From e72fe9c6f82d9cdcefeba5777feeb96375bcf5b1 Mon Sep 17 00:00:00 2001
From: gvanhoy <gmvanhoy@gmail.com>
Date: Mon, 25 Sep 2023 09:29:38 -0400
Subject: [PATCH] Minor edits

---
 .../00_Sig53DatasetTutorial.ipynb             | 24 +++++++------------
 1 file changed, 9 insertions(+), 15 deletions(-)

diff --git a/docs/00_dataset_tutorials/00_Sig53DatasetTutorial.ipynb b/docs/00_dataset_tutorials/00_Sig53DatasetTutorial.ipynb
index 7475a21..da586dd 100644
--- a/docs/00_dataset_tutorials/00_Sig53DatasetTutorial.ipynb
+++ b/docs/00_dataset_tutorials/00_Sig53DatasetTutorial.ipynb
@@ -28,7 +28,7 @@
     "#### Generating\n",
     "A Sig53-like dataset can be generated from scratch, on-the-fly, yielding a dataset with essentially inifinite number of unique exemplars (called samples in machine-learning circles) per class. The downside to generating things on-the-fly is that **training is slower.** \n",
     "\n",
-    "Before we jump into generating the whole dataset, let's start with a basic example which uses the ```ModulationsDataset```, which is the underlying class used to generate ```Sig53Dataset```. The class has a number of parameters"
+    "Before we jump into generating the whole dataset, let's start with a basic example which uses the ```ModulationsDataset```, which is the underlying class used to generate ```Sig53``` Dataset. The class has a number of parameters"
    ]
   },
   {
@@ -152,7 +152,7 @@
    "metadata": {},
    "source": [
     "#### Iterating Over the Dataset\n",
-    "Now the Sig53 dataset is a specific configuration of the ```ModulationsDataset``` class. To keep the configuration fixed, we implement classes with fixed parameters of type ```Sig53Conf```. We use the ```Sig53CleanTrianQAConfig```. In this configuration, the dataset ```label``` is actually a Python ```tuple``` which includes the estimated SNR of the produced sample. \n",
+    "Now the Sig53 dataset is a specific configuration of the ```ModulationsDataset``` class. To keep the configuration fixed, we implement classes with fixed parameters of type ```Sig53Conf```. We use the ```Sig53CleanTrainQAConfig```. In this configuration, the dataset ```label``` is actually a Python ```tuple``` which includes the estimated SNR of the produced sample. \n",
     "\n",
     "Using this example, you can choose one of the fixed configurations and generate a Sig53-like dataset on-the-fly as you iterate over it, and train an ML model, store it, or perform some analysis on it!"
    ]
@@ -180,8 +180,6 @@
     ")\n",
     "\n",
     "data, (modulation, snr) = ds[0]\n",
-    "print(data)\n",
-    "print(modulation, snr)\n",
     "\n",
     "# Plot it.\n",
     "plt.subplot(3, 1, 1)\n",
@@ -253,7 +251,7 @@
     "\n",
     "# You can iterate through this in a similar way.\n",
     "# The DataLoader produces a torch.Tensor\n",
-    "loader = DataLoader(ds, batch_size=16, num_workers=16)\n",
+    "loader = DataLoader(ds, batch_size=os.cpu_count() // 2, num_workers=os.cpu_count() // 2)\n",
     "\n",
     "if os.path.exists(\"data.fc32\"):\n",
     "    os.remove(\"data.fc32\")\n",
@@ -277,7 +275,7 @@
    "metadata": {},
    "source": [
     "#### Reading a Static Dataset\n",
-    "We've chosen to write our data as raw binary. We can write a Dataset class that allows us to iterate over that stored data. A Dataset class only needs to implement ```__init__``` and ```__getitem__```. We talk more about creating a custom dataset in a different tutorial."
+    "We've chosen to write our data as raw binary. We can write a Dataset class that allows us to iterate over that stored data. A Dataset class only needs to implement ```__init__```, ```__getitem__```, and ```__len__```. We talk more about creating a custom dataset in a different tutorial."
    ]
   },
   {
@@ -304,6 +302,10 @@
     "        self.label_file = os.path.join(path, \"label.int8\")\n",
     "\n",
     "        super().__init__(transform, target_transform, seed)\n",
+    "    \n",
+    "    def __len__(self) -> int:\n",
+    "        # Each sample is 8 * 4096 * 2 bytes large.\n",
+    "        return os.path.getsize(self.data_file) // 8 // 4092 // 2\n",
     "\n",
     "    def __getitem__(self, index: int) -> Tuple:\n",
     "        # the sample at index is 8192 double-precision floating-point numbers long.\n",
@@ -448,7 +450,7 @@
     "\n",
     "# Plot it.\n",
     "plt.subplot(3, 1, 1)\n",
-    "plt.title(\"Modulation Type = {}\".format(ModulationsDataset.default_classes[int(label)]))\n",
+    "plt.title(\"Modulation Type = {}\".format(ModulationsDataset.default_classes[int(mod)]))\n",
     "plt.ylabel(\"Time Domain\")\n",
     "plt.plot(data[:100].real, marker=\".\")\n",
     "plt.plot(data[:100].imag, marker=\".\")\n",
@@ -475,14 +477,6 @@
     "ax.set_xticks([])\n",
     "ax.set_yticks([])\n"
    ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "1fca0286",
-   "metadata": {},
-   "outputs": [],
-   "source": []
   }
  ],
  "metadata": {