{"id":"run_hL-HgU1CB7sW17kN","owner":"anonymous-EVjB2S","publishedAt":"2026-08-30T04:08:18.062Z","manifest":{"schema_version":"1.0","kind":"run","client_run_id":"rps-1fee0eac51b6f88ad8b2705a9af4a31b0c48e5a2","name":"gittensor-model-hub/Qwen3.8-27B-NVFP4-RTX5090 | baseline | 8192->256 | C3","notes":"Aggregate-only synthetic online serving measurement. It reports serving performance and does not evaluate model quality.","visibility":"public","license":"CC-BY-4.0","outcome":"completed","subject":{"type":"model","source":"huggingface","adapters":[],"revision":"0cc27958cefbbe231782ec8511de8c4eb5233348","identifier":"gittensor-model-hub/Qwen3.8-27B-NVFP4-RTX5090","modalities":["text"],"architecture":"Qwen3_5ForConditionalGeneration","quantization":{"format":"NVFP4 W4A4","recipe":"gittensor-model-hub RTX 5090 target checkpoint; BF16 speculative drafter is a runtime modification","checkpoint":"NVIDIA ModelOpt"},"modifications":["Speculative draft-bf16 checkpoint: incoai/Qwen3.8-27B-DFlash2@dedf8df68adfb1afeaf7b7480c0a0243108177b4."],"revision_kind":"commit","parameter_count_total":27000000000,"parameter_count_active":27000000000},"environment":{"os":"Linux-6.8.0-136-generic-x86_64-with-glibc2.39","cpu":{"model":"x86_64"},"kernel":"6.8.0-136-generic","topology":"One server engine on one GPU.","accelerators":[{"count":1,"model":"NVIDIA GeForce RTX 5090","driver":"580.173.02","vendor":"NVIDIA","architecture":"Blackwell SM120","interconnect":"single device","memory_bytes_per_device":34190917632}],"driver_versions":{"nvidia":"580.173.02"},"host_architecture":"x86_64","system_memory_bytes":269585743872},"runtime":{"engine":"sglang","version":"0.0.0.dev1+g5f55db35e","extensions":{"runpile.server_variant":"baseline","runpile.auxiliary_models":[{"role":"draft-bf16","revision":"dedf8df68adfb1afeaf7b7480c0a0243108177b4","repository":"incoai/Qwen3.8-27B-DFlash2","quantization":"BF16","weight_bytes":3849104785}]},"parameters":{"server":{"dtype":"bfloat16","context_length":9216,"kv_cache_dtype":"fp8_e4m3","attention_backend":"flashinfer","cuda_graph_max_bs":4,"mem_fraction_static":0.9,"chunked_prefill_size":1024,"max_running_requests":4}},"parallelism":{"data":1,"expert":1,"tensor":1,"pipeline":1,"replicas":1},"launch_command":"CUDA_VISIBLE_DEVICES=0 sglang serve --model-path gittensor-model-hub/Qwen3.8-27B-NVFP4-RTX5090 --revision 0cc27958cefbbe231782ec8511de8c4eb5233348 --served-model-name gittensor-model-hub/Qwen3.8-27B-NVFP4-RTX5090 --port 9000 --host 127.0.0.1 --dtype bfloat16 --mem-fraction-static 0.9 --context-length 9216 --max-running-requests 4 --tp 1 --kv-cache-dtype fp8_e4m3 --chunked-prefill-size 1024 --attention-backend flashinfer --cuda-graph-max-bs 4 --trust-remote-code --reasoning-parser qwen3 --tool-call-parser qwen3_coder --mamba-full-memory-ratio 5.61 --mamba-radix-cache-strategy extra_buffer_lazy --mamba-ssm-dtype bfloat16 --max-mamba-cache-size 16 --cuda-graph-bs-prefill 64 128 256 512 1024 --enable-memory-saver --enable-hierarchical-cache --hicache-ratio 2 --hicache-write-policy write_through --speculative-algorithm DFLASH --speculative-draft-model-path incoai/Qwen3.8-27B-DFlash2 --speculative-draft-model-revision dedf8df68adfb1afeaf7b7480c0a0243108177b4 --speculative-draft-model-quantization unquant --speculative-num-draft-tokens 8","build_identifier":"lmsysorg/sglang@sha256:616a3e97f45191af975896cfa644279096cb31bd408a071c2e99ca7209c3cafe","safe_environment":{"HF_HUB_OFFLINE":"1","CUDA_VISIBLE_DEVICES":"0","TOKENIZERS_PARALLELISM":"false"}},"workload":{"mode":"online","task":"text_generation","client":{"name":"sglang benchmark