{"id":"run_FR4A4S6PaeJtiExH","owner":"anonymous-TWTejS","publishedAt":"2026-08-23T22:38:43.669Z","manifest":{"schema_version":"1.0","kind":"run","client_run_id":"vllm-0b8ffbd9c7b1-1","name":"Muse-Glimmer-30B-GGUF | 1x RTX 5090 | 1024->256 | C1","notes":"Text-only synthetic serving measurement. Exact token lengths, greedy decoding, streaming responses, and EOS ignored. No prompts, generated text, request bodies, or response bodies are retained in the aggregate artifact.","visibility":"public","license":"CC-BY-4.0","outcome":"completed","subject":{"type":"model","source":"huggingface","adapters":[],"revision":"70bf1b61ac09f91b24d39038091b41c582bc5d7a","identifier":"meta-models/Muse-Glimmer-30B-GGUF","modalities":["text"],"architecture":"MuseGlimmerForConditionalGeneration","quantization":{"format":"Dynamic Q4_K_XL","checkpoint":"GGUF"},"modifications":[],"revision_kind":"commit","parameter_count_total":30000000000,"parameter_count_active":30000000000},"environment":{"os":"Ubuntu 24.04.4 LTS","cpu":{"model":"AMD Ryzen Threadripper PRO 9955WX 16-Cores","sockets":1,"numa_nodes":1,"logical_cpus":32,"physical_cores":16},"kernel":"6.8.0-136-generic","topology":"One process restricted to GPU 0; other host GPUs idle.","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","cuda_compiler":"Build cuda_13.1.r13.1/compiler.37061995_0"},"host_architecture":"x86_64","system_memory_bytes":269585743872},"runtime":{"engine":"llama.cpp","version":"b10353","parameters":{"server":{"dflash":false,"kv_cache":"Q8_0","slots_per_replica":128,"context_tokens_per_slot":2048,"context_tokens_per_replica":262144},"load_balancer":{"type":"llama_cpp_compat_proxy","cache_prompt":false,"sse_usage_split":true,"tokenizer_schema_translation":true}},"parallelism":{"tensor":1,"pipeline":1,"replicas":1},"launch_command":"CUDA_VISIBLE_DEVICES=0 llama-server -m Muse-Glimmer-30B-KQuant-Dynamic-Q4_K_XL.gguf -a meta-models/Muse-Glimmer-30B-GGUF -ngl 99 -c 262144 -np 128 --flash-attn on -ctk q8_0 -ctv q8_0 --host 127.0.0.1 --port 9001 --jinja --temp 1.0 --top-p 0.95 --top-k 64 -t 16 -tb 16\nRUNPILE_BACKENDS=<replica-urls> RUNPILE_PROXY_PORT=9000 python llama_compat_proxy.py","source_revision":"f8def7fe168bab245fbf15d3f18b26dbb1ef73c8","safe_environment":{"CUDA_VISIBLE_DEVICES":"0"}},"workload":{"mode":"online","task":"text_generation","client":{"name":"vllm bench serve","backend":"openai","version":"0.27.1"},"dataset":{"name":"vllm-random","range_ratio":0,"random_prefix_tokens":0},"sampling":{"mode":"greedy","temperature":0},"stopping":{"ignore_eos":true,"max_tokens":256},"streaming":true,"tokenizer":{"revision":"a4e59da52a7bc87ae7251dd5545c0dd437c44b68","identifier":"meta-models/Muse-Glimmer-30B"},"concurrency":1,"random_seed":20260823,"input_tokens":{"kind":"fixed","value":1024},"endpoint_type":"openai_completions","output_tokens":{"kind":"fixed","value":256},"request_count":200,"arrival_process":"closed_loop","prefix_behavior":{"shared_prefix_tokens":0,"server_prefix_caching":false},"warmup_request_count":16,"request_rate_per_second":null},"measurement":{"notes":"Aggregate-only output; --save-detailed was not enabled. For llama.cpp, the recorded compatibility proxy disables prompt caching, translates tokenize/detokenize schemas, and splits combined SSE usage so the harness verifies exact server token counts.","warmup":"16 explicit requests before the curve; the harness also issues one unmeasured endpoint test immediately before each timed point.","harness":"vllm bench serve","version":"0.27.1","repetitions":1,"timing_scope":"Client-observed, streaming online serving interval after the harness initial test request.","benchmark_command":"vllm bench serve --backend openai --host localhost --port 9000 --endpoint /v1/completions --model meta-models/Muse-Glimmer-30B-GGUF --tokenizer meta-models/Muse-Glimmer-30B --dataset-name random --random-input-len 1024 --random-output-len 256 --random-range-ratio 0 --num-prompts 200 --request-rate inf --max-concurrency 1 --ignore-eos --temperature 0 --seed 20260823 --percentile-metrics ttft,tpot,itl,e2el --metric-percentiles 50,90,95,99 --disable-tqdm --save-result --result-dir results --result-filename muse-glimmer-30b-dynamic-q4-k-xl__r1__i1024__o256__c1.json","retry_failure_treatment":"No client retries; failed requests are counted as failures.","client_overhead_included":true},"results":{"latency_ms":{"end_to_end":{"p50":3903.070593602024,"p90":4066.376924351789,"p95":4089.6655979333445,"p99":4159.546103754546,"mean":3942.3768382298294,"stddev":78.88258228290836,"sample_count":200,"measurement_status":"measured"},"inter_token":{"p50":14.060174115002155,"p90":14.111630385741591,"p95":14.155015582218766,"p99":28.137491303496063,"mean":14.52591154704435,"stddev":2.8327835203492584,"sample_count":200,"measurement_status":"measured"},"time_to_first_token":{"p50":315.09339599870145,"p90":477.0247323438525,"p95":501.92960684653366,"p99":572.1041093207895,"mean":353.9686300314497,"stddev":78.66647417379795,"sample_count":200,"measurement_status":"measured"},"time_per_output_token":{"p50":14.071763197745325,"p90":14.076214639981295,"p95":14.07919250926295,"p99":14.081575424675192,"mean":14.07218905175835,"stddev":0.003115568175893446,"sample_count":200,"measurement_status":"measured"}},"throughput":{"requests_per_second":{"value":0.2536384819382242,"sample_count":200,"measurement_status":"measured"},"total_tokens_per_second":{"value":324.6572568809269,"sample_count":200,"measurement_status":"measured"},"output_tokens_per_second":{"value":64.9314513761854,"sample_count":200,"measurement_status":"measured"}},"failed_requests":0,"duration_seconds":788.5238804130349,"total_input_tokens":204800,"successful_requests":200,"total_output_tokens":51200},"artifacts":[{"kind":"vllm_bench_serve_json","sha256":"0b8ffbd9c7b11b44fbd38f0d10b87f49a92a64c306a97e3c5497befed7ec6e9f","license":"CC-BY-4.0","filename":"muse-glimmer-30b-dynamic-q4-k-xl__r1__i1024__o256__c1.json","media_type":"application/json","size_bytes":1685,"visibility":"private","client_artifact_id":"raw-vllm-serve","raw_requests_included":false}],"provenance":{"tool":"Runpile vllm importer completed from local evidence","source":"import","generated_at":"2026-08-23T22:38:43.105188+00:00","tool_version":"1.0"}},"indexingState":"indexed","resultKey":"res_gMB24U1D6PfFMtymYp86cLpV","eligibilityReasons":[],"artifactBacked":true,"downloads":[]}