:24,  2.16it/s, train=0.0223, test=0.082] \u001b[A\n",
+      " 83%|████████▎ | 249/300 [01:54<00:23,  2.21it/s, train=0.0223, test=0.082]\u001b[A\n",
+      " 83%|████████▎ | 249/300 [01:54<00:23,  2.21it/s, train=0.0228, test=0.0786]\u001b[A\n",
+      " 83%|████████▎ | 250/300 [01:54<00:22,  2.20it/s, train=0.0228, test=0.0786]\u001b[A\n",
+      " 83%|████████▎ | 250/300 [01:54<00:22,  2.20it/s, train=0.0226, test=0.0781]\u001b[A\n",
+      " 84%|████████▎ | 251/300 [01:54<00:22,  2.20it/s, train=0.0226, test=0.0781]\u001b[A\n",
+      " 84%|████████▎ | 251/300 [01:55<00:22,  2.20it/s, train=0.0219, test=0.0825]\u001b[A\n",
+      " 84%|████████▍ | 252/300 [01:55<00:21,  2.19it/s, train=0.0219, test=0.0825]\u001b[A\n",
+      " 84%|████████▍ | 252/300 [01:55<00:21,  2.19it/s, train=0.0269, test=0.0816]\u001b[A\n",
+      " 84%|████████▍ | 253/300 [01:55<00:21,  2.20it/s, train=0.0269, test=0.0816]\u001b[A\n",
+      " 84%|████████▍ | 253/300 [01:56<00:21,  2.20it/s, train=0.0267, test=0.0997]\u001b[A\n",
+      " 85%|████████▍ | 254/300 [01:56<00:21,  2.17it/s, train=0.0267, test=0.0997]\u001b[A\n",
+      " 85%|████████▍ | 254/300 [01:56<00:21,  2.17it/s, train=0.0256, test=0.0792]\u001b[A\n",
+      " 85%|████████▌ | 255/300 [01:56<00:20,  2.15it/s, train=0.0256, test=0.0792]\u001b[A\n",
+      " 85%|████████▌ | 255/300 [01:57<00:20,  2.15it/s, train=0.025, test=0.0741] \u001b[A\n",
+      " 85%|████████▌ | 256/300 [01:57<00:20,  2.14it/s, train=0.025, test=0.0741]\u001b[A\n",
+      " 85%|████████▌ | 256/300 [01:57<00:20,  2.14it/s, train=0.0221, test=0.0866]\u001b[A\n",
+      " 86%|████████▌ | 257/300 [01:57<00:20,  2.14it/s, train=0.0221, test=0.0866]\u001b[A\n",
+      " 86%|████████▌ | 257/300 [01:58<00:20,  2.14it/s, train=0.0217, test=0.083] \u001b[A\n",
+      " 86%|████████▌ | 258/300 [01:58<00:19,  2.15it/s, train=0.0217, test=0.083]\u001b[A\n"
+     ]
+    },
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      " 86%|████████▌ | 258/300 [01:58<00:19,  2.15it/s, train=0.0221, test=0.0804]\u001b[A\n",
+      " 86%|████████▋ | 259/300 [01:58<00:19,  2.15it/s, train=0.0221, test=0.0804]\u001b[A\n",
+      " 86%|████████▋ | 259/300 [01:59<00:19,  2.15it/s, train=0.0229, test=0.0845]\u001b[A\n",
+      " 87%|████████▋ | 260/300 [01:59<00:18,  2.16it/s, train=0.0229, test=0.0845]\u001b[A\n",
+      " 87%|████████▋ | 260/300 [01:59<00:18,  2.16it/s, train=0.0233, test=0.0871]\u001b[A\n",
+      " 87%|████████▋ | 261/300 [01:59<00:18,  2.13it/s, train=0.0233, test=0.0871]\u001b[A\n",
+      " 87%|████████▋ | 261/300 [02:00<00:18,  2.13it/s, train=0.022, test=0.0799] \u001b[A\n",
