{ "title": "DeepSeek V4.1 Flash on one AMD Instinct MI350P", "revision": "2dd8058eab24484c9e2a54884244d68c825ccd7c", "scope": "One MI350P, configured GPU power limit 450 W, not measured consumption. Paiton proprietary runtime with CPU expert offload and host memory. No outside-engine ranking. GPU power excludes host CPUs, memory, PSU and cooling.", "sources": [ { "file": "README.md", "sha256": "754371cb8e7aeccb5f866e986a996cf8c4f40362b85602220516191d38203f63", "url": "https://huggingface.co/EliovpAI/DeepSeek-V4.1-Flash-Paiton-CDNA4/resolve/2dd8058eab24484c9e2a54884244d68c825ccd7c/README.md" }, { "file": "proof/SHA256SUMS", "sha256": "2120be43870e138de923d669b2d5a3151e1b3a8b2d43291b74e5af0e79fac7a6", "url": "https://huggingface.co/EliovpAI/DeepSeek-V4.1-Flash-Paiton-CDNA4/resolve/2dd8058eab24484c9e2a54884244d68c825ccd7c/proof/SHA256SUMS" }, { "file": "proof/interactive/kv-capacity.txt", "sha256": "487deb327315ff2ab4b03a7fa0eb8ea459fb0c5fcce6455408af63375ce91bad", "url": "https://huggingface.co/EliovpAI/DeepSeek-V4.1-Flash-Paiton-CDNA4/resolve/2dd8058eab24484c9e2a54884244d68c825ccd7c/proof/interactive/kv-capacity.txt" }, { "file": "proof/factory/kv-capacity.txt", "sha256": "2a86bcbb22dfebedfaf948556383becd948662926223baba9613230819dbebbf", "url": "https://huggingface.co/EliovpAI/DeepSeek-V4.1-Flash-Paiton-CDNA4/resolve/2dd8058eab24484c9e2a54884244d68c825ccd7c/proof/factory/kv-capacity.txt" }, { "file": "proof/interactive/lmx/inference-matrices/deepseek-ai-DeepSeek-V4.1-Flash-1017/matrix-20261102T130034Z.json", "sha256": "3e553a7067bd9926ded63c4df379eb06e264411d53cd517985c3ec1fb7b2b446", "url": "https://huggingface.co/EliovpAI/DeepSeek-V4.1-Flash-Paiton-CDNA4/resolve/2dd8058eab24484c9e2a54884244d68c825ccd7c/proof/interactive/lmx/inference-matrices/deepseek-ai-DeepSeek-V4.1-Flash-1017/matrix-20261102T130034Z.json" }, { "file": "proof/accuracy/interactive/rp_chat_kl.json", "sha256": "e6e37ff44d0c3991d9ff79eca1e0cd635868ee9a17699d0d8949f82eda6d1b33", "url": "https://huggingface.co/EliovpAI/DeepSeek-V4.1-Flash-Paiton-CDNA4/resolve/2dd8058eab24484c9e2a54884244d68c825ccd7c/proof/accuracy/interactive/rp_chat_kl.json" }, { "file": "proof/accuracy/interactive/long_ds1_kl.json", "sha256": "e87a8f8537e91fc8b9536a6442b39bad1729010d9dd4ba8542f889668ab1fea9", "url": "https://huggingface.co/EliovpAI/DeepSeek-V4.1-Flash-Paiton-CDNA4/resolve/2dd8058eab24484c9e2a54884244d68c825ccd7c/proof/accuracy/interactive/long_ds1_kl.json" }, { "file": "proof/accuracy/factory/rp_chat_kl.json", "sha256": "dfed407a611e64dad773df205b4cf5b98e8dd7ecdefa9c559235c28198a83656", "url": "https://huggingface.co/EliovpAI/DeepSeek-V4.1-Flash-Paiton-CDNA4/resolve/2dd8058eab24484c9e2a54884244d68c825ccd7c/proof/accuracy/factory/rp_chat_kl.json" }, { "file": "proof/accuracy/factory/long_ds1_kl.json", "sha256": "f0fae2f22