diff --git a/runs/frontier-fidelity-envelope-v1/.gitignore b/runs/frontier-fidelity-envelope-v1/.gitignore index acd9559..4de000f 100644 --- a/runs/frontier-fidelity-envelope-v1/.gitignore +++ b/runs/frontier-fidelity-envelope-v1/.gitignore @@ -5,9 +5,11 @@ simulator-a1/ simulator-a2/ simulator-a3/ simulator-f1/ +simulator-f1-steady/ simulator-f2/ simulator-t1/ fleet-state/ fleet-artifacts/ profiles/profile-v3-batch/ trace-exact-v1-public/ +workloads/f1-isl512-osl1-steady10s/*.csv diff --git a/runs/frontier-fidelity-envelope-v1/prepare_fixed_frontier_traces.py b/runs/frontier-fidelity-envelope-v1/prepare_fixed_frontier_traces.py index 606e5dd..60d62cc 100644 --- a/runs/frontier-fidelity-envelope-v1/prepare_fixed_frontier_traces.py +++ b/runs/frontier-fidelity-envelope-v1/prepare_fixed_frontier_traces.py @@ -26,6 +26,7 @@ def parse_args() -> argparse.Namespace: parser.add_argument("--input-tokens", type=int, required=True) parser.add_argument("--output-tokens", type=int, required=True) parser.add_argument("--requests", type=int, default=64) + parser.add_argument("--duration-seconds", type=float) parser.add_argument("--rate", type=float, action="append", required=True) parser.add_argument("--block-size", type=int, default=16) return parser.parse_args() @@ -47,6 +48,7 @@ def materialize( requests: int, rates: list[float], block_size: int, + duration_seconds: float | None = None, ) -> dict[str, object]: values = [input_tokens, output_tokens, requests, block_size] if any(value <= 0 for value in values): @@ -57,15 +59,21 @@ def materialize( raise ValueError("rates must be positive and finite") if len(set(rates)) != len(rates): raise ValueError("rates must be unique") + if duration_seconds is not None and duration_seconds <= 0: + raise ValueError("duration must be positive") root.mkdir(parents=True, exist_ok=True) anchors = [] blocks = math.ceil(input_tokens / block_size) for rate in rates: + request_count = max( + requests, + math.ceil(rate * duration_seconds) if duration_seconds is not None else 0, + ) path = root / f"{rate_key(rate)}.csv" with path.open("w", newline="") as output: writer = csv.DictWriter(output, fieldnames=FIELDS, lineterminator="\n") writer.writeheader() - for request_id in range(requests): + for request_id in range(request_count): first_block = request_id * blocks + 1 writer.writerow( { @@ -82,6 +90,7 @@ def materialize( { "label": rate_key(rate), "rate": rate, + "requests": request_count, "path": path.name, "sha256": sha256(path), } @@ -91,7 +100,8 @@ def materialize( "contract": { "input_tokens": input_tokens, "output_tokens": output_tokens, - "requests": requests, + "minimum_requests_per_anchor": requests, + "minimum_steady_duration_seconds": duration_seconds, "block_size": block_size, "arrival": "open_loop_uniform", "prefix_caching": False, @@ -114,6 +124,7 @@ def main() -> None: requests=args.requests, rates=args.rate, block_size=args.block_size, + duration_seconds=args.duration_seconds, ) print(json.dumps(manifest, sort_keys=True)) diff --git a/runs/frontier-fidelity-envelope-v1/results/f1-pilot/README.md b/runs/frontier-fidelity-envelope-v1/results/f1-pilot/README.md new file mode 100644 index 0000000..ef9e445 --- /dev/null +++ b/runs/frontier-fidelity-envelope-v1/results/f1-pilot/README.md @@ -0,0 +1,10 @@ +# F1 64-request pilot(非 decision-bearing) + +该 surface 使用固定 `ISL=512/OSL=1`、uniform QPS、prefix off 和每个 anchor +64 requests。它完成了 runner correctness smoke,但不能用于 capacity/ranking claim:在 +512 req/s 时,64 个请求只覆盖 0.123 秒,TP4 的 64/64 SLO pass 只说明这个有限 burst +能在 TTFT SLO 内排空,并不说明系统能持续承载 512 req/s。 + +Decision-bearing F1 改为每个 anchor 至少 64 requests 且至少 10 秒 steady arrivals; +物化 contract 位于 `../../workloads/f1-isl512-osl1-steady10s/manifest.json`。原始 pilot +manifest 保留用于审计,不与 steady surface 合并。 diff --git a/runs/frontier-fidelity-envelope-v1/results/f1-pilot/frontier_trace_surface_frozen.json b/runs/frontier-fidelity-envelope-v1/results/f1-pilot/frontier_trace_surface_frozen.json new file mode 100644 index 0000000..bb21c11 --- /dev/null +++ b/runs/frontier-fidelity-envelope-v1/results/f1-pilot/frontier_trace_surface_frozen.json @@ -0,0 +1,3912 @@ +{ + "collective": { + "allreduce_csv": "/home/gahow/phd/aituner/runs/frontier-fidelity-envelope-v1/profiles/measured-allreduce.csv", + "allreduce_csv_sha256": "9d693fd406616b599e57bcde66c980c7fc2831b3acf37d3eb633cec80ea0070d", + "backend": "vidur" + }, + "config_results": [ + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "loads": [ + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "elapsed_seconds": 9.23979640007019, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "ad907fe8c9476f468f30802108ab33b012d857f0b2082a55159b4645e201980a", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 216.29366351634627, + "ttft_p95_ms": 283.4691372063174 + }, + "status": "completed", + "trace_label": "r16", + "trace_sha256": "dd00640d57184b1607fb466d2735bd4edf42e585ac6d6f2d6c6363c76b34aac8" + }, + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "elapsed_seconds": 9.139918327331543, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "63476fae7272db8d2636c94e535f4f14bafa5e8b7b0e6d2cb9b1e840dd686115", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 259.0172330518654, + "ttft_p95_ms": 347.0536850751076 + }, + "status": "completed", + "trace_label": "r32", + "trace_sha256": "de448fa237c367c4910570586f8574a5020c3186a2adc1db5daa26acf0d2b9e1" + }, + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "elapsed_seconds": 9.040833234786987, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 64.0, + "request_count": 64, + "request_metrics_sha256": "5b277eafe865f738e74074ea50ad124c7acc79e034f09c78ec4c588db7bcac95", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.796875, + "passed": 51 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.796875, + "passed": 51 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.796875, + "passed": 51 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.796875, + "passed": 51 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 749.1180253669007, + "ttft_p95_ms": 1235.420300361212 + }, + "status": "completed", + "trace_label": "r64", + "trace_sha256": "c796965d7afa1a9df4b8ba523412cef2e4dc7ddfc3cef6bce7e91265b20aaf2d" + }, + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "elapsed_seconds": 8.89000153541565, + "offered_request_rate": 128.0, + "offered_request_rate_per_gpu": 128.0, + "request_count": 64, + "request_metrics_sha256": "07bd77c8416c5ff0e9b83861669f833c126ac2fb3b0a9a6ae2830fc8e2e9041b", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 952.2430253669007, + "ttft_p95_ms": 1704.170300361212 + }, + "status": "completed", + "trace_label": "r128", + "trace_sha256": "30706d59e020ad0dcb670d367201de77ca7337b5fb41da0efd5af8c1915b2f04" + }, + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "elapsed_seconds": 9.043700695037842, + "offered_request_rate": 256.0, + "offered_request_rate_per_gpu": 256.0, + "request_count": 64, + "request_metrics_sha256": "f939698c1a72ad9c6da784797fe3044c0eaeec87f311bcf2a9b35fc699aa9877", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 1053.8055253669008, + "ttft_p95_ms": 1938.545300361212 + }, + "status": "completed", + "trace_label": "r256", + "trace_sha256": "f93c2390f084c4c60e7ca0d3096c7bdf024b7b286fd7ac72f4dc21d3d2af8cbf" + }, + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "elapsed_seconds": 9.08841609954834, + "offered_request_rate": 512.0, + "offered_request_rate_per_gpu": 512.0, + "request_count": 64, + "request_metrics_sha256": "3734b160b72632c9d6eff25ab66e2db66a33685af385f22ff779454825604871", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.390625, + "passed": 25 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.390625, + "passed": 25 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.390625, + "passed": 25 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.390625, + "passed": 25 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 1104.5867753669008, + "ttft_p95_ms": 2055.732800361212 + }, + "status": "completed", + "trace_label": "r512", + "trace_sha256": "e2612362defdb10f876eb478a642ef78a9f6101833b7cc4ff1f0b02b8a453767" + } + ] + }, + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "loads": [ + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "elapsed_seconds": 9.090706586837769, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "ad907fe8c9476f468f30802108ab33b012d857f0b2082a55159b4645e201980a", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 216.29366351634627, + "ttft_p95_ms": 283.4691372063174 + }, + "status": "completed", + "trace_label": "r16", + "trace_sha256": "dd00640d57184b1607fb466d2735bd4edf42e585ac6d6f2d6c6363c76b34aac8" + }, + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "elapsed_seconds": 9.0899977684021, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "63476fae7272db8d2636c94e535f4f14bafa5e8b7b0e6d2cb9b1e840dd686115", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 259.0172330518654, + "ttft_p95_ms": 347.0536850751076 + }, + "status": "completed", + "trace_label": "r32", + "trace_sha256": "de448fa237c367c4910570586f8574a5020c3186a2adc1db5daa26acf0d2b9e1" + }, + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "elapsed_seconds": 9.190520524978638, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 64.0, + "request_count": 64, + "request_metrics_sha256": "63a37dec7df065e7e8c8ef5ada849898c99795af0873bbe32c4120596580e435", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 0.984375, + "passed": 63 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 0.984375, + "passed": 63 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 0.984375, + "passed": 63 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 0.984375, + "passed": 63 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 733.829341199415, + "ttft_p95_ms": 1045.2671022385387 + }, + "status": "completed", + "trace_label": "r64", + "trace_sha256": "c796965d7afa1a9df4b8ba523412cef2e4dc7ddfc3cef6bce7e91265b20aaf2d" + }, + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "elapsed_seconds": 9.088061571121216, + "offered_request_rate": 128.0, + "offered_request_rate_per_gpu": 128.0, + "request_count": 64, + "request_metrics_sha256": "043cb4e37ec062487ecf8bf2910d3660fe3d5846be13214ac2a8beb668e19192", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 870.0662724508372, + "ttft_p95_ms": 1477.3167945290845 + }, + "status": "completed", + "trace_label": "r128", + "trace_sha256": "30706d59e020ad0dcb670d367201de77ca7337b5fb41da0efd5af8c1915b2f04" + }, + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "elapsed_seconds": 9.192060708999634, + "offered_request_rate": 256.0, + "offered_request_rate_per_gpu": 256.0, + "request_count": 64, + "request_metrics_sha256": "e7f32c2152c42a95203201dba11f2c561eeaf3cfc9c6e9e68d89df2294475510", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 940.3787724508372, + "ttft_p95_ms": 1680.4417945290845 + }, + "status": "completed", + "trace_label": "r256", + "trace_sha256": "f93c2390f084c4c60e7ca0d3096c7bdf024b7b286fd7ac72f4dc21d3d2af8cbf" + }, + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "elapsed_seconds": 9.089672565460205, + "offered_request_rate": 512.0, + "offered_request_rate_per_gpu": 512.0, + "request_count": 64, + "request_metrics_sha256": "d879f833e5dc08d711f48fb6c6ccf48f7891fe7694470b107747ed0a1b95dac7", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 975.5350224508372, + "ttft_p95_ms": 1782.0042945290845 + }, + "status": "completed", + "trace_label": "r512", + "trace_sha256": "e2612362defdb10f876eb478a642ef78a9f6101833b7cc4ff1f0b02b8a453767" + } + ] + }, + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "loads": [ + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "elapsed_seconds": 9.191780090332031, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "ad907fe8c9476f468f30802108ab33b012d857f0b2082a55159b4645e201980a", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 216.29366351634627, + "ttft_p95_ms": 283.4691372063174 + }, + "status": "completed", + "trace_label": "r16", + "trace_sha256": "dd00640d57184b1607fb466d2735bd4edf42e585ac6d6f2d6c6363c76b34aac8" + }, + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "elapsed_seconds": 9.189058303833008, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "63476fae7272db8d2636c94e535f4f14bafa5e8b7b0e6d2cb9b1e840dd686115", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 259.0172330518654, + "ttft_p95_ms": 347.0536850751076 + }, + "status": "completed", + "trace_label": "r32", + "trace_sha256": "de448fa237c367c4910570586f8574a5020c3186a2adc1db5daa26acf0d2b9e1" + }, + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "elapsed_seconds": 9.190616607666016, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 64.0, + "request_count": 64, + "request_metrics_sha256": "63a37dec7df065e7e8c8ef5ada849898c99795af0873bbe32c4120596580e435", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 0.984375, + "passed": 63 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 0.984375, + "passed": 63 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 0.984375, + "passed": 63 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 0.984375, + "passed": 63 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 733.829341199415, + "ttft_p95_ms": 1045.2671022385387 + }, + "status": "completed", + "trace_label": "r64", + "trace_sha256": "c796965d7afa1a9df4b8ba523412cef2e4dc7ddfc3cef6bce7e91265b20aaf2d" + }, + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "elapsed_seconds": 9.140897035598755, + "offered_request_rate": 128.0, + "offered_request_rate_per_gpu": 128.0, + "request_count": 64, + "request_metrics_sha256": "043cb4e37ec062487ecf8bf2910d3660fe3d5846be13214ac2a8beb668e19192", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 870.0662724508372, + "ttft_p95_ms": 1477.3167945290845 + }, + "status": "completed", + "trace_label": "r128", + "trace_sha256": "30706d59e020ad0dcb670d367201de77ca7337b5fb41da0efd5af8c1915b2f04" + }, + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "elapsed_seconds": 9.195079803466797, + "offered_request_rate": 256.0, + "offered_request_rate_per_gpu": 256.0, + "request_count": 64, + "request_metrics_sha256": "e7f32c2152c42a95203201dba11f2c561eeaf3cfc9c6e9e68d89df2294475510", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 940.3787724508372, + "ttft_p95_ms": 1680.4417945290845 + }, + "status": "completed", + "trace_label": "r256", + "trace_sha256": "f93c2390f084c4c60e7ca0d3096c7bdf024b7b286fd7ac72f4dc21d3d2af8cbf" + }, + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "elapsed_seconds": 9.291805744171143, + "offered_request_rate": 512.0, + "offered_request_rate_per_gpu": 512.0, + "request_count": 64, + "request_metrics_sha256": "d879f833e5dc08d711f48fb6c6ccf48f7891fe7694470b107747ed0a1b95dac7", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 975.5350224508372, + "ttft_p95_ms": 1782.0042945290845 + }, + "status": "completed", + "trace_label": "r512", + "trace_sha256": "e2612362defdb10f876eb478a642ef78a9f6101833b7cc4ff1f0b02b8a453767" + } + ] + }, + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "loads": [ + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "elapsed_seconds": 9.336015462875366, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "ad907fe8c9476f468f30802108ab33b012d857f0b2082a55159b4645e201980a", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 216.29366351634627, + "ttft_p95_ms": 283.4691372063174 + }, + "status": "completed", + "trace_label": "r16", + "trace_sha256": "dd00640d57184b1607fb466d2735bd4edf42e585ac6d6f2d6c6363c76b34aac8" + }, + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "elapsed_seconds": 9.189040660858154, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "63476fae7272db8d2636c94e535f4f14bafa5e8b7b0e6d2cb9b1e840dd686115", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 259.0172330518654, + "ttft_p95_ms": 347.0536850751076 + }, + "status": "completed", + "trace_label": "r32", + "trace_sha256": "de448fa237c367c4910570586f8574a5020c3186a2adc1db5daa26acf0d2b9e1" + }, + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "elapsed_seconds": 9.28804874420166, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 64.0, + "request_count": 64, + "request_metrics_sha256": "63a37dec7df065e7e8c8ef5ada849898c99795af0873bbe32c4120596580e435", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 