Nine-id ops / TP2 subset (mobilenetv2, yolov8s, libface, pphuman_mobile, dbnet, dinov2_vits14, superpoint, realesrgan, metric3d_512). The 39-id product catalog skeleton is product_static_http_rps.md (python3 scripts/coverage_http_report.py --catalog product_static). This file is not the full catalog.
worktree: /mnt/g/Codex/mortred_wt_r1_p2_11 hardware: NVIDIA GeForce RTX 2070 SUPER, 8192 MiB date: 2026-10-14
PF5 recorded baselines for these rows: docs/perf/baselines/<id>.json (imported from the numbers below; not a second measurement). Thresholds and retest commands: docs/perf/pf5_retest.md.
checked=yes | worktree=/mnt/g/Codex/mortred_wt_r1_p2_11 | hardware="NVIDIA GeForce RTX 2070 SUPER, 8192 MiB" | date=2026-10-14 | server_uri=/mortred_ai_server_v1/classification/mobilenetv2 | conf_server=conf/server/classification/mobilenetv2/mobilenetv2_server_config.toml | conf_server_worker_nums=1 | worker_nums=4 | decode_auto=cpu | prepare="mortredctl prepare --id mobilenetv2" | calibrate="mortredctl calibrate --pack /tmp/mortred-p2-11-mobilenetv2-b3yy_35j/pack.toml --write-pack --duration 60s" | calibrate_raw="worker sweep and jpeg decode vote use --raw" | MORTRED_WORKER_NUMS=4 | MORTRED_DECODE_AUTO=cpu | process_workers=4 | process_workers_available=4.0 | process_workers_busy=0.0 | process_workers_source=after_load | process_workers_note="官方压测产生完成响应且进程仍在时，mortred_workers_available{model="/mortred_ai_server_v1/classification/mobilenetv2"} 与 mortred_workers_busy{model="/mortred_ai_server_v1/classification/mobilenetv2"} 之和；不是压测前刮取" | load_command="python3 scripts/server/http_infer_rps.py --raw --model mobilenetv2 --gateway http://127.0.0.1:8080 -c 8 -d 60s --warmup 5s --out /tmp/mortred-p2-11-mobilenetv2-b3yy_35j/load.json --quiet" | load_url=http://127.0.0.1:8080/v1/models/mobilenetv2/infer | wall_s=60.067 | warmup_s=5.0 | success_rps=894.751 | p50_ms=8.538 | p99_ms=14.556 | http_429=0 | error_rate=0.0 | requests=53745 | errors=0 | throughput_qps=1496.0 | ratio=0.5980955882352941 | mortred_queue_wait_duration_ms_count_delta=58509.0 | mortred_queue_wait_duration_ms_model=/mortred_ai_server_v1/classification/mobilenetv2 | mortred_inference_duration_ms_count_delta=58509.0 | mortred_inference_duration_ms_model=/mortred_ai_server_v1/classification/mobilenetv2 | jpeg_jpeggpu=0.0 | jpeg_cpu_reduced=58509.0 | jpeg_cpu_full=0.0 | jpeg_landing=cpu | gateway_gw_in_p50_ms=0.025208096370592493 | gateway_route_p50_ms=0.02505309582940824 | gateway_auth_p50_ms=0.025076718669638266 | gateway_fwd_p50_ms=0.027951939613988154 | gateway_upstream_p50_ms=7.586866758857099 | gateway_stage_p50_sum=7.690156609340725 | server_time_ms_p50=6.0 | dominant_stage=upstream | dominant_gauge=absent | image=- | decoded_width=- | decoded_height=- | engine="static shape batch=1"