+      " 87%|████████▋ | 262/300 [02:00<00:17,  2.15it/s, train=0.022, test=0.0799]\u001b[A\n",
+      " 87%|████████▋ | 262/300 [02:00<00:17,  2.15it/s, train=0.021, test=0.0789]\u001b[A\n",
+      " 88%|████████▊ | 263/300 [02:00<00:17,  2.15it/s, train=0.021, test=0.0789]\u001b[A\n",
+      " 88%|████████▊ | 263/300 [02:00<00:17,  2.15it/s, train=0.0201, test=0.0818]\u001b[A\n",
+      " 88%|████████▊ | 264/300 [02:00<00:16,  2.15it/s, train=0.0201, test=0.0818]\u001b[A\n",
+      " 88%|████████▊ | 264/300 [02:01<00:16,  2.15it/s, train=0.0196, test=0.0841]\u001b[A\n",
+      " 88%|████████▊ | 265/300 [02:01<00:16,  2.15it/s, train=0.0196, test=0.0841]\u001b[A\n",
+      " 88%|████████▊ | 265/300 [02:01<00:16,  2.15it/s, train=0.0197, test=0.0857]\u001b[A\n",
+      " 89%|████████▊ | 266/300 [02:01<00:15,  2.16it/s, train=0.0197, test=0.0857]\u001b[A\n",
+      " 89%|████████▊ | 266/300 [02:02<00:15,  2.16it/s, train=0.0199, test=0.0818]\u001b[A\n",
+      " 89%|████████▉ | 267/300 [02:02<00:15,  2.16it/s, train=0.0199, test=0.0818]\u001b[A\n",
+      " 89%|████████▉ | 267/300 [02:02<00:15,  2.16it/s, train=0.0199, test=0.0795]\u001b[A\n",
+      " 89%|████████▉ | 268/300 [02:02<00:14,  2.17it/s, train=0.0199, test=0.0795]\u001b[A\n",
+      " 89%|████████▉ | 268/300 [02:03<00:14,  2.17it/s, train=0.0193, test=0.0832]\u001b[A\n",
+      " 90%|████████▉ | 269/300 [02:03<00:14,  2.21it/s, train=0.0193, test=0.0832]\u001b[A\n",
+      " 90%|████████▉ | 269/300 [02:03<00:14,  2.21it/s, train=0.0194, test=0.0808]\u001b[A\n",
+      " 90%|█████████ | 270/300 [02:03<00:13,  2.22it/s, train=0.0194, test=0.0808]\u001b[A\n",
+      " 90%|█████████ | 270/300 [02:04<00:13,  2.22it/s, train=0.0191, test=0.0806]\u001b[A\n",
+      " 90%|█████████ | 271/300 [02:04<00:12,  2.26it/s, train=0.0191, test=0.0806]\u001b[A\n",
+      " 90%|█████████ | 271/300 [02:04<00:12,  2.26it/s, train=0.019, test=0.0809] \u001b[A\n",
+      " 91%|█████████ | 272/300 [02:04<00:12,  2.24it/s, train=0.019, test=0.0809]\u001b[A\n",
+      " 91%|█████████ | 272/300 [02:05<00:12,  2.24it/s, train=0.0187, test=0.0807]\u001b[A\n",
+      " 91%|█████████ | 273/300 [02:05<00:12,  2.24it/s, train=0.0187, test=0.0807]\u001b[A\n",
+      " 91%|█████████ | 273/300 [02:05<00:12,  2.24it/s, train=0.0188, test=0.089] \u001b[A\n",
+      " 91%|█████████▏| 274/300 [02:05<00:11,  2.20it/s, train=0.0188, test=0.089]\u001b[A\n",
+      " 91%|█████████▏| 274/300 [02:05<00:11,  2.20it/s, train=0.0193, test=0.0763]\u001b[A\n",
+      " 92%|█████████▏| 275/300 [02:05<00:11,  2.17it/s, train=0.0193, test=0.0763]\u001b[A\n",
+      " 92%|█████████▏| 275/300 [02:06<00:11,  2.17it/s, train=0.02, test=0.0876]  \u001b[A\n",
