Skip to content

Commit 50c55e0

Browse files
committed
latency script more stats
1 parent 5992a38 commit 50c55e0

File tree

1 file changed

+6
-2
lines changed
  • users/zeyer/experiments/exp2023_02_16_chunked_attention/scripts

1 file changed

+6
-2
lines changed

users/zeyer/experiments/exp2023_02_16_chunked_attention/scripts/latency.py

Lines changed: 6 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -483,13 +483,17 @@ def main():
483483
if max(seg_latencies) > args.chunk_left_padding + args.chunk_size:
484484
outliers.append((segment_name, max(seg_latencies)))
485485
res += seg_latencies
486-
print(f"avg latency: {sum(res) / len(res)}sec")
487-
print(f"max latency: {max(res)}sec")
488486
print("outliers:")
489487
for segment_name, latency in outliers:
490488
print(f" {segment_name}: {latency}sec")
491489
if not outliers:
492490
print(" (no outliers)")
491+
print(f"avg latency: {sum(res) / len(res)}sec")
492+
print(f"max latency: {max(res)}sec")
493+
res = sorted(res)
494+
print(f"median latency: {res[len(res) // 2]}sec")
495+
print(f"p90 latency: {res[int(len(res) * 0.9)]}sec")
496+
print(f"p95 latency: {res[int(len(res) * 0.95)]}sec")
493497

494498

495499
if __name__ == "__main__":

0 commit comments

Comments
 (0)