ed9e635654926e6de62ebddaebb2ce91f6cb526d7180a4c0eebe852", "url": "https://huggingface.co/EliovpAI/DeepSeek-V4.1-Flash-Paiton-CDNA4/resolve/2dd8058eab24484c9e2a54884244d68c825ccd7c/proof/accuracy/factory/long_ds1_kl.json" }, { "file": "proof/interactive/sweep-int-a/cells.jsonl", "sha256": "7d90d6378ce1b6319db02faf029c6bd6f1dd10164e218c9e5c74024f5c0ce691", "url": "https://huggingface.co/EliovpAI/DeepSeek-V4.1-Flash-Paiton-CDNA4/resolve/2dd8058eab24484c9e2a54884244d68c825ccd7c/proof/interactive/sweep-int-a/cells.jsonl" }, { "file": "proof/interactive/sweep-int-b/cells.jsonl", "sha256": "a2e250be01c5637d3b76951aa00b7b2e9a989c443e5bfebc064b5ee323a7b5e3", "url": "https://huggingface.co/EliovpAI/DeepSeek-V4.1-Flash-Paiton-CDNA4/resolve/2dd8058eab24484c9e2a54884244d68c825ccd7c/proof/interactive/sweep-int-b/cells.jsonl" }, { "file": "proof/factory/sweep-fac-a/cells.jsonl", "sha256": "385e6ab3e526918254b45b1e0c56c78e2fb40bbbd3b91bf550e547c5add1832f", "url": "https://huggingface.co/EliovpAI/DeepSeek-V4.1-Flash-Paiton-CDNA4/resolve/2dd8058eab24484c9e2a54884244d68c825ccd7c/proof/factory/sweep-fac-a/cells.jsonl" }, { "file": "proof/factory/sweep-fac-b/cells.jsonl", "sha256": "62de7636c93174ae5f69e4807a5de0fe2e754a348693cf8794074c9e8ed7fb2f", "url": "https://huggingface.co/EliovpAI/DeepSeek-V4.1-Flash-Paiton-CDNA4/resolve/2dd8058eab24484c9e2a54884244d68c825ccd7c/proof/factory/sweep-fac-b/cells.jsonl" }, { "file": "proof/factory/sweep-fac-c/cells.jsonl", "sha256": "86f5c1477d8553b942b33658bc376b40c250ed01906702721128ba1a31b8e644", "url": "https://huggingface.co/EliovpAI/DeepSeek-V4.1-Flash-Paiton-CDNA4/resolve/2dd8058eab24484c9e2a54884244d68c825ccd7c/proof/factory/sweep-fac-c/cells.jsonl" }, { "file": "proof/factory/sweep-fac-d/cells.jsonl", "sha256": "4323d9b4184b26f374f546b01672d600d12792004d5fc9fc1f414dd82fc02492", "url": "https://huggingface.co/EliovpAI/DeepSeek-V4.1-Flash-Paiton-CDNA4/resolve/2dd8058eab24484c9e2a54884244d68c825ccd7c/proof/factory/sweep-fac-d/cells.jsonl" }, { "file": "proof/factory/sweep-fac-e/cells.jsonl", "sha256": "cec1f986edc47f05ecdd7c7273eaab1fcae03cf54a2117a0a0c56c1d6df5a23d", "url": "https://huggingface.co/EliovpAI/DeepSeek-V4.1-Flash-Paiton-CDNA4/resolve/2dd8058eab24484c9e2a54884244d68c825ccd7c/proof/factory/sweep-fac-e/cells.jsonl" }, { "file": "proof/factory/sweep-fac-f/cells.jsonl", "sha256": "535d9705f6858ee805ffcc8853e5935bb17e0b7622b7cc94788978509df57046", "url": "https://huggingface.co/EliovpAI/DeepSeek-V4.1-Flash-Paiton-CDNA4/resolve/2dd8058eab24484c9e2a54884244d68c825ccd7c/proof/factory/sweep-fac-f/cells.jsonl" } ], "decode": { "source": "proof/interactive/lmx/inference-matrices/deepseek-ai-DeepSeek-V4.1-Flash-1017/matrix-20261102T130034Z.json", "concurrency": [ 1, 2, 4, 8, 16, 32, 64 ], "outputTokensPerSecond": [ 77.9, 112.7, 166.9, 241.4, 377.9, 