+      " 92%|█████████▏| 276/300 [02:06<00:10,  2.20it/s, train=0.02, test=0.0876]\u001b[A\n",
+      " 92%|█████████▏| 276/300 [02:06<00:10,  2.20it/s, train=0.0218, test=0.0797]\u001b[A\n",
+      " 92%|█████████▏| 277/300 [02:06<00:10,  2.18it/s, train=0.0218, test=0.0797]\u001b[A\n",
+      " 92%|█████████▏| 277/300 [02:07<00:10,  2.18it/s, train=0.0204, test=0.0821]\u001b[A\n",
+      " 93%|█████████▎| 278/300 [02:07<00:10,  2.18it/s, train=0.0204, test=0.0821]\u001b[A\n",
+      " 93%|█████████▎| 278/300 [02:07<00:10,  2.18it/s, train=0.0192, test=0.0777]\u001b[A\n",
+      " 93%|█████████▎| 279/300 [02:07<00:09,  2.20it/s, train=0.0192, test=0.0777]\u001b[A\n",
+      " 93%|█████████▎| 279/300 [02:08<00:09,  2.20it/s, train=0.0187, test=0.0831]\u001b[A\n",
+      " 93%|█████████▎| 280/300 [02:08<00:09,  2.20it/s, train=0.0187, test=0.0831]\u001b[A\n",
+      " 93%|█████████▎| 280/300 [02:08<00:09,  2.20it/s, train=0.0201, test=0.0979]\u001b[A\n",
+      " 94%|█████████▎| 281/300 [02:08<00:08,  2.17it/s, train=0.0201, test=0.0979]\u001b[A\n",
+      " 94%|█████████▎| 281/300 [02:09<00:08,  2.17it/s, train=0.0222, test=0.0816]\u001b[A\n",
+      " 94%|█████████▍| 282/300 [02:09<00:08,  2.12it/s, train=0.0222, test=0.0816]\u001b[A\n",
+      " 94%|█████████▍| 282/300 [02:09<00:08,  2.12it/s, train=0.0232, test=0.0854]\u001b[A\n",
+      " 94%|█████████▍| 283/300 [02:09<00:08,  2.10it/s, train=0.0232, test=0.0854]\u001b[A\n",
+      " 94%|█████████▍| 283/300 [02:10<00:08,  2.10it/s, train=0.0232, test=0.0852]\u001b[A\n",
+      " 95%|█████████▍| 284/300 [02:10<00:07,  2.09it/s, train=0.0232, test=0.0852]\u001b[A\n",
+      " 95%|█████████▍| 284/300 [02:10<00:07,  2.09it/s, train=0.0239, test=0.0837]\u001b[A\n",
+      " 95%|█████████▌| 285/300 [02:10<00:07,  2.12it/s, train=0.0239, test=0.0837]\u001b[A\n",
+      " 95%|█████████▌| 285/300 [02:11<00:07,  2.12it/s, train=0.0222, test=0.0846]\u001b[A\n",
+      " 95%|█████████▌| 286/300 [02:11<00:06,  2.10it/s, train=0.0222, test=0.0846]\u001b[A\n",
+      " 95%|█████████▌| 286/300 [02:11<00:06,  2.10it/s, train=0.0201, test=0.0848]\u001b[A\n",
+      " 96%|█████████▌| 287/300 [02:11<00:06,  2.13it/s, train=0.0201, test=0.0848]\u001b[A\n",
+      " 96%|█████████▌| 287/300 [02:12<00:06,  2.13it/s, train=0.0189, test=0.08]  \u001b[A\n",
+      " 96%|█████████▌| 288/300 [02:12<00:05,  2.15it/s, train=0.0189, test=0.08]\u001b[A\n",
+      " 96%|█████████▌| 288/300 [02:12<00:05,  2.15it/s, train=0.0186, test=0.0839]\u001b[A\n",
+      " 96%|█████████▋| 289/300 [02:12<00:05,  2.17it/s, train=0.0186, test=0.0839]\u001b[A\n",
+      " 96%|█████████▋| 289/300 [02:12<00:05,  2.17it/s, train=0.0184, test=0.0803]\u001b[A\n",
