marketneutral · May 14, 2025 00:59 · May 14, 2025
diff --git a/tensorscrape.py b/tensorscrape.py
@@ -0,0 +1,82 @@
+import argparse
+import os
+import matplotlib.pyplot as plt
+from tensorboard.backend.event_processing import event_accumulator
+import numpy as np
+
+def export_scalars(logdir, output_dir, tags=None, fmt='png', dpi=300, stem=None):
+    os.makedirs(output_dir, exist_ok=True)
+    ea = event_accumulator.EventAccumulator(logdir)
+    ea.Reload()
+
+    available_tags = ea.Tags().get('scalars', [])
+    logdir_name = os.path.basename(os.path.normpath(logdir))
+
+    if tags:
+        tags_to_export = [t for t in available_tags if t in tags]
+    else:
+        tags_to_export = available_tags
+
+    if not tags_to_export:
+        print("No matching tags found.")
+        return
+
+    for tag in tags_to_export:
+        events = ea.Scalars(tag)
+        steps = [e.step for e in events]
+        values = [e.value for e in events]
+
+        values_np = np.array(values)
+        summary_text = (
+            f"Mean: {values_np.mean():.4f}\n"
+            f"Max: {values_np.max():.4f}\n"
+            f"Min: {values_np.min():.4f}\n"
+            f"Final: {values_np[-1]:.4f}"
+        )
+
+        plt.figure(figsize=(8, 5))
+        plt.plot(steps, values, label=tag)
+        plt.xlabel("Step")
+        plt.ylabel("Value")
+        plt.title(tag)
+        plt.suptitle(f"Source: {logdir_name}", fontsize=10, y=0.92)
+
+        # Add summary box
+        props = dict(boxstyle='round', facecolor='white', alpha=0.8)
+        plt.gca().text(0.98, 0.02, summary_text,
+                       transform=plt.gca().transAxes,
+                       fontsize=9,
+                       verticalalignment='bottom',
+                       horizontalalignment='right',
+                       bbox=props)
+
+        plt.legend()
+        plt.tight_layout(rect=[0, 0, 1, 0.95])
+
+        # Build output filename
+        tag_sanitized = tag.replace('/', '_')
+        if stem:
+            filename = os.path.join(output_dir, f"{stem}_{tag_sanitized}.{fmt}")
+        else:
+            filename = os.path.join(output_dir, f"{tag_sanitized}.{fmt}")
+
+        plt.savefig(filename, dpi=dpi)
+        plt.close()
+
+    print(f"Exported {len(tags_to_export)} plots to '{output_dir}' as .{fmt} files")
+
+def main():
+    parser = argparse.ArgumentParser(description="Export TensorBoard scalar plots to images.")
+    parser.add_argument("--logdir", required=True, help="Path to the TensorBoard log directory")
+    parser.add_argument("--output_dir", default="./exported_plots", help="Directory to save the exported plots")
+    parser.add_argument("--format", default="png", choices=["png", "pdf", "svg"], help="Image format to export")
+    parser.add_argument("--dpi", type=int, default=300, help="Resolution of exported images")
+    parser.add_argument("--tags", type=str, help="Comma-separated list of scalar tags to export (e.g., loss,accuracy)")
+    parser.add_argument("--stem", type=str, help="Optional filename prefix (e.g., 'exp1' -> 'exp1_loss.png')")
+
+    args = parser.parse_args()
+    tag_list = [tag.strip() for tag in args.tags.split(',')] if args.tags else None
+    export_scalars(args.logdir, args.output_dir, tag_list, args.format, args.dpi, args.stem)
+
+if __name__ == "__main__":
+    main()