{ "argv": [ "/workshop/bench-quant-ladder/coi_quant.py", "--arm-id", "q-q6_K", "--model", "mistral-nemo:12b-instruct-2407-q6_K", "--num-ctx", "4096", "--think", "off", "--tasks", "c,d", "--probes", "--out", "/workshop/bench-quant-ladder/results/tasks/q-q6_K" ], "arm_id": "q-q6_K", "baseline_rows": [ { "memory_mib": 1.0, "sm_mhz": 210.0, "temp_c": 28.0, "uuid": "", "watts": 23.04 }, { "memory_mib": 1.0, "sm_mhz": 210.0, "temp_c": 27.0, "uuid": "", "watts": 25.75 } ], "board": "", "box": "benchbox", "box_class": "gpu-ql3090-24g", "box_class_check": "gpu-ql3090-24g", "disk_floor": { "checked_utc": "2026-10-15T10:33:20Z", "filesystems": { "ollama model store": { "above_floor": true, "df_raw": "Avail Mounted on\n31137951744 /", "df_rc": 0, "free_gib": 29.0, "path": "/workshop/bench-quant-ladder/store", "readable": true }, "results": { "above_floor": true, "df_raw": "Avail Mounted on\n31137951744 /", "df_rc": 0, "free_gib": 29.0, "path": "/workshop/bench-quant-ladder/results/tasks/q-q6_K", "readable": true } }, "floor_gib": 20.0, "floor_source": "BENCH_STORE_MIN_FREE_GIB -- the same floor pull_bank.check_floors gates its pulls with, read through benchlib.params_value", "ok": true, "why": "free 29.0 GiB on /workshop/bench-quant-ladder/results/tasks/q-q6_K (results); free 29.0 GiB on /workshop/bench-quant-ladder/store (ollama model store) -- all at or above the 20.0 GiB floor" }, "ended": "2026-10-15T10:36:35Z", "host": "http://127.0.0.1:withheld-2026-10-16", "idle_preload": { "board": { "cards": [ "" ], "joules": 361.5910344898701, "mean_watts": 25.791086660613125, "peak_watts": 26.01, "per_card": { "": { "joules": 361.5910344898701, "mean_watts": 25.791086660613125, "min_watts": 25.66, "peak_watts": 26.01, "samples": 15, "seconds": 14.019999980926514 } }, "samples": 15, "seconds": 14.019999980926514 }, "wall": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 1531.7080035209656, "mean_watts": 117.76933711972686, "median_s_between_changes": null, "min_watts": 97.0, "peak_watts": 133.0, "samples": 14, "seconds": 13.00600004196167, "value_changes": 1 }, "window_epoch": [ 1792060406.244, 1792060420.264 ] }, "instrument": { "benchlib_py": "28adb0c5036dee50b1315bd38827a68ac20007aff14054a98b33fb32c4da3ab8", "coi3090_py": "f301e824d8a86faeb49ef6cde1398c5b3b307c5fae63d189b32080dcabd8c058", "jev/kit/doorman_planted.json": "4daf059c26a47a84fa3b73630a64ba1a348da72c2e5df7f3c28c007b67da22f7", "jev/kit/task_c.json": "61a0b1d6e7cf66fd91992c0c3a4572f6fbae90e6471048fd7dc1eb41ff2d7f39", "jev/run.py": "withheld-2026-10-16", "run_addenda_py": "a34aca61833ed0c5860c91c7e2e227429e8ea9e2e50fec68d7b7fec15d89acab" }, "limit": null, "model": "mistral-nemo:12b-instruct-2407-q6_K", "num_ctx": 4096, "ollama_version": "0.32.13", "pin": { "api_ps_model": { "context_length": 4096, "details": { "families": [ "llama" ], "family": "llama", "format": "gguf", "parameter_size": "12.2B", "parent_model": "", "quantization_level": "Q6_K" }, "digest": "b6435edbd7917b5503544812b0d283185985367f9c9a5c4dba1ad1cf737cf99c", "expires_at": "2319-01-25T10:21:02.698372691Z", "model": "mistral-nemo:12b-instruct-2407-q6_K", "name": "mistral-nemo:12b-instruct-2407-q6_K", "size": 10290557091, "size_vram": 10290557091 }, "card": "", "cards": [ "" ], "fit": { "growth_bytes": 10612637696.0, "ps_size": 10290557091, "ps_size_vram": 10290557091, "ps_whole": true, "rule": "WHOLE when the named board grew by at least 90 %% of the store's bytes for this tag (weights + projector/draft; KV rides on top) AND /api/ps does not report a CPU share; ps alone is corroboration, never the verdict", "store_bytes": 10056222667, "whole": true }, "floor_mib": 200.0, "growth_mib": { "": 0.0, "": 10121.0 }, "memory_after_mib": { "": 1.0, "": 10122.0 }, "named_grew": true, "ok": true, "others_quiet": true, "per_card_grew": { "": true } }, "postflight": { "after_unload": { "api_ps": { "models": [] }, "rows": [ { "memory_mib": 1.0, "sm_mhz": 210.0, "temp_c": 28.0, "uuid": "", "watts": 23.28 }, { "memory_mib": 1.0, "sm_mhz": 210.0, "temp_c": 59.0, "uuid": "", "watts": 58.8 } ], "stamp": "2026-10-15T10:36:35Z" }, "posture_after": { "boards_on_bus": [ "0, , NVIDIA GeForce RTX 3090, withheld-2026-10-16", "1, , NVIDIA GeForce RTX 