Random code push

simulator/configs/arvix_linear.json

@@ -0,0 +1,8 @@
+{
+    "dataset_name" : "ogbn_arxiv",
+    "features_stats" : {
+        "page_size" : "16 KB",
+        "feature_dimension" : 128,
+        "feature_size" : "float32"
+    }
+}

simulator/configs/arvix_random.json

@@ -0,0 +1,9 @@
+{
+    "dataset_name" : "ogbn_arxiv",
+    "features_stats" : {
+        "feature_layout" : "random",
+        "page_size" : "16 KB",
+        "feature_dimension" : 128,
+        "feature_size" : "float32"
+    }
+}

simulator/main.py

@@ -44,7 +44,14 @@ def main():
 
     # Save the histogram
     os.makedirs(os.path.dirname(arguments.save_path), exist_ok=True)
-    visualize_results(pages_loaded, arguments.save_path, arguments.graph_title, config["dataset_name"])
+    visualize_arguments = {
+        "pages_loaded": pages_loaded,
+        "save_path": arguments.save_path,
+        "graph_title": arguments.graph_title,
+        "total_space": features_loader.get_total_file_size(),
+        "dataset_name": config["dataset_name"],
+    }
+    visualize_results(visualize_arguments)
 
 
 if __name__ == "__main__":

simulator/src/features_loader.py

@@ -21,8 +21,8 @@ def __init__(self, data_loader, features_stat):
 
     def get_node_page(self, node_id):
         node_location = self.node_location_map[node_id]
-        return int(node_id / self.nodes_per_page)
+        return int(node_location / self.nodes_per_page)
 
     def get_total_file_size(self):
-        total_bytes = self.page_size * self.total_bytes
+        total_bytes = self.page_size * self.total_pages
         return humanfriendly.format_size(total_bytes)

simulator/src/visualizer.py

@@ -1,18 +1,44 @@
 import matplotlib.pyplot as plt
 import os
+import numpy as np
 
 
-def visualize_results(pages_loaded, save_path, graph_title, dataset_name, num_bins=50):
+def visualize_results(visualize_args, num_bins=60, x_range=(0, 75), write_location=(0.75, 0.6)):
+    # Get the number of pages read
+    pages_loaded = visualize_args["pages_loaded"]
+    np_arr = np.array(pages_loaded)
+    page_mean, page_std = round(np.mean(np_arr), 2), round(np.std(np_arr), 2)
+
     # Create the histogram
     plt.figure()
     plt.ecdf(pages_loaded, label="CDF")
     plt.hist(pages_loaded, bins=num_bins, histtype="step", density=True, cumulative=True, label="Cumulative histogram")
     plt.xlabel("Number of pages loaded for node inference")
     plt.ylabel("Percentage of nodes")
-    plt.title(graph_title)
-    plt.xlim(0, 50)
+    plt.title(visualize_args["graph_title"] + " for dataset " + visualize_args["dataset_name"])
+    plt.xlim(x_range)
     plt.legend()
 
+    # Write some resulting text
+    text_to_write = "Mean Pages Loaded: " + str(page_mean) + "\n"
+    text_to_write += "Std Dev of Pages Loaded: " + str(page_std) + "\n"
+    text_to_write += "Feature File Size: " + visualize_args["total_space"]
+
+    # Get the current axis limits
+    xlim = plt.xlim()
+    ylim = plt.ylim()
+    actual_x = write_location[0] * (xlim[1] - xlim[0]) + xlim[0]
+    actual_y = write_location[1] * (ylim[1] - ylim[0]) + ylim[0]
+    plt.text(
+        actual_x,
+        actual_y,
+        text_to_write,
+        fontsize=10,
+        horizontalalignment="center",
+        verticalalignment="center",
+        bbox=dict(facecolor="red", alpha=0.5),
+    )
+
     # Save the result
-    print("Saving the result to", save_path)
-    plt.savefig(save_path)
+    plt.tight_layout()
+    plt.savefig(visualize_args["save_path"])