0.984375, + "passed": 63 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 0.984375, + "passed": 63 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 0.984375, + "passed": 63 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 0.984375, + "passed": 63 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 733.829341199415, + "ttft_p95_ms": 1045.2671022385387 + }, + "status": "completed", + "trace_label": "r64", + "trace_sha256": "c796965d7afa1a9df4b8ba523412cef2e4dc7ddfc3cef6bce7e91265b20aaf2d" + }, + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "elapsed_seconds": 9.286862850189209, + "offered_request_rate": 128.0, + "offered_request_rate_per_gpu": 128.0, + "request_count": 64, + "request_metrics_sha256": "043cb4e37ec062487ecf8bf2910d3660fe3d5846be13214ac2a8beb668e19192", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 870.0662724508372, + "ttft_p95_ms": 1477.3167945290845 + }, + "status": "completed", + "trace_label": "r128", + "trace_sha256": "30706d59e020ad0dcb670d367201de77ca7337b5fb41da0efd5af8c1915b2f04" + }, + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "elapsed_seconds": 9.135990619659424, + "offered_request_rate": 256.0, + "offered_request_rate_per_gpu": 256.0, + "request_count": 64, + "request_metrics_sha256": "e7f32c2152c42a95203201dba11f2c561eeaf3cfc9c6e9e68d89df2294475510", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 940.3787724508372, + "ttft_p95_ms": 1680.4417945290845 + }, + "status": "completed", + "trace_label": "r256", + "trace_sha256": "f93c2390f084c4c60e7ca0d3096c7bdf024b7b286fd7ac72f4dc21d3d2af8cbf" + }, + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "elapsed_seconds": 9.236075401306152, + "offered_request_rate": 512.0, + "offered_request_rate_per_gpu": 512.0, + "request_count": 64, + "request_metrics_sha256": "d879f833e5dc08d711f48fb6c6ccf48f7891fe7694470b107747ed0a1b95dac7", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.515625, + "passed": 33 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 975.5350224508372, + "ttft_p95_ms": 1782.0042945290845 + }, + "status": "completed", + "trace_label": "r512", + "trace_sha256": "e2612362defdb10f876eb478a642ef78a9f6101833b7cc4ff1f0b02b8a453767" + } + ] + }, + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "loads": [ + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "elapsed_seconds": 9.188935279846191, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "ef8e3157f73a0421abb1d6c8ef036db2c3858b52a81c47c52bf2b24803674ab6", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 145.3083076454922, + "ttft_p95_ms": 191.94509020206718 + }, + "status": "completed", + "trace_label": "r16", + "trace_sha256": "dd00640d57184b1607fb466d2735bd4edf42e585ac6d6f2d6c6363c76b34aac8" + }, + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "elapsed_seconds": 9.238474130630493, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "b3e01bb71b1a148c2586b6f5a9cfe05bb9b31e656f4e7993ebd27a4d14d6c994", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 160.71212013197345, + "ttft_p95_ms": 215.40746023330092 + }, + "status": "completed", + "trace_label": "r32", + "trace_sha256": "de448fa237c367c4910570586f8574a5020c3186a2adc1db5daa26acf0d2b9e1" + }, + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "elapsed_seconds": 9.044495582580566, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "be6729dc2729694448aa8a59685d2a1cfd6863cdc40baa8d8aff7edc86633c63", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 342.3859670790913, + "ttft_p95_ms": 516.3686608843987 + }, + "status": "completed", + "trace_label": "r64", + "trace_sha256": "c796965d7afa1a9df4b8ba523412cef2e4dc7ddfc3cef6bce7e91265b20aaf2d" + }, + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "elapsed_seconds": 9.19176721572876, + "offered_request_rate": 128.0, + "offered_request_rate_per_gpu": 64.0, + "request_count": 64, + "request_metrics_sha256": "4a23d18f2b8a29fd0e341ab68c7499b93864e91aebf164a39a819450d728e7fc", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 568.5763491205233, + "ttft_p95_ms": 976.9340429258309 + }, + "status": "completed", + "trace_label": "r128", + "trace_sha256": "30706d59e020ad0dcb670d367201de77ca7337b5fb41da0efd5af8c1915b2f04" + }, + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "elapsed_seconds": 9.190044641494751, + "offered_request_rate": 256.0, + "offered_request_rate_per_gpu": 128.0, + "request_count": 64, + "request_metrics_sha256": "f1d8c6820f95946bcf3a4977585b99514ae35d870d2ff09bb9f7c7baef96464e", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.796875, + "passed": 51 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.796875, + "passed": 51 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.796875, + "passed": 51 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.796875, + "passed": 51 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 670.1388491205233, + "ttft_p95_ms": 1211.309042925831 + }, + "status": "completed", + "trace_label": "r256", + "trace_sha256": "f93c2390f084c4c60e7ca0d3096c7bdf024b7b286fd7ac72f4dc21d3d2af8cbf" + }, + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "elapsed_seconds": 9.139874696731567, + "offered_request_rate": 512.0, + "offered_request_rate_per_gpu": 256.0, + "request_count": 64, + "request_metrics_sha256": "307d2d128d6032dd613bc67d5f78e0d3d9c9f05e7f1a72c5751fd1cec0b589b3", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.765625, + "passed": 49 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.765625, + "passed": 49 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.765625, + "passed": 49 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.765625, + "passed": 49 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 720.9200991205233, + "ttft_p95_ms": 1328.496542925831 + }, + "status": "completed", + "trace_label": "r512", + "trace_sha256": "e2612362defdb10f876eb478a642ef78a9f6101833b7cc4ff1f0b02b8a453767" + } + ] + }, + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "loads": [ + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "elapsed_seconds": 9.338894128799438, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "ef8e3157f73a0421abb1d6c8ef036db2c3858b52a81c47c52bf2b24803674ab6", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 145.3083076454922, + "ttft_p95_ms": 191.94509020206718 + }, + "status": "completed", + "trace_label": "r16", + "trace_sha256": "dd00640d57184b1607fb466d2735bd4edf42e585ac6d6f2d6c6363c76b34aac8" + }, + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "elapsed_seconds": 9.092449188232422, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "b3e01bb71b1a148c2586b6f5a9cfe05bb9b31e656f4e7993ebd27a4d14d6c994", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 160.71212013197345, + "ttft_p95_ms": 215.40746023330092 + }, + "status": "completed", + "trace_label": "r32", + "trace_sha256": "de448fa237c367c4910570586f8574a5020c3186a2adc1db5daa26acf0d2b9e1" + }, + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "elapsed_seconds": 9.191263914108276, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "82720ff92616f5f6544e03f52ce3635328d72bd28cbfb26e18a0ad8f43e21abf", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 308.93499355358654, + "ttft_p95_ms": 484.23333144654856 + }, + "status": "completed", + "trace_label": "r64", + "trace_sha256": "c796965d7afa1a9df4b8ba523412cef2e4dc7ddfc3cef6bce7e91265b20aaf2d" + }, + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "elapsed_seconds": 9.09642219543457, + "offered_request_rate": 128.0, + "offered_request_rate_per_gpu": 64.0, + "request_count": 64, + "request_metrics_sha256": "e8dae4b0df880217f1c36c91435bf1afe920ccec1d776428611a7a49f3861d79", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 504.6855225446456, + "ttft_p95_ms": 731.2781755076838 + }, + "status": "completed", + "trace_label": "r128", + "trace_sha256": "30706d59e020ad0dcb670d367201de77ca7337b5fb41da0efd5af8c1915b2f04" + }, + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "elapsed_seconds": 9.189780473709106, + "offered_request_rate": 256.0, + "offered_request_rate_per_gpu": 128.0, + "request_count": 