serving","backend":"sglang-oai","version":"0.0.0.dev1+g5f55db35e","placement":"same-host"},"dataset":{"sha256":"bf8e1edf5dcbcc6d9e19ff7fa2ecd06a04491fc4d22d9a758f5913706d71ab84","generator":"runpile-sweeps-tokenizer-verified-jsonl","synthetic":true,"generator_revision":"b62df3a992da3d4a1dc6f2f12ce2706bbeef61bc5ca0c03c917d7597606c3312","shared_prefix_tokens":0},"sampling":{"mode":"greedy","temperature":0},"stopping":{"ignore_eos":true,"max_output_tokens":256},"streaming":true,"tokenizer":{"revision":"0cc27958cefbbe231782ec8511de8c4eb5233348","identifier":"gittensor-model-hub/Qwen3.8-27B-NVFP4-RTX5090"},"extensions":{"runpile.profile_id":"serve-fixed-8k-256-v1"},"concurrency":3,"random_seed":0,"input_tokens":{"kind":"fixed","value":8192},"endpoint_type":"openai-compatible-completions","output_tokens":{"kind":"fixed","value":256},"request_count":128,"arrival_process":"closed_loop","prefix_behavior":{"cache_state":"warm-host","cache_enabled":false,"shared_prefix_tokens":0},"warmup_request_count":8,"request_rate_per_second":null},"measurement":{"notes":"Aggregate-only output.","warmup":"8 unmeasured warmups per invocation after one readiness request.","harness":"sglang benchmark serving","version":"0.0.0.dev1+g5f55db35e","extensions":{"runpile.declared_warmup_count":8,"runpile.max_gpu_temperature_c":58,"runpile.preemption_log_mentions":2,"runpile.readiness_request_count":1,"runpile.controller_source_sha256":"b62df3a992da3d4a1dc6f2f12ce2706bbeef61bc5ca0c03c917d7597606c3312","runpile.speculative_accept_length":6.711866844241026,"runpile.independent_repetition_index":1},"repetitions":1,"timing_scope":"client-observed streaming request latency","benchmark_command":"sglang bench serving --backend sglang-oai --base-url http://127.0.0.1:9000 --model gittensor-model-hub/Qwen3.8-27B-NVFP4-RTX5090 --served-model-name gittensor-model-hub/Qwen3.8-27B-NVFP4-RTX5090 --tokenizer gittensor-model-hub/Qwen3.8-27B-NVFP4-RTX5090 --dataset-name custom --dataset-path <EXACT_SYNTHETIC_DATASET> --sharegpt-output-len 256 --num-prompts 128 --max-concurrency 3 --request-rate inf --seed 0 --temperature 0 --disable-tqdm --output-file <local-path> --warmup-requests 8 --tag runpile-rtx5090-qwen38-dflash2-concurrency-20260829-v5:qwen38-27b-nvfp4-sglang-dflash2-bf16-tp1:baseline:serve-fixed-8k-256-v1:c3:r1","retry_failure_treatment":"Failed requests counted; no client retries.","client_overhead_included":true},"results":{"latency_ms":{"end_to_end":{"sample_count":128,"measurement_status":"measured"},"inter_token":{"p50":2.2664308780804276,"p90":8.84498975938186,"p95":18.30510003492236,"p99":71.93838449893519,"mean":6.993769063155499,"stddev":26.287357450343325,"sample_count":128,"measurement_status":"measured"},"time_to_first_token":{"p50":659.3930011149496,"p90":1274.9187833163887,"p95":1278.1397788319737,"p99":1876.562482654117,"mean":776.3442426621623,"stddev":266.81173978603505,"sample_count":128,"measurement_status":"measured"},"time_per_output_token":{"p50":7.000689385641439,"p90":7.310698792959253,"p95":11.280852340015706,"p99":31.40259933214198,"mean":6.947298356566998,"stddev":5.018292424193758,"sample_count":128,"measurement_status":"measured"}},"throughput":{"requests_per_second":{"value":1.1702902513528894,"sample_count":128,"measurement_status":"measured"},"input_tokens_per_second":{"value":9587.01773908287,"sample_count":128,"measurement_status":"measured"},"output_tokens_per_second":{"value":299.5943043463397,"sample_count":128,"measurement_status":"measured"}},"failed_requests":0,"duration_seconds":109.37457596696913,"total_input_tokens":1048576,"successful_requests":128,"total_output_tokens":32768},"artifacts":[{"kind":"sglang_bench_serving_json","sha256":"6d8ddad59b0018f9a0a7b9311aa66430f079ed7c5b55010ac9f4a557825660a0","license":"CC-BY-4.0","filename":"qwen38-27b-nvfp4-sglang-dflash2-bf16-tp1__baseline__serve-fixed-8k-256-v1__c3__r1.json","media_type":"application/json","size_bytes":43679,"visibility":"private","client_artifact_id":"raw-sglang-serve","raw_requests_included":false}],"provenance":{"tool":"runpile HTTP importer","source":"import","tool_version":"1.0"},"warnings":["Review imported facts against the local model, hardware, launch command, and benchmark command before publication."]},"indexingState":"indexed","resultKey":"res_sUQTmzhWKqfy5GYRPNMk2-st","eligibilityReasons":[],"artifactBacked":true,"downloads":[]}