518.4, 709.6 ], "unit": "aggregate output tokens/s", "method": "LocalMaxxing 0.1.54 / llm-inference-bench 0.7.7; context=0; 30-second sustained-decode cells; selected whole median-c1 run of three. Not fixed-length end-to-end serving throughput.", "prefill": { "8192": { "client_tok_per_sec": 4366, "client_ttft_seconds": 1.86, "hardware_summary": {}, "method": "scout_only", "prefill_seconds": 1.86, "prompt_tokens": 8119, "samples": 1, "server_validation": { "cached_tokens": 0, "invalid_reason": "", "method": "", "prefill_seconds": 0, "prompt_tokens": 0, "request_prompt_tokens": 0, "samples": 0, "tok_per_sec": 0, "token_source": "" }, "tok_per_sec": 4366, "ttft_seconds": 1.86 }, "65536": { "client_tok_per_sec": 4227, "client_ttft_seconds": 15.079, "hardware_summary": {}, "method": "scout_only", "prefill_seconds": 15.079, "prompt_tokens": 63742, "samples": 1, "server_validation": { "cached_tokens": 0, "invalid_reason": "", "method": "", "prefill_seconds": 0, "prompt_tokens": 0, "request_prompt_tokens": 0, "samples": 0, "tok_per_sec": 0, "token_source": "" }, "tok_per_sec": 4227, "ttft_seconds": 15.079 } } }, "cache": { "unit": "GPU KV tokens", "method": "Startup-reported physical cache allocation, not one model context or demonstrated simultaneous maximum-length requests.", "rows": [ { "profile": "interactive", "tokens": 4484095, "slots": 64, "maxTokensPerRequest": 70000, "source": "proof/interactive/kv-capacity.txt" }, { "profile": "factory", "tokens": 6324224, "slots": 512, "maxTokensPerRequest": 12288, "source": "proof/factory/kv-capacity.txt" } ] }, "accuracy": { "metric": "teacher-forced top-64 bucket KL, not task accuracy", "rows": [ { "profile": "interactive", "chatKL": 0.030668283281383858, "longKL": 0.04379518725794706, "chatPositions": 1767, "longPositions": 1000, "chatTop1Agreement": 0.9496321448783248, "longTop1Agreement": 0.942, "measuredConcurrency": 8, "sources": [ "proof/accuracy/interactive/rp_chat_kl.json", "proof/accuracy/interactive/long_ds1_kl.json" ] }, { "profile": "factory", "chatKL": 0.031129499544153096, "longKL": 0.041762516589460154, "chatPositions": 1767, "longPositions": 1000, "chatTop1Agreement": 0.944538766270515, "longTop1Agreement": 0.942, "measuredConcurrency": 8, "sources": [ "proof/accuracy/factory/rp_chat_kl.json", "proof/accuracy/factory/long_ds1_kl.json" ] } ], "gate": { "longKL": 0.045, "emulatorFullSetKL": 0.0444, "emulatorFullSetPositions": 171200, "projectedServedFullSetKL": 0.0456, "measured": false, "source": "README.md", "limit": "Full-set emulator and served projection are model-card narrative; raw full-set emulator outputs are not in the public proof. No confidence interval is assigned to the served projection." } }, "serving": { "aggregateUnit": "output tokens/s per GPU", "userUnit": "output tokens/s per request = reciprocal