64, + "request_metrics_sha256": "172a6cae007a956ca9640e313a3c82b2ffbb8b19c05bb1db808d1f390e9c8cc3", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 565.5028311011398, + "ttft_p95_ms": 970.7870068870639 + }, + "status": "completed", + "trace_label": "r256", + "trace_sha256": "f93c2390f084c4c60e7ca0d3096c7bdf024b7b286fd7ac72f4dc21d3d2af8cbf" + }, + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "elapsed_seconds": 9.236796617507935, + "offered_request_rate": 512.0, + "offered_request_rate_per_gpu": 256.0, + "request_count": 64, + "request_metrics_sha256": "18b637ba04030fc15b1b39a20e3c651c829f4bd06bc908ac441ebb640653ef93", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.875, + "passed": 56 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.875, + "passed": 56 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.875, + "passed": 56 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.875, + "passed": 56 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 600.6590811011398, + "ttft_p95_ms": 1072.3495068870639 + }, + "status": "completed", + "trace_label": "r512", + "trace_sha256": "e2612362defdb10f876eb478a642ef78a9f6101833b7cc4ff1f0b02b8a453767" + } + ] + }, + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "loads": [ + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "elapsed_seconds": 9.48096251487732, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "ef8e3157f73a0421abb1d6c8ef036db2c3858b52a81c47c52bf2b24803674ab6", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 145.3083076454922, + "ttft_p95_ms": 191.94509020206718 + }, + "status": "completed", + "trace_label": "r16", + "trace_sha256": "dd00640d57184b1607fb466d2735bd4edf42e585ac6d6f2d6c6363c76b34aac8" + }, + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "elapsed_seconds": 9.192594051361084, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "b3e01bb71b1a148c2586b6f5a9cfe05bb9b31e656f4e7993ebd27a4d14d6c994", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 160.71212013197345, + "ttft_p95_ms": 215.40746023330092 + }, + "status": "completed", + "trace_label": "r32", + "trace_sha256": "de448fa237c367c4910570586f8574a5020c3186a2adc1db5daa26acf0d2b9e1" + }, + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "elapsed_seconds": 9.440584182739258, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "82720ff92616f5f6544e03f52ce3635328d72bd28cbfb26e18a0ad8f43e21abf", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 308.93499355358654, + "ttft_p95_ms": 484.23333144654856 + }, + "status": "completed", + "trace_label": "r64", + "trace_sha256": "c796965d7afa1a9df4b8ba523412cef2e4dc7ddfc3cef6bce7e91265b20aaf2d" + }, + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "elapsed_seconds": 9.340449810028076, + "offered_request_rate": 128.0, + "offered_request_rate_per_gpu": 64.0, + "request_count": 64, + "request_metrics_sha256": "e8dae4b0df880217f1c36c91435bf1afe920ccec1d776428611a7a49f3861d79", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 504.6855225446456, + "ttft_p95_ms": 731.2781755076838 + }, + "status": "completed", + "trace_label": "r128", + "trace_sha256": "30706d59e020ad0dcb670d367201de77ca7337b5fb41da0efd5af8c1915b2f04" + }, + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "elapsed_seconds": 9.138880968093872, + "offered_request_rate": 256.0, + "offered_request_rate_per_gpu": 128.0, + "request_count": 64, + "request_metrics_sha256": "172a6cae007a956ca9640e313a3c82b2ffbb8b19c05bb1db808d1f390e9c8cc3", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 565.5028311011398, + "ttft_p95_ms": 970.7870068870639 + }, + "status": "completed", + "trace_label": "r256", + "trace_sha256": "f93c2390f084c4c60e7ca0d3096c7bdf024b7b286fd7ac72f4dc21d3d2af8cbf" + }, + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "elapsed_seconds": 9.097595930099487, + "offered_request_rate": 512.0, + "offered_request_rate_per_gpu": 256.0, + "request_count": 64, + "request_metrics_sha256": "18b637ba04030fc15b1b39a20e3c651c829f4bd06bc908ac441ebb640653ef93", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.875, + "passed": 56 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.875, + "passed": 56 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.875, + "passed": 56 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.875, + "passed": 56 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 600.6590811011398, + "ttft_p95_ms": 1072.3495068870639 + }, + "status": "completed", + "trace_label": "r512", + "trace_sha256": "e2612362defdb10f876eb478a642ef78a9f6101833b7cc4ff1f0b02b8a453767" + } + ] + }, + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "loads": [ + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "elapsed_seconds": 8.998417615890503, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "ef8e3157f73a0421abb1d6c8ef036db2c3858b52a81c47c52bf2b24803674ab6", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 145.3083076454922, + "ttft_p95_ms": 191.94509020206718 + }, + "status": "completed", + "trace_label": "r16", + "trace_sha256": "dd00640d57184b1607fb466d2735bd4edf42e585ac6d6f2d6c6363c76b34aac8" + }, + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "elapsed_seconds": 9.042184114456177, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "b3e01bb71b1a148c2586b6f5a9cfe05bb9b31e656f4e7993ebd27a4d14d6c994", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 160.71212013197345, + "ttft_p95_ms": 215.40746023330092 + }, + "status": "completed", + "trace_label": "r32", + "trace_sha256": "de448fa237c367c4910570586f8574a5020c3186a2adc1db5daa26acf0d2b9e1" + }, + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "elapsed_seconds": 9.095743656158447, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "82720ff92616f5f6544e03f52ce3635328d72bd28cbfb26e18a0ad8f43e21abf", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 308.93499355358654, + "ttft_p95_ms": 484.23333144654856 + }, + "status": "completed", + "trace_label": "r64", + "trace_sha256": "c796965d7afa1a9df4b8ba523412cef2e4dc7ddfc3cef6bce7e91265b20aaf2d" + }, + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "elapsed_seconds": 9.151461124420166, + "offered_request_rate": 128.0, + "offered_request_rate_per_gpu": 64.0, + "request_count": 64, + "request_metrics_sha256": "e8dae4b0df880217f1c36c91435bf1afe920ccec1d776428611a7a49f3861d79", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 504.6855225446456, + "ttft_p95_ms": 731.2781755076838 + }, + "status": "completed", + "trace_label": "r128", + "trace_sha256": "30706d59e020ad0dcb670d367201de77ca7337b5fb41da0efd5af8c1915b2f04" + }, + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "elapsed_seconds": 9.000789642333984, + "offered_request_rate": 256.0, + "offered_request_rate_per_gpu": 128.0, + "request_count": 64, + "request_metrics_sha256": "172a6cae007a956ca9640e313a3c82b2ffbb8b19c05bb1db808d1f390e9c8cc3", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 565.5028311011398, + "ttft_p95_ms": 970.7870068870639 + }, + "status": "completed", + "trace_label": "r256", + "trace_sha256": "f93c2390f084c4c60e7ca0d3096c7bdf024b7b286fd7ac72f4dc21d3d2af8cbf" + }, + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "elapsed_seconds": 8.99738073348999, + "offered_request_rate": 512.0, + "offered_request_rate_per_gpu": 256.0, + "request_count": 64, + "request_metrics_sha256": "18b637ba04030fc15b1b39a20e3c651c829f4bd06bc908ac441ebb640653ef93", + "score": { + "slos": { + "tpot_100ms": { + "feasible": false, + "pass_rate": 0.875, + "passed": 56 + }, + "tpot_150ms": { + "feasible": false, + "pass_rate": 0.875, + "passed": 56 + }, + "tpot_180ms": { + "feasible": false, + "pass_rate": 0.875, + "passed": 56 + }, + "tpot_50ms": { + "feasible": false, + "pass_rate": 0.875, + "passed": 56 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 600.6590811011398, + "ttft_p95_ms": 1072.3495068870639 + }, + "status": "completed", + "trace_label": "r512", + "trace_sha256": "e2612362defdb10f876eb478a642ef78a9f6101833b7cc4ff1f0b02b8a453767" + } + ] + }, + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "loads": [ + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "elapsed_seconds": 