+      " 97%|█████████▋| 290/300 [02:12<00:04,  2.19it/s, train=0.0184, test=0.0803]\u001b[A\n",
+      " 97%|█████████▋| 290/300 [02:13<00:04,  2.19it/s, train=0.0175, test=0.0907]\u001b[A\n",
+      " 97%|█████████▋| 291/300 [02:13<00:04,  2.18it/s, train=0.0175, test=0.0907]\u001b[A\n",
+      " 97%|█████████▋| 291/300 [02:13<00:04,  2.18it/s, train=0.0172, test=0.0795]\u001b[A\n",
+      " 97%|█████████▋| 292/300 [02:13<00:03,  2.20it/s, train=0.0172, test=0.0795]\u001b[A\n",
+      " 97%|█████████▋| 292/300 [02:14<00:03,  2.20it/s, train=0.0172, test=0.0835]\u001b[A\n",
+      " 98%|█████████▊| 293/300 [02:14<00:03,  2.20it/s, train=0.0172, test=0.0835]\u001b[A\n",
+      " 98%|█████████▊| 293/300 [02:14<00:03,  2.20it/s, train=0.018, test=0.0822] \u001b[A\n",
+      " 98%|█████████▊| 294/300 [02:14<00:02,  2.21it/s, train=0.018, test=0.0822]\u001b[A\n",
+      " 98%|█████████▊| 294/300 [02:15<00:02,  2.21it/s, train=0.0197, test=0.0794]\u001b[A\n",
+      " 98%|█████████▊| 295/300 [02:15<00:02,  2.20it/s, train=0.0197, test=0.0794]\u001b[A\n",
+      " 98%|█████████▊| 295/300 [02:15<00:02,  2.20it/s, train=0.0213, test=0.0986]\u001b[A\n",
+      " 99%|█████████▊| 296/300 [02:15<00:01,  2.22it/s, train=0.0213, test=0.0986]\u001b[A\n",
+      " 99%|█████████▊| 296/300 [02:16<00:01,  2.22it/s, train=0.0199, test=0.0852]\u001b[A\n",
+      " 99%|█████████▉| 297/300 [02:16<00:01,  2.24it/s, train=0.0199, test=0.0852]\u001b[A\n",
+      " 99%|█████████▉| 297/300 [02:16<00:01,  2.24it/s, train=0.0208, test=0.0836]\u001b[A\n",
+      " 99%|█████████▉| 298/300 [02:16<00:00,  2.24it/s, train=0.0208, test=0.0836]\u001b[A\n",
+      " 99%|█████████▉| 298/300 [02:16<00:00,  2.24it/s, train=0.0187, test=0.0863]\u001b[A\n",
+      "100%|█████████▉| 299/300 [02:16<00:00,  2.24it/s, train=0.0187, test=0.0863]\u001b[A\n",
+      "100%|█████████▉| 299/300 [02:17<00:00,  2.24it/s, train=0.0179, test=0.0827]\u001b[A\n",
+      "100%|██████████| 300/300 [02:17<00:00,  2.18it/s, train=0.0179, test=0.0827]\u001b[A\n"
      ]
     }
    ],
@@ -424,7 +1057,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 59,
+   "execution_count": 15,
    "metadata": {
     "ExecuteTime": {
      "end_time": "2020-11-19T20:06:21.401827Z",
@@ -438,7 +1071,7 @@
    "outputs": [
     {
      "data": {
-      "image/png": "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\n",
+      "image/png": "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\n",
       "text/plain": [
        "<Figure size 432x288 with 1 Axes>"
       ]
@@ -707,7 +1340,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.7.7"
+   "version": "3.7.10"
   },
   "toc": {
    "base_numbering": 1,