median TPOT after the first token; not aggregate output/concurrency or end-to-end speed", "ttftUnit": "seconds", "powerUnit": "mean GPU SOCKET_POWER in W; not whole-server power", "workloads": [ { "kind": "serving-short", "inputTokens": 1024, "outputTokens": 1024, "cells": [ { "t_start": 1793629483.4, "t_end": 1793629573.7, "isl": 1024, "osl": 1024, "conc": 1, "n": 8, "errors": 0, "wall_s": 90.3, "short": 0, "out_tok_s": 90.7, "user_tok_s": 95.4, "gpus": 1, "power_w": 251.8, "tok_s_per_kw": 360.3, "ttft_p50_ms": 687.8, "ttft_p99_ms": 773, "tpot_p50_ms": 10.5, "tpot_p99_ms": 12, "e2e_p50_ms": 11382, "e2e_p99_ms": 12958.4, "itl_p50_ms": 27.44, "itl_p99_ms": 37.01, "profile": "interactive", "ttftMedianS": 0.6878, "source": "proof/interactive/sweep-int-a/cells.jsonl" }, { "t_start": 1793629579.9, "t_end": 1793629633, "isl": 1024, "osl": 1024, "conc": 4, "n": 8, "errors": 0, "wall_s": 53.1, "short": 0, "out_tok_s": 154.3, "user_tok_s": 42.4, "gpus": 1, "power_w": 243.1, "tok_s_per_kw": 634.5, "ttft_p50_ms": 796.3, "ttft_p99_ms": 2873.4, "tpot_p50_ms": 23.6, "tpot_p99_ms": 27.5, "e2e_p50_ms": 24844.9, "e2e_p99_ms": 31014.4, "itl_p50_ms": 59.47, "itl_p99_ms": 70.88, "profile": "interactive", "ttftMedianS": 0.7963, "source": "proof/interactive/sweep-int-a/cells.jsonl" }, { "t_start": 1793629657.2, "t_end": 1793629818, "isl": 1024, "osl": 1024, "conc": 16, "n": 32, "errors": 0, "wall_s": 160.8, "short": 0, "out_tok_s": 203.8, "user_tok_s": 14.4, "gpus": 1, "power_w": 231.6, "tok_s_per_kw": 879.9, "ttft_p50_ms": 8913.3, "ttft_p99_ms": 11421.8, "tpot_p50_ms": 69.2, "tpot_p99_ms": 77.5, "e2e_p50_ms": 80131.1, "e2e_p99_ms": 84579.3, "itl_p50_ms": 67.48, "itl_p99_ms": 75.26, "profile": "interactive", "ttftMedianS": 8.9133, "source": "proof/interactive/sweep-int-a/cells.jsonl" }, { "t_start": 1793629910.3, "t_end": 1793630130.8, "isl": 1024, "osl": 1024, "conc": 64, "n": 64, "errors": 0, "wall_s": 220.5, "short": 0, "out_tok_s": 297.2, "user_tok_s": 5.2, "gpus": 1, "power_w": 207.6, "tok_s_per_kw": 1432, "ttft_p50_ms": 28249.7, "ttft_p99_ms": 44638.7, "tpot_p50_ms": 192.9, "tpot_p99_ms": 208.7, "e2e_p50_ms": 220202.8, "e2e_p99_ms": 220451.5, "itl_p50_ms": 172.2, "itl_p99_ms": 187.84, "profile": "interactive", "ttftMedianS": 28.2497, "source": "proof/interactive/sweep-int-a/cells.jsonl" }, { "t_start": 1793634768.8, "t_end": 1793635251.9, "isl": 1024, "osl": 1024, "conc": 256, "n": 256, "errors": 0, "wall_s": 483, "short": 0, "out_tok_s": 542.8, "user_tok_s": 2.6, "gpus": 1, "power_w": 200.8, "tok_s_per_kw": 2702.7, "ttft_p50_ms": 91562.4, "ttft_p99_ms": 178513.5, "tpot_p50_ms": 381, "tpot_p99_ms": 459.3, "e2e_p50_ms": 481310.2, "e2e_p99_ms": 482941.2, "itl_p50_ms": 298.85, "itl_p99_ms": 4057.86, "profile": "factory", "ttftMedianS": 91.5624, "source": "proof/factory/sweep-fac-a/cells.jsonl" }, { "t_start": 