9.145608186721802, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 4.0, + "request_count": 64, + "request_metrics_sha256": "cf7604290c74aa06e82791dccac8f57adbb43b861ec03601cf7353bcdb5cf85d", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 113.95995004232296, + "ttft_p95_ms": 148.6208545233651 + }, + "status": "completed", + "trace_label": "r16", + "trace_sha256": "dd00640d57184b1607fb466d2735bd4edf42e585ac6d6f2d6c6363c76b34aac8" + }, + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "elapsed_seconds": 9.04130220413208, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "667039850bf100a1214dff99ba4a6f15ff47f9d3986736dfa2460979e2e644e6", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 123.08122913633257, + "ttft_p95_ms": 155.48122649863294 + }, + "status": "completed", + "trace_label": "r32", + "trace_sha256": "de448fa237c367c4910570586f8574a5020c3186a2adc1db5daa26acf0d2b9e1" + }, + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "elapsed_seconds": 9.094721555709839, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "8fc12fe9c0c8ee1d2cae00e41bf57f36e90ecda35fed2e6c1b4fcaa52afb096b", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 132.5292253513205, + "ttft_p95_ms": 179.06372047949105 + }, + "status": "completed", + "trace_label": "r64", + "trace_sha256": "c796965d7afa1a9df4b8ba523412cef2e4dc7ddfc3cef6bce7e91265b20aaf2d" + }, + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "elapsed_seconds": 9.047569274902344, + "offered_request_rate": 128.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "a657986eb6c2f0500d902176c3344861ffca678da2498bca87bbaf0989097229", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 369.5552485735299, + "ttft_p95_ms": 599.7693411300143 + }, + "status": "completed", + "trace_label": "r128", + "trace_sha256": "30706d59e020ad0dcb670d367201de77ca7337b5fb41da0efd5af8c1915b2f04" + }, + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "elapsed_seconds": 9.150342464447021, + "offered_request_rate": 256.0, + "offered_request_rate_per_gpu": 64.0, + "request_count": 64, + "request_metrics_sha256": "7f8328f0bba2873fb721543cb20ef17fd1c43610e8adff9d26f4033157f86d92", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 471.1177485735299, + "ttft_p95_ms": 834.1443411300143 + }, + "status": "completed", + "trace_label": "r256", + "trace_sha256": "f93c2390f084c4c60e7ca0d3096c7bdf024b7b286fd7ac72f4dc21d3d2af8cbf" + }, + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "elapsed_seconds": 9.043598413467407, + "offered_request_rate": 512.0, + "offered_request_rate_per_gpu": 128.0, + "request_count": 64, + "request_metrics_sha256": "056ab2df7050c5080c7e03ba1bb9875d82fbcad2e8b1dc69b34a5c184fd89cb1", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 521.89899857353, + "ttft_p95_ms": 951.3318411300143 + }, + "status": "completed", + "trace_label": "r512", + "trace_sha256": "e2612362defdb10f876eb478a642ef78a9f6101833b7cc4ff1f0b02b8a453767" + } + ] + }, + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "loads": [ + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "elapsed_seconds": 9.041609525680542, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 4.0, + "request_count": 64, + "request_metrics_sha256": "cf7604290c74aa06e82791dccac8f57adbb43b861ec03601cf7353bcdb5cf85d", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 113.95995004232296, + "ttft_p95_ms": 148.6208545233651 + }, + "status": "completed", + "trace_label": "r16", + "trace_sha256": "dd00640d57184b1607fb466d2735bd4edf42e585ac6d6f2d6c6363c76b34aac8" + }, + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "elapsed_seconds": 9.040033102035522, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "667039850bf100a1214dff99ba4a6f15ff47f9d3986736dfa2460979e2e644e6", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 123.08122913633257, + "ttft_p95_ms": 155.48122649863294 + }, + "status": "completed", + "trace_label": "r32", + "trace_sha256": "de448fa237c367c4910570586f8574a5020c3186a2adc1db5daa26acf0d2b9e1" + }, + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "elapsed_seconds": 9.14271068572998, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "8fc12fe9c0c8ee1d2cae00e41bf57f36e90ecda35fed2e6c1b4fcaa52afb096b", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 132.5292253513205, + "ttft_p95_ms": 179.06372047949105 + }, + "status": "completed", + "trace_label": "r64", + "trace_sha256": "c796965d7afa1a9df4b8ba523412cef2e4dc7ddfc3cef6bce7e91265b20aaf2d" + }, + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "elapsed_seconds": 9.04430103302002, + "offered_request_rate": 128.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "a1ad1f58cce44b8bc7b72b2997413758c4d40e971ce51fc68f9530bb2279d281", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 302.6860632179096, + "ttft_p95_ms": 411.8879773090427 + }, + "status": "completed", + "trace_label": "r128", + "trace_sha256": "30706d59e020ad0dcb670d367201de77ca7337b5fb41da0efd5af8c1915b2f04" + }, + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "elapsed_seconds": 8.998274326324463, + "offered_request_rate": 256.0, + "offered_request_rate_per_gpu": 64.0, + "request_count": 64, + "request_metrics_sha256": "6271e503f9158379f1b1043acab957d96c452adb195e45c63b311667677dfd75", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 381.3555701081787, + "ttft_p95_ms": 623.3699841993117 + }, + "status": "completed", + "trace_label": "r256", + "trace_sha256": "f93c2390f084c4c60e7ca0d3096c7bdf024b7b286fd7ac72f4dc21d3d2af8cbf" + }, + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "elapsed_seconds": 9.042869806289673, + "offered_request_rate": 512.0, + "offered_request_rate_per_gpu": 128.0, + "request_count": 64, + "request_metrics_sha256": "1bddc1f7a64764acb948087958ba9c20ed08b818545ccde67a3cec4cb48a9c35", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 416.5118201081787, + "ttft_p95_ms": 724.9324841993117 + }, + "status": "completed", + "trace_label": "r512", + "trace_sha256": "e2612362defdb10f876eb478a642ef78a9f6101833b7cc4ff1f0b02b8a453767" + } + ] + }, + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "loads": [ + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "elapsed_seconds": 9.03968071937561, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 4.0, + "request_count": 64, + "request_metrics_sha256": "cf7604290c74aa06e82791dccac8f57adbb43b861ec03601cf7353bcdb5cf85d", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 113.95995004232296, + "ttft_p95_ms": 148.6208545233651 + }, + "status": "completed", + "trace_label": "r16", + "trace_sha256": "dd00640d57184b1607fb466d2735bd4edf42e585ac6d6f2d6c6363c76b34aac8" + }, + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "elapsed_seconds": 9.040271520614624, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "667039850bf100a1214dff99ba4a6f15ff47f9d3986736dfa2460979e2e644e6", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 123.08122913633257, + "ttft_p95_ms": 155.48122649863294 + }, + "status": "completed", + "trace_label": "r32", + "trace_sha256": "de448fa237c367c4910570586f8574a5020c3186a2adc1db5daa26acf0d2b9e1" + }, + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "elapsed_seconds": 9.089779615402222, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "8fc12fe9c0c8ee1d2cae00e41bf57f36e90ecda35fed2e6c1b4fcaa52afb096b", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 132.5292253513205, + "ttft_p95_ms": 179.06372047949105 + }, + "status": "completed", + "trace_label": "r64", + "trace_sha256": "c796965d7afa1a9df4b8ba523412cef2e4dc7ddfc3cef6bce7e91265b20aaf2d" + }, + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "elapsed_seconds": 9.046946287155151, + "offered_request_rate": 128.