core: sort benchmarks by speed

README.md

@@ -11,8 +11,8 @@ Running on an A100 80G SXM hosted at [fal.ai](https://fal.ai).
 ### SD1.5 Benchmarks
 |                  | mean (s) | median (s) | min (s) | max (s) | speed (it/s) |
 |------------------|----------|------------|---------|---------|--------------|
-| Diffusers (fp16, SDPA) |   1.591s |     1.590s |  1.581s |  1.601s |   31.44 it/s |
 | Diffusers (fp16, xformers) |   1.758s |     1.759s |  1.746s |  1.772s |   28.43 it/s |
+| Diffusers (fp16, SDPA) |   1.591s |     1.590s |  1.581s |  1.601s |   31.44 it/s |
 | Diffusers (fp16, SDPA, compiled) |   1.352s |     1.351s |  1.348s |  1.356s |   37.01 it/s |
 | Diffusers (fp16, SDPA, compiled, NCHW channels last) |   1.066s |     1.065s |  1.062s |  1.076s |   46.95 it/s |
 | TensorRT 9.0 (cuda graphs, static shapes) |   0.819s |     0.818s |  0.817s |  0.821s |   61.14 it/s |

benchmarks/update_table.py

@@ -33,7 +33,11 @@ def main():
 
     all_rows = defaultdict(list)
     steps = results["parameters"]["steps"]
-    for timing in results["timings"]:
+    for timing in sorted(
+        results["timings"],
+        key=lambda timing: statistics.mean(timing["timings"]),
+        reverse=True,
+    ):
         benchmark_name = timing["name"]
         benchmark_timings = timing["timings"]
         row = TABLE_ROW_FORMAT.format(
@@ -47,7 +51,7 @@ def main():
         all_rows[timing["category"]].append(row)
 
     tables = []
-    for category, rows in all_rows.items():
+    for category, rows in sorted(all_rows.items(), key=lambda kv: kv[0]):
         tables.append(f"### {category} Benchmarks\n")
         tables.append(TABLE_HEADER)
         tables.append(TABLE_DIVIDER)