{"id":"run_F4Kz3F-9yzkw5N-l","owner":"anonymous-EVjB2S","publishedAt":"2026-08-30T04:08:23.523Z","manifest":{"schema_version":"1.0","kind":"run","client_run_id":"rps-53e96ec82d1522fbc686c2d094f977bf75de94e9","name":"gittensor-model-hub/Qwen3.8-27B-NVFP4-RTX5090 | baseline | 1024->256 | C1","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; calibrated NVFP4 speculative drafter is a runtime modification","checkpoint":"NVIDIA ModelOpt"},"modifications":["Speculative draft-nvfp4 checkpoint: maurienne-ai/Qwen3.8-27B-DFlash2-NVFP4-RTNcal@b45a9e0fe2d62963a91158e6fb776fb86532c190."],"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-nvfp4","revision":"b45a9e0fe2d62963a91158e6fb776fb86532c190","repository":"maurienne-ai/Qwen3.8-27B-DFlash2-NVFP4-RTNcal","quantization":"ModelOpt NVFP4 W4A4, calibrated","weight_bytes":1550153248}]},"parameters":{"server":{"dtype":"bfloat16","context_length":9216,"kv_cache_dtype":"fp8_e4m3","attention_backend":"flashinfer","cuda_graph_max_bs":8,"mem_fraction_static":0.9,"chunked_prefill_size":1024,"max_running_requests":8}},"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 8 --tp 1 --kv-cache-dtype fp8_e4m3 --chunked-prefill-size 1024 --attention-backend flashinfer --cuda-graph-max-bs 8 --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 32 --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 maurienne-ai/Qwen3.8-27B-DFlash2-NVFP4-RTNcal --speculative-draft-model-revision b45a9e0fe2d62963a91158e6fb776fb86532c190 --speculative-draft-model-quantization modelopt_fp4 --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":"7f10b755bb8a97e1f56cbe8e2e66b6af3cc8078b04f2dd3474db54ddba85214b","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-1k-256-v1"},"concurrency":1,"random_seed":0,"input_tokens":{"kind":"fixed","value":1024},"endpoint_type":"openai-compatible-completions","output_tokens":{"kind":"fixed","value":256},"request_count":256,"arrival_process":"closed_loop","prefix_behavior":{"cache_state":"warm-host","cache_enabled":false,"shared_prefix_tokens":0},"warmup_request_count":16,"request_rate_per_second":null},"measurement":{"notes":"Aggregate-only output.","warmup":"16 unmeasured warmups per invocation after one readiness request.","harness":"sglang benchmark serving","version":"0.0.0.dev1+g5f55db35e","extensions":{"runpile.declared_warmup_count":16,"runpile.max_gpu_temperature_c":53,"runpile.preemption_log_mentions":2,"runpile.readiness_request_count":1,"runpile.controller_source_sha256":"b62df3a992da3d4a1dc6f2f12ce2706bbeef61bc5ca0c03c917d7597606c3312","runpile.speculative_accept_length":7.065197334700154,"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 256 --max-concurrency 1 --request-rate inf --seed 0 --temperature 0 --disable-tqdm --output-file <local-path> --warmup-requests 16 --tag runpile-rtx5090-qwen38-dflash2-concurrency-20260829-v5:qwen38-27b-nvfp4-sglang-dflash2-nvfp4-tp1:baseline:serve-fixed-1k-256-v1:c1:r1","retry_failure_treatment":"Failed requests counted; no client retries.","client_overhead_included":true},"results":{"latency_ms":{"end_to_end":{"sample_count":256,"measurement_status":"measured"},"inter_token":{"p50":1.85642036376521,"p90":1.8972736434079707,"p95":2.4726014429082475,"p99":14.749905630014839,"mean":2.1029689214567,"stddev":1.6458050812176364,"sample_count":256,"measurement_status":"measured"},"time_to_first_token":{"p50":81.2315121293068,"p90":81.69691357761621,"p95":81.823744578287,"p99":82.23283735569566,"mean":81.0871836165461,"stddev":3.144307266092219,"sample_count":256,"measurement_status":"measured"},"time_per_output_token":{"p50":1.8369200128112353,"p90":1.8953628043698914,"p95":2.5646597763304326,"p99":11.553626507520656,"mean":2.1020423658317706,"stddev":1.4853279243907587,"sample_count":256,"measurement_status":"measured"}},"throughput":{"requests_per_second":{"value":1.6198224625793924,"sample_count":256,"measurement_status":"measured"},"input_tokens_per_second":{"value":1658.6982016812979,"sample_count":256,"measurement_status":"measured"},"output_tokens_per_second":{"value":414.67455042032446,"sample_count":256,"measurement_status":"measured"}},"failed_requests":0,"duration_seconds":158.04201134014875,"total_input_tokens":262144,"successful_requests":256,"total_output_tokens":65536},"artifacts":[{"kind":"sglang_bench_serving_json","sha256":"179c5726c5054d421c380905a00402b2c6da75d16ad3fe49a7cf51cf723f79b7","license":"CC-BY-4.0","filename":"qwen38-27b-nvfp4-sglang-dflash2-nvfp4-tp1__baseline__serve-fixed-1k-256-v1__c1__r1.json","media_type":"application/json","size_bytes":43854,"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_N-8vzLqyJP05Y8EqryzfSg4i","eligibilityReasons":[],"artifactBacked":true,"downloads":[]}