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "a1ad1f58cce44b8bc7b72b2997413758c4d40e971ce51fc68f9530bb2279d281", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 302.6860632179096, + "ttft_p95_ms": 411.8879773090427 + }, + "status": "completed", + "trace_label": "r128", + "trace_sha256": "30706d59e020ad0dcb670d367201de77ca7337b5fb41da0efd5af8c1915b2f04" + }, + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "elapsed_seconds": 9.149730682373047, + "offered_request_rate": 256.0, + "offered_request_rate_per_gpu": 64.0, + "request_count": 64, + "request_metrics_sha256": "6271e503f9158379f1b1043acab957d96c452adb195e45c63b311667677dfd75", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 381.3555701081787, + "ttft_p95_ms": 623.3699841993117 + }, + "status": "completed", + "trace_label": "r256", + "trace_sha256": "f93c2390f084c4c60e7ca0d3096c7bdf024b7b286fd7ac72f4dc21d3d2af8cbf" + }, + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "elapsed_seconds": 9.099130392074585, + "offered_request_rate": 512.0, + "offered_request_rate_per_gpu": 128.0, + "request_count": 64, + "request_metrics_sha256": "1bddc1f7a64764acb948087958ba9c20ed08b818545ccde67a3cec4cb48a9c35", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 416.5118201081787, + "ttft_p95_ms": 724.9324841993117 + }, + "status": "completed", + "trace_label": "r512", + "trace_sha256": "e2612362defdb10f876eb478a642ef78a9f6101833b7cc4ff1f0b02b8a453767" + } + ] + }, + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "loads": [ + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "elapsed_seconds": 9.091226577758789, + "offered_request_rate": 16.0, + "offered_request_rate_per_gpu": 4.0, + "request_count": 64, + "request_metrics_sha256": "cf7604290c74aa06e82791dccac8f57adbb43b861ec03601cf7353bcdb5cf85d", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 113.95995004232296, + "ttft_p95_ms": 148.6208545233651 + }, + "status": "completed", + "trace_label": "r16", + "trace_sha256": "dd00640d57184b1607fb466d2735bd4edf42e585ac6d6f2d6c6363c76b34aac8" + }, + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "elapsed_seconds": 9.049186944961548, + "offered_request_rate": 32.0, + "offered_request_rate_per_gpu": 8.0, + "request_count": 64, + "request_metrics_sha256": "667039850bf100a1214dff99ba4a6f15ff47f9d3986736dfa2460979e2e644e6", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 123.08122913633257, + "ttft_p95_ms": 155.48122649863294 + }, + "status": "completed", + "trace_label": "r32", + "trace_sha256": "de448fa237c367c4910570586f8574a5020c3186a2adc1db5daa26acf0d2b9e1" + }, + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "elapsed_seconds": 9.14399766921997, + "offered_request_rate": 64.0, + "offered_request_rate_per_gpu": 16.0, + "request_count": 64, + "request_metrics_sha256": "8fc12fe9c0c8ee1d2cae00e41bf57f36e90ecda35fed2e6c1b4fcaa52afb096b", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 132.5292253513205, + "ttft_p95_ms": 179.06372047949105 + }, + "status": "completed", + "trace_label": "r64", + "trace_sha256": "c796965d7afa1a9df4b8ba523412cef2e4dc7ddfc3cef6bce7e91265b20aaf2d" + }, + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "elapsed_seconds": 9.000115871429443, + "offered_request_rate": 128.0, + "offered_request_rate_per_gpu": 32.0, + "request_count": 64, + "request_metrics_sha256": "a1ad1f58cce44b8bc7b72b2997413758c4d40e971ce51fc68f9530bb2279d281", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 302.6860632179096, + "ttft_p95_ms": 411.8879773090427 + }, + "status": "completed", + "trace_label": "r128", + "trace_sha256": "30706d59e020ad0dcb670d367201de77ca7337b5fb41da0efd5af8c1915b2f04" + }, + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "elapsed_seconds": 9.093112468719482, + "offered_request_rate": 256.0, + "offered_request_rate_per_gpu": 64.0, + "request_count": 64, + "request_metrics_sha256": "6271e503f9158379f1b1043acab957d96c452adb195e45c63b311667677dfd75", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 381.3555701081787, + "ttft_p95_ms": 623.3699841993117 + }, + "status": "completed", + "trace_label": "r256", + "trace_sha256": "f93c2390f084c4c60e7ca0d3096c7bdf024b7b286fd7ac72f4dc21d3d2af8cbf" + }, + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "elapsed_seconds": 9.14099645614624, + "offered_request_rate": 512.0, + "offered_request_rate_per_gpu": 128.0, + "request_count": 64, + "request_metrics_sha256": "1bddc1f7a64764acb948087958ba9c20ed08b818545ccde67a3cec4cb48a9c35", + "score": { + "slos": { + "tpot_100ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_150ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_180ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + }, + "tpot_50ms": { + "feasible": true, + "pass_rate": 1.0, + "passed": 64 + } + }, + "tpot_p50_ms": null, + "tpot_p95_ms": null, + "ttft_p50_ms": 416.5118201081787, + "ttft_p95_ms": 724.9324841993117 + }, + "status": "completed", + "trace_label": "r512", + "trace_sha256": "e2612362defdb10f876eb478a642ef78a9f6101833b7cc4ff1f0b02b8a453767" + } + ] + } + ], + "contract": { + "arrival": "original_trace_timestamp_and_order", + "input_output": "exact_source_values", + "prefix_caching": false, + "primary_tpot_slo_ms": 150.0, + "rate_contract": "uniform-spacing", + "target_pass_rate": 0.95, + "tpot_slos_ms": [ + 50.0, + 100.0, + 150.0, + 180.0 + ], + "ttft_slo": "1000ms + 1000ms * input_tokens / 8000", + "window_seconds": null + }, + "frontier": { + "git_head": "d9cfeb6d8791fbf2f295dd9744c56a666171776e", + "git_status_short": " M frontier/cc_backend/backends/vidur_cc_backend.py\n M frontier/config/config.py\n M frontier/entities/request.py\n M frontier/events/cluster_schedule_event.py\n M frontier/execution_time_predictor/sklearn_execution_time_predictor.py\n M frontier/metrics/constants.py\n M frontier/metrics/metrics_store.py\n M frontier/profiling/common/layers/rotary_embedding.py\n M frontier/profiling/moe/moe_impl.py\n M frontier/profiling/moe/moe_vllm_kernel.py\n M frontier/scheduler/cluster_scheduler/__init__.py\n M frontier/scheduler/cluster_scheduler/base_cluster_scheduler.py\n M frontier/scheduler/cluster_scheduler/cluster_scheduler_registry.py\n M frontier/scheduler/cluster_scheduler/sticky_lor_cluster_scheduler.py\n M frontier/scheduler/replica_scheduler/base_replica_scheduler.py\n M frontier/scheduler/replica_scheduler/vllm_v1_engine_replica_scheduler.py\n M frontier/scheduler/replica_stage_scheduler/replica_stage_schduler.py\n M frontier/simulator.py\n M frontier/types/cluster_scheduler_type.py\n?? data/profiling/compute/h20/\n?? frontier/scheduler/cluster_scheduler/prefix_lor_cluster_scheduler.py\n?? runs/\n?? tests/unit/test_attn_prefill_prediction_fallback.py\n?? tests/unit/test_vidur_cc_large_payload.py\n", + "source": "/tmp/replayserve-frontier-rs1b" + }, + "profiles": { + "coverage": { + "attention": { + "1": { + "exact_prefill_2048_rows": 1, + "profile_batch_size": 1 + }, + "2": { + "exact_prefill_2048_rows": 1, + "profile_batch_size": 1 + }, + "4": { + "exact_prefill_2048_rows": 1, + "profile_batch_size": 1 + } + }, + "manifest": { + "attention_tp_coverage": [ + 1, + 2, + 4 + ], + "environment_contract": { + "dtype": "bfloat16", + "frontier_commit": "d9cfeb6d8791fbf2f295dd9744c56a666171776e", + "hardware": "NVIDIA H20", + "model": "Qwen3-30B-A3B", + "tensor_parallel_sizes": [ + 1, + 2, + 4 + ], + "vllm_source_commit": "88d34c6409e9fb3c7b8ca0c04756f061d2099eb1", + "vllm_version": "0.20.0" + }, + "inputs": { + "/home/gahow/phd/aituner/runs/frontier-fidelity-envelope-v1/fleet-artifacts/qwen30-attention-composition-tp1-20260717-v1-20260717T014550833983Z/artifacts/artifacts/attention-composition-tp1-v1/raw/flashattn-composition-tp1.json": "ce461be92a5059e2d168b2cbf21b55d9a55ee71c3b0586a3e63cd8288938168d", + "/home/gahow/phd/aituner/runs/frontier-fidelity-envelope-v1/fleet-artifacts/qwen30-attention-composition-tp2-20260717-v1-20260717T014551831545Z/artifacts/artifacts/attention-composition-tp2-v1/raw/flashattn-composition-tp2.json": "997d67fb3a28b488c3aa4945ae7c6d4ecf8a08726ee3cb4652d16c221ff395ad", + "/home/gahow/phd/aituner/runs/frontier-fidelity-envelope-v1/fleet-artifacts/qwen30-attention-composition-tp4-20260717-v1-20260717T014552897206Z/artifacts/artifacts/attention-composition-tp4-v1/raw/flashattn-composition-tp4.json": "046cffd4c11d8a36513167ec25d3e1ca016324715679ea688a58d4a71dfe62b2", + "/home/gahow/phd/aituner/runs/frontier-qwen30-vllm020-profile-v1/fleet-artifacts/qwen30-vllm020-allreduce-full-tp2-20260716-v1-dispatch-aware-20260716T140743025781Z/artifacts/artifacts/allreduce-full-tp2-v1/raw/allreduce-tp2.json": "97c3c76b5a04e95bd9192423c2b891667c668f39cc0dfecbd097d749939f2d0a", + "/home/gahow/phd/aituner/runs/frontier-qwen30-vllm020-profile-v1/fleet-artifacts/qwen30-vllm020-allreduce-full-tp4-20260716-v1-dispatch-aware-20260716T141106009788Z/artifacts/artifacts/allreduce-full-tp4-v1/raw/allreduce-tp4.json": "809df9baa6f468cf12bf0c99827475acc67894dd9f3f948976590b665fac0e76", + "/home/gahow/phd/aituner/runs/frontier-qwen30-vllm020-profile-v1/fleet-artifacts/qwen30-vllm020-flashattn-kv-full-tp1-20260716-v2-20260716T135132587012Z/artifacts/artifacts/flashattn-kv-full-v2-tp1/raw/flashattn-tp1.json": "dcb4c1bf7e76b9c765f78ddd2b8a734f2d7ba2adac13ce017689a8a77fe69a27", + "/home/gahow/phd/aituner/runs/frontier-qwen30-vllm020-profile-v1/fleet-artifacts/qwen30-vllm020-flashattn-kv-full-tp2-20260716-v2-20260716T135134194295Z/artifacts/artifacts/flashattn-kv-full-v2-tp2/raw/flashattn-tp2.json": "43ce042556ba887c8860614b43ccf0f564e5cebc1a0cffbce299d0acb9fa8d07", + "/home/gahow/phd/aituner/runs/frontier-qwen30-vllm020-profile-v1/fleet-artifacts/qwen30-vllm020-flashattn-kv-full-tp4-20260716-v2-20260716T135135197200Z/artifacts/artifacts/flashattn-kv-full-v2-tp4/raw/flashattn-tp4.json": "84eef31bcad0f556907a093318a420959d14fdc94474823d11f659704bdfec73", + "/home/gahow/phd/aituner/runs/frontier-qwen30-vllm020-profile-v1/fleet-artifacts/qwen30-vllm020-frontier-linear-full-20260716-v2-max-tokens-20260716T144444676943Z/artifacts/artifacts/frontier-linear-full-v2/profiles/compute/h20/qwen3-a3b-30b-moe/linear_op.csv": "67666cb0a4901b74599d468df2e31bcaa2a11a7842cc0cefba24ffce62508e0c", + "/home/gahow/phd/aituner/runs/frontier-qwen30-vllm020-profile-v1/fleet-artifacts/qwen30-vllm020-moe-full-20260716-v1-local-shard-20260716T141334565164Z/artifacts/artifacts/moe-full-v1/raw/moe-full.json": "588f6ad0d69c9636d1b852e3df0a12d13cfe731f050ea7ec7aea457cceefbde8", + "/home/gahow/phd/aituner/runs/frontier-qwen30-vllm020-profile-v1/fleet-artifacts/qwen30-vllm020-router-full-20260716-v3-tp-context-20260716T145446098505Z/artifacts/artifacts/router-full-v3/raw/router.json": "1962972e983bff3e06a721ef4ae4ec65728ff669681497a4a7e7f769b88b4931" + }, + "outputs": { + "allreduce.json": "b38d14f990578d668523d25b107aceed433da5020d8ada3b6e44d3562261a3b3", + "attention.csv": "03180eec28087a301c9d8c185874c50bdb28b31e5e9af35fcce5a6a5a9e62147", + "attention_true_mixed_fused.csv": "cccf76e6cb6a9359f188b07d4dc2942442568a6e8cf3ff3d77ab3d35a052cecb", + "linear_op.csv": "67666cb0a4901b74599d468df2e31bcaa2a11a7842cc0cefba24ffce62508e0c", + "moe.csv": "dd698113975e5ed77f51dfd8a69aebf764ed651c3f6f34b71914d25d6e001415" + }, + "profile_id": "qwen3-30b-a3b-bf16-vllm020-h20-tp1-2-4-fused-mixed-total-conserving-pure-prefill-batch-composition", + "projection_contract": { + "allreduce": "Frozen exact runtime measurements; base profile-only comparison keeps the historical Frontier CC backend fixed to isolate compute profile fidelity", + "attention": "Pure prefill/extend/decode FA3 core plus separately measured KV update; input/output reshape assumed zero; exported mean is used as median target; true mixed rows use a total-conserving compatibility projection", + "attention_pure_prefill_batch_composition": "Direct FA3 measurements for 2/4 requests at query length 2048 and 2/4/8/16 requests at query length 512 for each TP; included only when batch-composition attention inputs are supplied", + "attention_true_mixed": "The directly measured fused total is preserved in diagnostics. Frontier's two targets are projected by the same-TP pure prefill/decode reference ratio, with projected prefill + decode exactly equal to the fused total; the split is a schema compatibility attribution, not an observation", + "linear": "Frontier profiler using vLLM 0.20 CUDA operators", + "moe": "Replicated gate and fused top-k plus TP-local modular expert kernel; expert measurement already includes prepare/finalize so shuffling is zero" + }, + "row_counts": { + "allreduce": 24, + "attention_frontier_compatible": 150, + "attention_true_mixed_fused_diagnostic": 30, + "linear": 36, + "moe": 72 + }, + "schema_version": "frontier_qwen30_vllm020_frozen_profile.v3" + } + }, + "root": "/home/gahow/phd/aituner/runs/frontier-fidelity-envelope-v1/profiles/profile-v3-batch-final", + "sha256": { + "attention": "03180eec28087a301c9d8c185874c50bdb28b31e5e9af35fcce5a6a5a9e62147", + "linear": "67666cb0a4901b74599d468df2e31bcaa2a11a7842cc0cefba24ffce62508e0c", + "manifest": "70aecb6d70f9e509e9816b1a0929163b9505760c2ac30bbd4cd8e51badfb7635", + "moe": "dd698113975e5ed77f51dfd8a69aebf764ed651c3f6f34b71914d25d6e001415" + } + }, + "rankings": { + "tpot_100ms": [ + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 256.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": false + }, + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 256.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": false + }, + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 256.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": false + }, + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 512.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": true + }, + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 512.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": true + }, + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 512.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": true + }, + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 512.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": true + }, + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 64.0, + "maximum_tested_feasible_request_rate_per_gpu": 64.0, + "upper_censored": false + }, + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 64.0, + "maximum_tested_feasible_request_rate_per_gpu": 64.0, + "upper_censored": false + }, + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 64.0, + "maximum_tested_feasible_request_rate_per_gpu": 64.0, + "upper_censored": false + }, + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 128.0, + "maximum_tested_feasible_request_rate_per_gpu": 64.0, + "upper_censored": false + }, + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 32.0, + "maximum_tested_feasible_request_rate_per_gpu": 32.0, + "upper_censored": false + } + ], + "tpot_150ms": [ + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 256.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": false + }, + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 256.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": false + }, + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 256.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": false + }, + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 512.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": true + }, + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 512.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": true + }, + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 512.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": true + }, + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 512.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": true + }, + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 64.0, + "maximum_tested_feasible_request_rate_per_gpu": 64.0, + "upper_censored": false + }, + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 64.0, + "maximum_tested_feasible_request_rate_per_gpu": 64.0, + "upper_censored": false + }, + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 64.0, + "maximum_tested_feasible_request_rate_per_gpu": 64.0, + "upper_censored": false + }, + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 128.0, + "maximum_tested_feasible_request_rate_per_gpu": 64.0, + "upper_censored": false + }, + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 32.0, + "maximum_tested_feasible_request_rate_per_gpu": 32.0, + "upper_censored": false + } + ], + "tpot_180ms": [ + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 256.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": false + }, + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 256.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": false + }, + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 256.