1793635750.9, "t_end": 1793636384.5, "isl": 1024, "osl": 1024, "conc": 384, "n": 384, "errors": 0, "wall_s": 633.4, "short": 0, "out_tok_s": 620.8, "user_tok_s": 2.1, "gpus": 1, "power_w": 208.6, "tok_s_per_kw": 2976.6, "ttft_p50_ms": 131764, "ttft_p99_ms": 249569.4, "tpot_p50_ms": 487.5, "tpot_p99_ms": 603.3, "e2e_p50_ms": 630469.2, "e2e_p99_ms": 633251.3, "itl_p50_ms": 379.13, "itl_p99_ms": 7575.9, "profile": "factory", "ttftMedianS": 131.764, "source": "proof/factory/sweep-fac-a/cells.jsonl" }, { "t_start": 1793637050.5, "t_end": 1793637818.9, "isl": 1024, "osl": 1024, "conc": 512, "n": 512, "errors": 0, "wall_s": 768.2, "short": 0, "out_tok_s": 682.5, "user_tok_s": 1.7, "gpus": 1, "power_w": 212.5, "tok_s_per_kw": 3212, "ttft_p50_ms": 167904.6, "ttft_p99_ms": 331243.9, "tpot_p50_ms": 582.3, "tpot_p99_ms": 730.8, "e2e_p50_ms": 763601.6, "e2e_p99_ms": 768057.8, "itl_p50_ms": 433.3, "itl_p99_ms": 7617.15, "profile": "factory", "ttftMedianS": 167.9046, "source": "proof/factory/sweep-fac-a/cells.jsonl" } ] }, { "kind": "serving-long-input", "inputTokens": 8192, "outputTokens": 1024, "cells": [ { "t_start": 1793630134.9, "t_end": 1793630238, "isl": 8192, "osl": 1024, "conc": 1, "n": 8, "errors": 0, "wall_s": 103.1, "short": 0, "out_tok_s": 79.5, "user_tok_s": 95.2, "gpus": 1, "power_w": 256.6, "tok_s_per_kw": 309.6, "ttft_p50_ms": 1828.8, "ttft_p99_ms": 1881.4, "tpot_p50_ms": 10.5, "tpot_p99_ms": 12.9, "e2e_p50_ms": 12471.3, "e2e_p99_ms": 14928.4, "itl_p50_ms": 27.82, "itl_p99_ms": 36.08, "profile": "interactive", "ttftMedianS": 1.8288, "source": "proof/interactive/sweep-int-a/cells.jsonl" }, { "t_start": 1793630253.6, "t_end": 1793630316.7, "isl": 8192, "osl": 1024, "conc": 4, "n": 8, "errors": 0, "wall_s": 63.1, "short": 0, "out_tok_s": 129.8, "user_tok_s": 39.9, "gpus": 1, "power_w": 260.2, "tok_s_per_kw": 498.9, "ttft_p50_ms": 3866, "ttft_p99_ms": 7389.7, "tpot_p50_ms": 25.1, "tpot_p99_ms": 35.4, "e2e_p50_ms": 32605.9, "e2e_p99_ms": 37966.2, "itl_p50_ms": 60.52, "itl_p99_ms": 73.04, "profile": "interactive", "ttftMedianS": 3.866, "source": "proof/interactive/sweep-int-a/cells.jsonl" }, { "t_start": 1793630381.6, "t_end": 1793630496.8, "isl": 8192, "osl": 1024, "conc": 16, "n": 17, "errors": 0, "wall_s": 115.2, "short": 0, "out_tok_s": 151.1, "user_tok_s": 12.1, "gpus": 1, "power_w": 259.4, "tok_s_per_kw": 582.7, "ttft_p50_ms": 17093.3, "ttft_p99_ms": 31817.6, "tpot_p50_ms": 82.4, "tpot_p99_ms": 95.6, "e2e_p50_ms": 103027.3, "e2e_p99_ms": 103589, "itl_p50_ms": 68.95, "itl_p99_ms": 130.79, "profile": "interactive", "ttftMedianS": 17.0933, "source": "proof/interactive/sweep-int-a/cells.jsonl" }, { "t_start": 1793630775.9, "t_end": 1793631101, "isl": 8192, "osl": 1024, "conc": 64, "n": 64, "errors": 0, "wall_s": 325.1, "short": 0, "out_tok_s": 