checked=yes | worktree=/mnt/g/Codex/mortred_wt_r1_p2_11 | hardware="NVIDIA GeForce RTX 2070 SUPER, 8192 MiB" | date=2026-10-14 | server_uri=/mortred_ai_server_v1/obj_detection/yolov8s | conf_server=conf/server/object_detection/yolov8/yolov8s_server_config.toml | conf_server_worker_nums=1 | worker_nums=4 | decode_auto=cpu | prepare="mortredctl prepare --id yolov8s" | calibrate="mortredctl calibrate --pack /tmp/mortred-p2-11-yolov8s-r7zbzezu/pack.toml --write-pack --duration 60s" | calibrate_raw="worker sweep and jpeg decode vote use --raw" | MORTRED_WORKER_NUMS=4 | MORTRED_DECODE_AUTO=cpu | process_workers=4 | process_workers_available=4.0 | process_workers_busy=0.0 | process_workers_source=after_load | process_workers_note="官方压测产生完成响应且进程仍在时，mortred_workers_available{model="/mortred_ai_server_v1/obj_detection/yolov8s"} 与 mortred_workers_busy{model="/mortred_ai_server_v1/obj_detection/yolov8s"} 之和；不是压测前刮取" | load_command="python3 scripts/server/http_infer_rps.py --raw --model yolov8s --gateway http://127.0.0.1:8080 -c 8 -d 60s --warmup 5s --out /tmp/mortred-p2-11-yolov8s-r7zbzezu/load.json --quiet" | load_url=http://127.0.0.1:8080/v1/models/yolov8s/infer | wall_s=60.0726 | warmup_s=5.0 | success_rps=272.321 | p50_ms=28.053 | p99_ms=42.064 | http_429=0 | error_rate=0.0 | requests=16359 | errors=0 | throughput_qps=400.677 | ratio=0.6796521886706749 | mortred_queue_wait_duration_ms_count_delta=17447.0 | mortred_queue_wait_duration_ms_model=/mortred_ai_server_v1/obj_detection/yolov8s | mortred_inference_duration_ms_count_delta=17447.0 | mortred_inference_duration_ms_model=/mortred_ai_server_v1/obj_detection/yolov8s | jpeg_jpeggpu=0.0 | jpeg_cpu_reduced=17447.0 | jpeg_cpu_full=0.0 | jpeg_landing=cpu | gateway_gw_in_p50_ms=0.025145566701256773 | gateway_route_p50_ms=0.025028691111493664 | gateway_auth_p50_ms=0.02505888773985982 | gateway_fwd_p50_ms=0.09657998718496369 | gateway_upstream_p50_ms=35.76745631830377 | gateway_stage_p50_sum=35.93926945104135 | server_time_ms_p50=25.0 | dominant_stage=upstream | dominant_gauge=absent | image=- | decoded_width=- | decoded_height=- | engine="static shape batch=1"
checked=yes | worktree=/mnt/g/Codex/mortred_wt_r1_p2_11 | hardware="NVIDIA GeForce RTX 2070 SUPER, 8192 MiB" | date=2026-10-14 | server_uri=/mortred_ai_server_v1/obj_detection/libface | conf_server=conf/server/object_detection/libface_det/libface_server_config.toml | conf_server_worker_nums=1 | worker_nums=4 | decode_auto=gpu | prepare="mortredctl prepare --id libface" | calibrate="mortredctl calibrate --pack /tmp/mortred-p2-11-libface-5y7pumc3/pack.toml --write-pack --duration 60s" | calibrate_raw="worker sweep and jpeg decode vote use --raw" | MORTRED_WORKER_NUMS=4 | MORTRED_DECODE_AUTO=gpu | process_workers=4 | process_workers_available=4.0 | process_workers_busy=0.0 | process_workers_source=after_load | process_workers_note="官方压测产生完成响应且进程仍在时，mortred_workers_available{model="/mortred_ai_server_v1/obj_detection/libface"} 与 mortred_workers_busy{model="/mortred_ai_server_v1/obj_detection/libface"} 之和；不是压测前刮取" | load_command="python3 scripts/server/http_infer_rps.py --raw --model libface --gateway http://127.0.0.1:8080 -c 8 -d 60s --warmup 5s --out /tmp/mortred-p2-11-libface-5y7pumc3/load.json --quiet" | load_url=http://127.0.0.1:8080/v1/models/libface/infer | wall_s=60.0734 | warmup_s=5.0 | success_rps=389.507 | p50_ms=20.419 | p99_ms=28.819 | http_429=0 | error_rate=0.0 | requests=23399 | errors=0 | throughput_qps=1668.97 | ratio=0.23338166653684608 | mortred_queue_wait_duration_ms_count_delta=25336.0 | mortred_queue_wait_duration_ms_model=/mortred_ai_server_v1/obj_detection/libface | mortred_inference_duration_ms_count_delta=25336.0 | mortred_inference_duration_ms_model=/mortred_ai_server_v1/obj_detection/libface | jpeg_jpeggpu=25336.0 | jpeg_cpu_reduced=0.0 | jpeg_cpu_full=0.0 | jpeg_landing=gpu | gateway_gw_in_p50_ms=0.025005921831819978 | gateway_route_p50_ms=0.025002960565270595 | gateway_auth_p50_ms=0.02500690907655257 | gateway_fwd_p50_ms=0.02511200095151251 | gateway_upstream_p50_ms=17.98017321898616 | gateway_stage_p50_sum=18.080301011411315 | server_time_ms_p50=19.0 | dominant_stage=upstream | dominant_gauge=absent | image=- | decoded_width=- | decoded_height=- | engine="static shape batch=1"
checked=yes | worktree=/mnt/g/Codex/mortred_wt_r1_p2_11 | hardware="NVIDIA GeForce RTX 2070 SUPER, 8192 MiB" | date=2026-10-14 | server_uri=/mortred_ai_server_v1/scene_segmentation/pphuman_mobile | conf_server=conf/server/scene_segmentation/pphuman_seg/pphuman_mobile_server_config.toml | conf_server_worker_nums=1 | worker_nums=2 | decode_auto=gpu | prepare="mortredctl prepare --id pphuman_mobile" | calibrate="mortredctl calibrate --pack /tmp/mortred-p2-11-pphuman_mobile-q2mtq1su/pack.toml --write-pack --duration 60s" | calibrate_raw="worker sweep and jpeg decode vote use --raw" | MORTRED_WORKER_NUMS=2 | MORTRED_DECODE_AUTO=gpu | process_workers=2 | process_workers_available=2.0 | process_workers_busy=0.0 | process_workers_source=after_load | process_workers_note="官方压测产生完成响应且进程仍在时，mortred_workers_available{model="/mortred_ai_server_v1/scene_segmentation/pphuman_mobile"} 与 mortred_workers_busy{model="/mortred_ai_server_v1/scene_segmentation/pphuman_mobile"} 之和；不是压测前刮取" | load_command="python3 scripts/server/http_infer_rps.py --raw --model pphuman_mobile --gateway http://127.0.0.1:8080 -c 4 -d 60s --warmup 5s --image demo_data/model_test_input/scene_segmentation/face_w_mask.jpg --out /tmp/mortred-p2-11-pphuman_mobile-q2mtq1su/load.json --quiet" | load_url=http://127.0.0.1:8080/v1/models/pphuman_mobile/infer | wall_s=60.0682 | warmup_s=5.0 | success_rps=374.841 | p50_ms=10.672 | p99_ms=12.521 | http_429=0 | error_rate=0.0 | requests=22516 | errors=0 | throughput_qps=968.012 | ratio=0.38722763767391316 | mortred_queue_wait_duration_ms_count_delta=24882.0 | mortred_queue_wait_duration_ms_model=/mortred_ai_server_v1/scene_segmentation/pphuman_mobile | mortred_inference_duration_ms_count_delta=24882.0 | mortred_inference_duration_ms_model=/mortred_ai_server_v1/scene_segmentation/pphuman_mobile | jpeg_jpeggpu=24882.0 | jpeg_cpu_reduced=0.0 | jpeg_cpu_full=0.0 | jpeg_landing=gpu | gateway_gw_in_p50_ms=0.025206661804035985 | gateway_route_p50_ms=0.025090755082284612 | gateway_auth_p50_ms=0.025136380167293006 | gateway_fwd_p50_ms=0.025743905971940574 | gateway_upstream_p50_ms=12.213502398903358 | gateway_stage_p50_sum=12.314680101928912 | server_time_ms_p50=6.0 | dominant_stage=upstream | dominant_gauge=absent | image=demo_data/model_test_input/scene_segmentation/face_w_mask.jpg | decoded_width=- | decoded_height=- | engine="static shape batch=1"
checked=yes | worktree=/mnt/g/Codex/mortred_wt_r1_p2_11 | hardware="NVIDIA GeForce RTX 2070 SUPER, 8192 MiB" | date=2026-10-14 | server_uri=/mortred_ai_server_v1/ocr/dbnet | conf_server=conf/server/ocr/dbnet/dbnet_server_config.toml | conf_server_worker_nums=1 | worker_nums=2 | decode_auto=gpu | prepare="mortredctl prepare --id dbnet" | calibrate="mortredctl calibrate --pack /tmp/mortred-p2-11-dbnet-zb8nyfcy/pack.toml --write-pack --duration 60s" | calibrate_raw="worker sweep and jpeg decode vote use --raw" | MORTRED_WORKER_NUMS=2 | MORTRED_DECODE_AUTO=gpu | process_workers=2 | process_workers_available=2.0 | process_workers_busy=0.0 | process_workers_source=after_load | process_workers_note="官方压测产生完成响应且进程仍在时，mortred_workers_available{model="/mortred_ai_server_v1/ocr/dbnet"} 与 mortred_workers_busy{model="/mortred_ai_server_v1/ocr/dbnet"} 之和；不是压测前刮取" | load_command="python3 scripts/server/http_infer_rps.py --raw --model dbnet --gateway http://127.0.0.1:8080 -c 4 -d 60s --warmup 5s --image demo_data/model_test_input/ocr/coco_text_test.jpg --out /tmp/mortred-p2-11-dbnet-zb8nyfcy/load.json --quiet" | load_url=http://127.0.0.1:8080/v1/models/dbnet/infer | wall_s=60.0819 | warmup_s=5.0 | success_rps=168.67 | p50_ms=23.651 | p99_ms=25.311 | http_429=0 | error_rate=0.0 | requests=10134 | errors=0 | throughput_qps=315.002 | ratio=0.5354569177338556 | mortred_queue_wait_duration_ms_count_delta=11214.0 | mortred_queue_wait_duration_ms_model=/mortred_ai_server_v1/ocr/dbnet | mortred_inference_duration_ms_count_delta=11214.0 | mortred_inference_duration_ms_model=/mortred_ai_server_v1/ocr/dbnet | jpeg_jpeggpu=11214.0 | jpeg_cpu_reduced=0.0 | jpeg_cpu_full=0.0 | jpeg_landing=gpu | gateway_gw_in_p50_ms=0.02500668985817501 | gateway_route_p50_ms=0.025002229554980826 | gateway_auth_p50_ms=0.025002229554980826 | gateway_fwd_p50_ms=0.025361859960195406 | gateway_upstream_p50_ms=17.5335901110713 | gateway_stage_p50_sum=17.633963119999635 | server_time_ms_p50=22.0 | dominant_stage=upstream | dominant_gauge=absent | image=demo_data/model_test_input/ocr/coco_text_test.jpg | decoded_width=- | decoded_height=- | engine="static shape batch=1"
