Efficient net and non working transformer model.

1 files changed, 54 insertions, 9 deletions

diff --git a/notebooks/03-look-at-iam-paragraphs.ipynb b/notebooks/03-look-at-iam-paragraphs.ipynb
index eaf5397..add0b80 100644
--- a/notebooks/03-look-at-iam-paragraphs.ipynb
+++ b/notebooks/03-look-at-iam-paragraphs.ipynb
@@ -31,7 +31,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 6,
+   "execution_count": 2,
    "id": "726ac25b",
    "metadata": {},
    "outputs": [],
@@ -56,8 +56,8 @@
      "name": "stderr",
      "output_type": "stream",
      "text": [
-      "2021-04-16 23:01:52.352 | INFO     | text_recognizer.data.iam_paragraphs:setup:107 - Loading IAM paragraph regions and lines for None...\n",
-      "2021-04-16 23:02:08.521 | INFO     | text_recognizer.data.iam_synthetic_paragraphs:setup:79 - IAM Synthetic dataset steup for stage None\n"
+      "2021-04-25 23:17:44.177 | INFO     | text_recognizer.data.iam_paragraphs:setup:107 - Loading IAM paragraph regions and lines for None...\n",
+      "2021-04-25 23:18:00.750 | INFO     | text_recognizer.data.iam_synthetic_paragraphs:setup:79 - IAM Synthetic dataset steup for stage None\n"
      ]
     },
     {
@@ -68,9 +68,9 @@
       "Num classes: 84\n",
       "Dims: (1, 576, 640)\n",
       "Output dims: (682, 1)\n",
-      "Train/val/test sizes: 19912, 262, 231\n",
-      "Train Batch x stats: (torch.Size([1, 1, 576, 640]), torch.float32, tensor(0.), tensor(0.0043), tensor(0.0333), tensor(0.8588))\n",
-      "Train Batch y stats: (torch.Size([1, 682]), torch.int64, tensor(1), tensor(78))\n",
+      "Train/val/test sizes: 19948, 262, 231\n",
+      "Train Batch x stats: (torch.Size([1, 1, 576, 640]), torch.float32, tensor(0.), tensor(0.0109), tensor(0.0499), tensor(0.8314))\n",
+      "Train Batch y stats: (torch.Size([1, 682]), torch.int64, tensor(1), tensor(83))\n",
       "Test Batch x stats: (torch.Size([1, 1, 576, 640]), torch.float32, tensor(0.), tensor(0.0372), tensor(0.0767), tensor(0.8118))\n",
       "Test Batch y stats: (torch.Size([1, 682]), torch.int64, tensor(1), tensor(83))\n",
       "\n"
@@ -86,10 +86,34 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 4,
    "id": "42501428",
    "metadata": {},
-   "outputs": [],
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "2021-04-25 23:18:14.449 | INFO     | text_recognizer.data.iam_paragraphs:setup:107 - Loading IAM paragraph regions and lines for None...\n"
+     ]
+    },
+    {
+     "name": "stdout",
+     "output_type": "stream",
+     "text": [
+      "IAM Paragraphs Dataset\n",
+      "Num classes: 84\n",
+      "Input dims: (1, 576, 640)\n",
+      "Output dims: (682, 1)\n",
+      "Train/val/test sizes: 1046, 262, 231\n",
+      "Train Batch x stats: (torch.Size([16, 1, 576, 640]), torch.float32, tensor(0.), tensor(0.0393), tensor(0.0924), tensor(1.))\n",
+      "Train Batch y stats: (torch.Size([16, 682]), torch.int64, tensor(1), tensor(83))\n",
+      "Test Batch x stats: (torch.Size([16, 1, 576, 640]), torch.float32, tensor(0.), tensor(0.0312), tensor(0.0817), tensor(0.9294))\n",
+      "Test Batch y stats: (torch.Size([16, 682]), torch.int64, tensor(1), tensor(83))\n",
+      "\n"
+     ]
+    }
+   ],
    "source": [
     "dataset = IAMParagraphs()\n",
     "dataset.prepare_data()\n",
@@ -99,7 +123,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 5,
    "id": "0cf22683",
    "metadata": {},
    "outputs": [],
@@ -109,6 +133,27 @@
   },
   {
    "cell_type": "code",
+   "execution_count": 6,
+   "id": "af7747a8",
+   "metadata": {},
+   "outputs": [
+    {
+     "data": {
+      "text/plain": [
+       "torch.Size([682])"
+      ]
+     },
+     "execution_count": 6,
+     "metadata": {},
+     "output_type": "execute_result"
+    }
+   ],
+   "source": [
+    "y.shape"
+   ]
+  },
+  {
+   "cell_type": "code",
    "execution_count": 7,
    "id": "e7778ae2",
    "metadata": {