201.6, "user_tok_s": 4, "gpus": 1, "power_w": 251.2, "tok_s_per_kw": 802.3, "ttft_p50_ms": 70448.4, "ttft_p99_ms": 146612.4, "tpot_p50_ms": 249.1, "tpot_p99_ms": 303.1, "e2e_p50_ms": 323027.8, "e2e_p99_ms": 324963.9, "itl_p50_ms": 177.98, "itl_p99_ms": 2573.14, "profile": "interactive", "ttftMedianS": 70.4484, "source": "proof/interactive/sweep-int-a/cells.jsonl" }, { "t_start": 1793639405.7, "t_end": 1793640282.4, "isl": 8192, "osl": 1024, "conc": 256, "n": 256, "errors": 0, "wall_s": 876.5, "short": 0, "out_tok_s": 299.1, "user_tok_s": 1.8, "gpus": 1, "power_w": 267.6, "tok_s_per_kw": 1117.4, "ttft_p50_ms": 285419.4, "ttft_p99_ms": 573378.9, "tpot_p50_ms": 564.7, "tpot_p99_ms": 811.1, "e2e_p50_ms": 863146.1, "e2e_p99_ms": 876279.7, "itl_p50_ms": 306.24, "itl_p99_ms": 3707.66, "profile": "factory", "ttftMedianS": 285.4194, "source": "proof/factory/sweep-fac-b/cells.jsonl" }, { "t_start": 1793641946.9, "t_end": 1793643184.6, "isl": 8192, "osl": 1024, "conc": 384, "n": 384, "errors": 0, "wall_s": 1237.5, "short": 0, "out_tok_s": 317.8, "user_tok_s": 1.3, "gpus": 1, "power_w": 271.8, "tok_s_per_kw": 1169.1, "ttft_p50_ms": 434135.9, "ttft_p99_ms": 870754.9, "tpot_p50_ms": 765, "tpot_p99_ms": 1134.7, "e2e_p50_ms": 1213539.1, "e2e_p99_ms": 1237067.9, "itl_p50_ms": 382.57, "itl_p99_ms": 3748.86, "profile": "factory", "ttftMedianS": 434.1359, "source": "proof/factory/sweep-fac-b/cells.jsonl" }, { "t_start": 1793645886.3, "t_end": 1793647480.7, "isl": 8192, "osl": 1024, "conc": 512, "n": 512, "errors": 0, "wall_s": 1594.2, "short": 0, "out_tok_s": 328.9, "user_tok_s": 1.1, "gpus": 1, "power_w": 273.4, "tok_s_per_kw": 1202.8, "ttft_p50_ms": 586390.1, "ttft_p99_ms": 1191703.6, "tpot_p50_ms": 952.1, "tpot_p99_ms": 1446.1, "e2e_p50_ms": 1557370.8, "e2e_p99_ms": 1593561.6, "itl_p50_ms": 429.26, "itl_p99_ms": 3862.31, "profile": "factory", "ttftMedianS": 586.3901, "source": "proof/factory/sweep-fac-c/cells.jsonl" } ] }, { "kind": "serving-long-output", "inputTokens": 1024, "outputTokens": 8192, "cells": [ { "t_start": 1793631555.7, "t_end": 1793631706, "isl": 1024, "osl": 8192, "conc": 1, "n": 2, "errors": 0, "wall_s": 150.3, "short": 0, "out_tok_s": 109, "user_tok_s": 120, "gpus": 1, "power_w": 264.5, "tok_s_per_kw": 412.2, "ttft_p50_ms": 718.3, "ttft_p99_ms": 820.6, "tpot_p50_ms": 8.3, "tpot_p99_ms": 9.8, "e2e_p50_ms": 68995.7, "e2e_p99_ms": 81275.1, "itl_p50_ms": 26.66, "itl_p99_ms": 34.8, "profile": "interactive", "ttftMedianS": 0.7183, "source": "proof/interactive/sweep-int-b/cells.jsonl" }, { "t_start": 1793631712.4, "t_end": 1793631876.8, "isl": 1024, "osl": 8192, "conc": 4, "n": 4, "errors": 0, "wall_s": 164.4, "short": 0, "out_tok_s": 199.3, "user_tok_s": 52.3, "gpus": 1, "power_w": 261, "tok_s_per_kw": 763.4, "ttft_p50_ms": 2990.5, "ttft_p99_ms": 2991.5, "tpot_p50_ms": 19.1, "tpot_p99_ms": 