Add Torch CUDA sync to fix timing code in cli.py

black-forest-labs · Sep 11, 2024 · fe0992e · fe0992e
1 parent d013dc4
commit fe0992e
Showing 1 changed file with 3 additions and 0 deletions.
diff --git a/src/flux/cli.py b/src/flux/cli.py
@@ -215,6 +215,9 @@ def main(
         x = unpack(x.float(), opts.height, opts.width)
         with torch.autocast(device_type=torch_device.type, dtype=torch.bfloat16):
             x = ae.decode(x)
+
+        if torch.cuda.is_available():
+            torch.cuda.synchronize()
         t1 = time.perf_counter()
 
         fn = output_name.format(idx=idx)