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": false + }, + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 512.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": true + }, + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 512.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": true + }, + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 512.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": true + }, + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 512.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": true + }, + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 64.0, + "maximum_tested_feasible_request_rate_per_gpu": 64.0, + "upper_censored": false + }, + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 64.0, + "maximum_tested_feasible_request_rate_per_gpu": 64.0, + "upper_censored": false + }, + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 64.0, + "maximum_tested_feasible_request_rate_per_gpu": 64.0, + "upper_censored": false + }, + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 128.0, + "maximum_tested_feasible_request_rate_per_gpu": 64.0, + "upper_censored": false + }, + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 32.0, + "maximum_tested_feasible_request_rate_per_gpu": 32.0, + "upper_censored": false + } + ], + "tpot_50ms": [ + { + "config": { + "mns": 16, + "name": "tp2_mns16", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 256.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": false + }, + { + "config": { + "mns": 32, + "name": "tp2_mns32", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 256.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": false + }, + { + "config": { + "mns": 64, + "name": "tp2_mns64", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 256.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": false + }, + { + "config": { + "mns": 16, + "name": "tp4_mns16", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 512.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": true + }, + { + "config": { + "mns": 32, + "name": "tp4_mns32", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 512.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": true + }, + { + "config": { + "mns": 64, + "name": "tp4_mns64", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 512.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": true + }, + { + "config": { + "mns": 8, + "name": "tp4_mns8", + "tp": 4 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 512.0, + "maximum_tested_feasible_request_rate_per_gpu": 128.0, + "upper_censored": true + }, + { + "config": { + "mns": 16, + "name": "tp1_mns16", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 64.0, + "maximum_tested_feasible_request_rate_per_gpu": 64.0, + "upper_censored": false + }, + { + "config": { + "mns": 32, + "name": "tp1_mns32", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 64.0, + "maximum_tested_feasible_request_rate_per_gpu": 64.0, + "upper_censored": false + }, + { + "config": { + "mns": 64, + "name": "tp1_mns64", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 64.0, + "maximum_tested_feasible_request_rate_per_gpu": 64.0, + "upper_censored": false + }, + { + "config": { + "mns": 8, + "name": "tp2_mns8", + "tp": 2 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 128.0, + "maximum_tested_feasible_request_rate_per_gpu": 64.0, + "upper_censored": false + }, + { + "config": { + "mns": 8, + "name": "tp1_mns8", + "tp": 1 + }, + "lower_censored": false, + "maximum_tested_feasible_request_rate": 32.0, + "maximum_tested_feasible_request_rate_per_gpu": 32.0, + "upper_censored": false + } + ] + }, + "schema": "frontier-qwen30-exact-trace-surface-v1", + "status": "frozen_before_real", + "traces": [ + { + "first_arrival_s": 0.0, + "label": "r16", + "last_arrival_s": 3.9375, + "offered_request_rate": 16.0, + "path": "/home/gahow/phd/aituner/runs/frontier-fidelity-envelope-v1/workloads/f1-isl512-osl1/r16.csv", + "requests": 64, + "sha256": "dd00640d57184b1607fb466d2735bd4edf42e585ac6d6f2d6c6363c76b34aac8" + }, + { + "first_arrival_s": 0.0, + "label": "r32", + "last_arrival_s": 1.96875, + "offered_request_rate": 32.0, + "path": "/home/gahow/phd/aituner/runs/frontier-fidelity-envelope-v1/workloads/f1-isl512-osl1/r32.csv", + "requests": 64, + "sha256": "de448fa237c367c4910570586f8574a5020c3186a2adc1db5daa26acf0d2b9e1" + }, + { + "first_arrival_s": 0.0, + "label": "r64", + "last_arrival_s": 0.984375, + "offered_request_rate": 64.0, + "path": "/home/gahow/phd/aituner/runs/frontier-fidelity-envelope-v1/workloads/f1-isl512-osl1/r64.csv", + "requests": 64, + "sha256": "c796965d7afa1a9df4b8ba523412cef2e4dc7ddfc3cef6bce7e91265b20aaf2d" + }, + { + "first_arrival_s": 0.0, + "label": "r128", + "last_arrival_s": 0.4921875, + "offered_request_rate": 128.0, + "path": "/home/gahow/phd/aituner/runs/frontier-fidelity-envelope-v1/workloads/f1-isl512-osl1/r128.csv", + "requests": 64, + "sha256": "30706d59e020ad0dcb670d367201de77ca7337b5fb41da0efd5af8c1915b2f04" + }, + { + "first_arrival_s": 0.0, + "label": "r256", + "last_arrival_s": 0.24609375, + "offered_request_rate": 256.0, + "path": "/home/gahow/phd/aituner/runs/frontier-fidelity-envelope-v1/workloads/f1-isl512-osl1/r256.csv", + "requests": 64, + "sha256": "f93c2390f084c4c60e7ca0d3096c7bdf024b7b286fd7ac72f4dc21d3d2af8cbf" + }, + { + "first_arrival_s": 0.0, + "label": "r512", + "last_arrival_s": 0.123046875, + "offered_request_rate": 512.0, + "path": "/home/gahow/phd/aituner/runs/frontier-fidelity-envelope-v1/workloads/f1-isl512-osl1/r512.csv", + "requests": 64, + "sha256": "e2612362defdb10f876eb478a642ef78a9f6101833b7cc4ff1f0b02b8a453767" + } + ] +} diff --git a/runs/frontier-fidelity-envelope-v1/test_fidelity_envelope.py b/runs/frontier-fidelity-envelope-v1/test_fidelity_envelope.py index 20f4e32..4d8d224 100644 --- a/runs/frontier-fidelity-envelope-v1/test_fidelity_envelope.py +++ b/runs/frontier-fidelity-envelope-v1/test_fidelity_envelope.py @@ -203,6 +203,16 @@ class FidelityEnvelopeTest(unittest.TestCase): self.assertEqual(rows[1]["block_hash_ids"], "3|4") self.assertEqual(rows[1]["arrived_at"], "0.500000000000") self.assertEqual(manifest["contract"]["prefix_caching"], False) + steady = module.materialize( + root / "steady", + input_tokens=32, + output_tokens=4, + requests=2, + rates=[2.0], + block_size=16, + duration_seconds=10.0, + ) + self.assertEqual(steady["anchors"][0]["requests"], 20) if __name__ == "__main__": diff --git a/runs/frontier-fidelity-envelope-v1/workloads/f1-isl512-osl1-steady10s/manifest.json b/runs/frontier-fidelity-envelope-v1/workloads/f1-isl512-osl1-steady10s/manifest.json new file mode 100644 index 0000000..744a672 --- /dev/null +++ b/runs/frontier-fidelity-envelope-v1/workloads/f1-isl512-osl1-steady10s/manifest.json @@ -0,0 +1,57 @@ +{ + "anchors": [ + { + "label": "r16", + "path": "r16.csv", + "rate": 16.0, + "requests": 160, + "sha256": "931604bdf085b1e0c2c70ffed9d10cbaeb6bfb3523976ce25c9bb3faf5fb20a9" + }, + { + "label": "r32", + "path": "r32.csv", + "rate": 32.0, + "requests": 320, + "sha256": "3f839823181c75149184df43326df4139c1f00cbc607dd305323d34e92467140" + }, + { + "label": "r64", + "path": "r64.csv", + "rate": 64.0, + "requests": 640, + "sha256": "b5bea5cf977f6a875a863a135eca70e643653b07ab92233f933f8b5e44e4a8c7" + }, + { + "label": "r128", + "path": "r128.csv", + "rate": 128.0, + "requests": 1280, + "sha256": "ea94212548621233172e430649e523f494064d035a7609d0b9ec3b1930222a3b" + }, + { + "label": "r256", + "path": "r256.csv", + "rate": 256.0, + "requests": 2560, + "sha256": "1e580f6ca30317f1f1ff81a86f318347341607cf1afcfef7ba53123335f02295" + }, + { + "label": "r512", + "path": "r512.csv", + "rate": 512.0, + "requests": 5120, + "sha256": "992ad52e6e6c08bcb0843d59ff7d529d637ad8ed812aca96d1318cb1ecfeb71b" + } + ], + "contract": { + "arrival": "open_loop_uniform", + "block_size": 16, + "input_tokens": 512, + "minimum_requests_per_anchor": 64, + "minimum_steady_duration_seconds": 10.0, + "output_tokens": 1, + "prefix_caching": false, + "prefix_relation": "all request block identities are disjoint" + }, + "schema": "qwen30-fixed-frontier-traces-v1" +}