checked=yes | worktree=/mnt/g/Codex/mortred_wt_r1_p2_11 | hardware="NVIDIA GeForce RTX 2070 SUPER, 8192 MiB" | date=2026-10-14 | server_uri=/mortred_ai_server_v1/feature_embedding/dinov2_vits14 | conf_server=conf/server/feature_embedding/dinov2/dinov2_vits14_server_config.toml | conf_server_worker_nums=1 | worker_nums=2 | decode_auto=cpu | prepare="mortredctl prepare --id dinov2_vits14" | calibrate="mortredctl calibrate --pack /tmp/mortred-p2-11-dinov2_vits14-j6juuu5q/pack.toml --write-pack --duration 60s" | calibrate_raw="worker sweep and jpeg decode vote use --raw" | MORTRED_WORKER_NUMS=2 | MORTRED_DECODE_AUTO=cpu | process_workers=2 | process_workers_available=2.0 | process_workers_busy=0.0 | process_workers_source=after_load | process_workers_note="官方压测产生完成响应且进程仍在时，mortred_workers_available{model="/mortred_ai_server_v1/feature_embedding/dinov2_vits14"} 与 mortred_workers_busy{model="/mortred_ai_server_v1/feature_embedding/dinov2_vits14"} 之和；不是压测前刮取" | load_command="python3 scripts/server/http_infer_rps.py --raw --model dinov2_vits14 --gateway http://127.0.0.1:8080 -c 4 -d 60s --warmup 5s --out /tmp/mortred-p2-11-dinov2_vits14-j6juuu5q/load.json --quiet" | load_url=http://127.0.0.1:8080/v1/models/dinov2_vits14/infer | wall_s=60.0641 | warmup_s=5.0 | success_rps=179.725 | p50_ms=21.939 | p99_ms=23.838 | http_429=0 | error_rate=0.0 | requests=10795 | errors=0 | throughput_qps=287.409 | ratio=0.6253283648041641 | mortred_queue_wait_duration_ms_count_delta=11944.0 | mortred_queue_wait_duration_ms_model=/mortred_ai_server_v1/feature_embedding/dinov2_vits14 | mortred_inference_duration_ms_count_delta=11944.0 | mortred_inference_duration_ms_model=/mortred_ai_server_v1/feature_embedding/dinov2_vits14 | jpeg_jpeggpu=0.0 | jpeg_cpu_reduced=11944.0 | jpeg_cpu_full=0.0 | jpeg_landing=cpu | gateway_gw_in_p50_ms=0.02501466029990785 | gateway_route_p50_ms=0.025002093276396214 | gateway_auth_p50_ms=0.025 | gateway_fwd_p50_ms=0.025451755881350152 | gateway_upstream_p50_ms=17.504398089972355 | gateway_stage_p50_sum=17.60486659943001 | server_time_ms_p50=20.0 | dominant_stage=upstream | dominant_gauge=absent | image=- | decoded_width=- | decoded_height=- | engine="static shape batch=1"
checked=yes | worktree=/mnt/g/Codex/mortred_wt_r1_p2_11 | hardware="NVIDIA GeForce RTX 2070 SUPER, 8192 MiB" | date=2026-10-15 | server_uri=/mortred_ai_server_v1/feature_point/superpoint | conf_server=conf/server/feature_point/superpoint/superpoint_server_config.toml | conf_server_worker_nums=1 | worker_nums=1 | decode_auto=gpu | prepare="mortredctl prepare --id superpoint" | calibrate="mortredctl calibrate --pack /tmp/mortred-p2-11-superpoint-7pokmrlc/pack.toml --write-pack --duration 60s" | calibrate_raw="worker sweep and jpeg decode vote use --raw" | MORTRED_WORKER_NUMS=1 | MORTRED_DECODE_AUTO=gpu | process_workers=1 | process_workers_available=1.0 | process_workers_busy=0.0 | process_workers_source=after_load | process_workers_note="官方压测产生完成响应且进程仍在时，mortred_workers_available{model="/mortred_ai_server_v1/feature_point/superpoint"} 与 mortred_workers_busy{model="/mortred_ai_server_v1/feature_point/superpoint"} 之和；不是压测前刮取" | load_command="python3 