19.7, "e2e_p50_ms": 158194.4, "e2e_p99_ms": 164436.7, "itl_p50_ms": 55.47, "itl_p99_ms": 64.92, "profile": "interactive", "ttftMedianS": 2.9905, "source": "proof/interactive/sweep-int-b/cells.jsonl" }, { "t_start": 1793631901.5, "t_end": 1793632439.7, "isl": 1024, "osl": 8192, "conc": 16, "n": 16, "errors": 0, "wall_s": 538.1, "short": 0, "out_tok_s": 243.6, "user_tok_s": 15.4, "gpus": 1, "power_w": 240.6, "tok_s_per_kw": 1012.5, "ttft_p50_ms": 11015.9, "ttft_p99_ms": 11017.3, "tpot_p50_ms": 64.9, "tpot_p99_ms": 65.6, "e2e_p50_ms": 538063.9, "e2e_p99_ms": 538070.8, "itl_p50_ms": 63.99, "itl_p99_ms": 71.86, "profile": "interactive", "ttftMedianS": 11.0159, "source": "proof/interactive/sweep-int-b/cells.jsonl" }, { "t_start": 1793632536.7, "t_end": 1793633951.9, "isl": 1024, "osl": 8192, "conc": 64, "n": 64, "errors": 0, "wall_s": 1415, "short": 0, "out_tok_s": 370.5, "user_tok_s": 5.9, "gpus": 1, "power_w": 216.1, "tok_s_per_kw": 1714.9, "ttft_p50_ms": 29091.3, "ttft_p99_ms": 45833.3, "tpot_p50_ms": 169.8, "tpot_p99_ms": 171.9, "e2e_p50_ms": 1414699.3, "e2e_p99_ms": 1414940.5, "itl_p50_ms": 167.35, "itl_p99_ms": 178.95, "profile": "interactive", "ttftMedianS": 29.0913, "source": "proof/interactive/sweep-int-b/cells.jsonl" }, { "t_start": 1793673771.8, "t_end": 1793676351.7, "isl": 1024, "osl": 8192, "conc": 256, "n": 256, "errors": 0, "wall_s": 2578.9, "short": 0, "out_tok_s": 813.2, "user_tok_s": 3.3, "gpus": 1, "power_w": 209.3, "tok_s_per_kw": 3886.2, "ttft_p50_ms": 92903.5, "ttft_p99_ms": 181242.6, "tpot_p50_ms": 303.3, "tpot_p99_ms": 313.2, "e2e_p50_ms": 2577395.9, "e2e_p99_ms": 2578893.1, "itl_p50_ms": 292.17, "itl_p99_ms": 320.3, "profile": "factory", "ttftMedianS": 92.9035, "source": "proof/factory/sweep-fac-d/cells.jsonl" }, { "t_start": 1793677344.2, "t_end": 1793680645.5, "isl": 1024, "osl": 8192, "conc": 384, "n": 384, "errors": 0, "wall_s": 3299.7, "short": 0, "out_tok_s": 953.4, "user_tok_s": 2.6, "gpus": 1, "power_w": 210.1, "tok_s_per_kw": 4536.5, "ttft_p50_ms": 138964.7, "ttft_p99_ms": 263735.6, "tpot_p50_ms": 385.5, "tpot_p99_ms": 400.8, "e2e_p50_ms": 3296788.4, "e2e_p99_ms": 3299571.8, "itl_p50_ms": 371.65, "itl_p99_ms": 406.5, "profile": "factory", "ttftMedianS": 138.9647, "source": "proof/factory/sweep-fac-e/cells.jsonl" }, { "t_start": 1793681809.8, "t_end": 1793685598.6, "isl": 1024, "osl": 8192, "conc": 512, "n": 512, "errors": 0, "wall_s": 3786.7, "short": 0, "out_tok_s": 1107.6, "user_tok_s": 2.3, "gpus": 1, "power_w": 214.7, "tok_s_per_kw": 5159.4, "ttft_p50_ms": 180835.3, "ttft_p99_ms": 355272.3, "tpot_p50_ms": 439.7, "tpot_p99_ms": 459.7, "e2e_p50_ms": 3782370.6, "e2e_p99_ms": 3786605.1, "itl_p50_ms": 419.24, "itl_p99_ms": 648.08, "profile": "factory", "ttftMedianS": 180.8353, "source": "proof/factory/sweep-fac-f/cells.jsonl" } ] } ] } }