scripts/server/http_infer_rps.py --raw --model superpoint --gateway http://127.0.0.1:8080 -c 2 -d 60s --warmup 5s --image demo_data/model_test_input/feature_point/test.jpg --out /tmp/mortred-p2-11-superpoint-7pokmrlc/load.json --quiet" | load_url=http://127.0.0.1:8080/v1/models/superpoint/infer | wall_s=60.0691 | warmup_s=5.0 | success_rps=206.978 | p50_ms=9.044 | p99_ms=17.245 | http_429=0 | error_rate=0.0 | requests=12433 | errors=0 | throughput_qps=3474.19 | ratio=0.05957590114530294 | mortred_queue_wait_duration_ms_count_delta=14440.0 | mortred_queue_wait_duration_ms_model=/mortred_ai_server_v1/feature_point/superpoint | mortred_inference_duration_ms_count_delta=14440.0 | mortred_inference_duration_ms_model=/mortred_ai_server_v1/feature_point/superpoint | jpeg_jpeggpu=14440.0 | jpeg_cpu_reduced=0.0 | jpeg_cpu_full=0.0 | jpeg_landing=gpu | gateway_gw_in_p50_ms=0.025008659508139935 | gateway_route_p50_ms=0.025001731421843622 | gateway_auth_p50_ms=0.025001731421843622 | gateway_fwd_p50_ms=0.025092097032042817 | gateway_upstream_p50_ms=3.693775414451048 | gateway_stage_p50_sum=3.7938796338349183 | server_time_ms_p50=2.0 | dominant_stage=upstream | dominant_gauge=absent | image=demo_data/model_test_input/feature_point/test.jpg | decoded_width=- | decoded_height=- | engine="static shape batch=1"
checked=yes | worktree=/mnt/g/Codex/mortred_wt_r1_p2_11 | hardware="NVIDIA GeForce RTX 2070 SUPER, 8192 MiB" | date=2026-10-15 | server_uri=/mortred_ai_server_v1/enhancement/realesrgan | conf_server=conf/server/enhancement/real_esrgan/realesrgan_server_config.toml | conf_server_worker_nums=1 | worker_nums=4 | decode_auto=cpu | prepare="mortredctl prepare --id realesrgan" | calibrate="mortredctl calibrate --pack /tmp/mortred-p2-11-realesrgan-895l6oc9/pack.toml --write-pack --duration 60s" | calibrate_raw="worker sweep and jpeg decode vote use --raw" | MORTRED_WORKER_NUMS=4 | MORTRED_DECODE_AUTO=cpu | process_workers=4 | process_workers_available=4.0 | process_workers_busy=0.0 | process_workers_source=after_load | process_workers_note="官方压测产生完成响应且进程仍在时，mortred_workers_available{model="/mortred_ai_server_v1/enhancement/realesrgan"} 与 mortred_workers_busy{model="/mortred_ai_server_v1/enhancement/realesrgan"} 之和；不是压测前刮取" | load_command="python3 scripts/server/http_infer_rps.py --raw --model realesrgan --gateway http://127.0.0.1:8080 -c 8 -d 60s --warmup 5s --image logs/p2-11-realesrgan-256.jpg --out /tmp/mortred-p2-11-realesrgan-895l6oc9/load.json --quiet" | load_url=http://127.0.0.1:8080/v1/models/realesrgan/infer | wall_s=60.3619 | warmup_s=5.0 | success_rps=24.568 | p50_ms=322.613 | p99_ms=383.273 | http_429=0 | error_rate=0.0 | requests=1483 | errors=0 | throughput_qps=50.7951 | ratio=0.48366870032739384 | mortred_queue_wait_duration_ms_count_delta=1620.0 | mortred_queue_wait_duration_ms_model=/mortred_ai_server_v1/enhancement/realesrgan | mortred_inference_duration_ms_count_delta=1620.0 | mortred_inference_duration_ms_model=/mortred_ai_server_v1/enhancement/realesrgan | jpeg_jpeggpu=0.0 | jpeg_cpu_reduced=0.0 | jpeg_cpu_full=1620.0 | jpeg_landing=cpu | gateway_gw_in_p50_ms=0.025108493490390578 | gateway_route_p50_ms=0.025 | gateway_auth_p50_ms=0.025 | gateway_fwd_p50_ms=0.025139664804469275 | gateway_upstream_p50_ms=374.5353159851301 | gateway_stage_p50_sum=374.635564143425 | server_time_ms_p50=291.0 | dominant_stage=upstream | dominant_gauge=upstream | image=logs/p2-11-realesrgan-256.jpg | decoded_width=256 | decoded_height=256 | engine="static shape batch=1"
checked=yes | worktree=/mnt/g/Codex/mortred_wt_r1_p2_11 | hardware="NVIDIA GeForce RTX 2070 SUPER, 8192 MiB" | date=2026-10-15 | server_uri=/mortred_ai_server_v1/mono_depth_estimation/metric3d_512 | conf_server=conf/server/mono_depth_estimation/metric3d/metric3d_512_server_config.toml | conf_server_worker_nums=1 | worker_nums=1 | decode_auto=gpu | prepare="mortredctl prepare --id metric3d_512" | calibrate="mortredctl calibrate --pack /tmp/mortred-p2-11-metric3d_512-921r6eb9/pack.toml --write-pack --duration 60s" | calibrate_raw="worker sweep and jpeg decode vote use --raw" | MORTRED_WORKER_NUMS=1 | MORTRED_DECODE_AUTO=gpu | process_workers=1 | process_workers_available=1.0 | process_workers_busy=0.0 | process_workers_source=after_load | process_workers_note="官方压测产生完成响应且进程仍在时，mortred_workers_available{model="/mortred_ai_server_v1/mono_depth_estimation/metric3d_512"} 与 mortred_workers_busy{model="/mortred_ai_server_v1/mono_depth_estimation/metric3d_512"} 之和；不是压测前刮取" | load_command="python3 scripts/server/http_infer_rps.py --raw --model metric3d_512 --gateway http://127.0.0.1:8080 -c 2 -d 60s --warmup 5s --image demo_data/model_test_input/mono_depth_estimation/0000000005.jpg --out /tmp/mortred-p2-11-metric3d_512-921r6eb9/load.json --quiet" | load_url=http://127.0.0.1:8080/v1/models/metric3d_512/infer | wall_s=60.1394 | warmup_s=5.0 | success_rps=22.514 | p50_ms=83.138 | p99_ms=126.457 | http_429=0 | error_rate=0.0 | requests=1354 | errors=0 | throughput_qps=20.9717 | ratio=1.0735419636939305 | mortred_queue_wait_duration_ms_count_delta=1471.0 | mortred_queue_wait_duration_ms_model=/mortred_ai_server_v1/mono_depth_estimation/metric3d_512 | mortred_inference_duration_ms_count_delta=1471.0 | mortred_inference_duration_ms_model=/mortred_ai_server_v1/mono_depth_estimation/metric3d_512 | jpeg_jpeggpu=1471.0 | jpeg_cpu_reduced=0.0 | jpeg_cpu_full=0.0 | jpeg_landing=gpu | gateway_gw_in_p50_ms=0.02501700680272109 | gateway_route_p50_ms=0.02501700680272109 | gateway_auth_p50_ms=0.02501700680272109 | gateway_fwd_p50_ms=0.02636200716845878 | gateway_upstream_p50_ms=78.8657770800628 | gateway_stage_p50_sum=78.96719010763942 | server_time_ms_p50=76.0 | dominant_stage=upstream | dominant_gauge=upstream | image=demo_data/model_test_input/mono_depth_estimation/0000000005.jpg | decoded_width=- | decoded_height=- | engine="static shape batch=1"