3090, withheld-2026-10-16" ], "card": { "clocks.max.sm": "2130", "clocks.mem": "9501", "clocks.sm": "1575", "clocks_throttle_reasons.active": "0x0000000000000004", "driver_version": "595.84", "enforced.power.limit": "300.00", "memory.total": "24576", "memory.used": "10130", "name": "NVIDIA GeForce RTX 3090", "pci.bus_id": "withheld-2026-10-16", "pcie.link.gen.current": "4", "pcie.link.gen.max": "4", "pcie.link.width.current": "16", "pcie.link.width.max": "16", "persistence_mode": "Enabled", "power.default_limit": "350.00", "power.draw": "209.38", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P2", "temperature.gpu": "65", "utilization.gpu": "94", "uuid": "", "vbios_version": "94.02.42.C0.10" }, "compute_apps": [ ", 29550, /usr/local/lib/ollama/llama-server, 10120 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T10:36:29Z" } }, "posture_before": { "boards_on_bus": [ "0, , NVIDIA GeForce RTX 3090, withheld-2026-10-16", "1, , NVIDIA GeForce RTX 3090, withheld-2026-10-16" ], "card": { "clocks.max.sm": "2130", "clocks.mem": "405", "clocks.sm": "210", "clocks_throttle_reasons.active": "0x0000000000000001", "driver_version": "595.84", "enforced.power.limit": "300.00", "memory.total": "24576", "memory.used": "1", "name": "NVIDIA GeForce RTX 3090", "pci.bus_id": "withheld-2026-10-16", "pcie.link.gen.current": "1", "pcie.link.gen.max": "4", "pcie.link.width.current": "16", "pcie.link.width.max": "16", "persistence_mode": "Enabled", "power.default_limit": "350.00", "power.draw": "30.20", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P8", "temperature.gpu": "28", "utilization.gpu": "0", "uuid": "", "vbios_version": "94.02.42.C0.10" }, "compute_apps": [], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T10:33:20Z" }, "posture_verdict": "300.00 W enforced, persistence Enabled, the named card on the bus (2 board(s)), other boards idle", "probes": { "decode": { "board": { "cards": [ "" ], "joules": 3816.9985073041917, "mean_watts": 272.50649786681157, "peak_watts": 305.16, "per_card": { "": { "joules": 3816.9985073041917, "mean_watts": 272.50649786681157, "min_watts": 157.26, "peak_watts": 305.16, "samples": 15, "seconds": 14.006999969482422 } }, "samples": 15, "seconds": 14.006999969482422 }, "num_predict": 256, "prompt": "Describe, in about three hundred words, a lighthouse on a rocky coast during a winter storm.", "runs": [ { "decode_tok_s": 65.7997887209909, "done_reason": "length", "eval_count": 256, "eval_s": 3.89059, "load_s": 0.462019559, "prefill_tok_s": 1044.3864229765013, "prompt_eval_count": 24, "prompt_eval_s": 0.02298, "text_head": "In the heart of a winter storm, the lighthouse stands sentinel, a solitary guardian against the relentless fury of the s", "thinking_chars": 0, "total_s": 4.421407493 }, { "decode_tok_s": 65.94144811541402, "done_reason": "length", "eval_count": 256, "eval_s": 3.882232, "load_s": 0.389150132, "prefill_tok_s": 1665.3944903198944, "prompt_eval_count": 24, "prompt_eval_s": 0.014411, "text_head": "In the heart of a winter storm, the lighthouse stands sentinel, a solitary guardian against the relentless fury of the s", "thinking_chars": 0, "total_s": 4.314952049 }, { "decode_tok_s": 65.78264386612821, "done_reason": "length", "eval_count": 256, "eval_s": 3.891604, "load_s": 0.376284254, "prefill_tok_s": 1628.6644951140065, "prompt_eval_count": 24, "prompt_eval_s": 0.014736, "text_head": "In the heart of a winter storm, the lighthouse stands sentinel, a solitary guardian against the relentless fury of the s", "thinking_chars": 0, "total_s": 4.315266906 } ], "wall": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 5096.912970066071, "mean_watts": 392.04007372362435, "median_s_between_changes": null, "min_watts": 304.0, "peak_watts": 413.0, "samples": 14, "seconds": 13.000999927520752, "value_changes": 1 } }, "prefill": { "board": { "cards": [ "" ], "joules": 2921.2643991315363, "mean_watts": 224.52266267124824, "peak_watts": 262.57, "per_card": { "": { "joules": 2921.2643991315363, "mean_watts": 224.52266267124824, "min_watts": 139.58, "peak_watts": 262.57, "samples": 14, "seconds": 13.011000156402588 } }, "samples": 14, "seconds": 13.011000156402588 }, "runs": [ { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.437992627, "prefill_tok_s": 2890.6841543813666, "prompt_eval_count": 2241, "prompt_eval_s": 0.775249, "text_head": "Under", "thinking_chars": 0, "total_s": 1.245352109 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.432438526, "prefill_tok_s": 2888.5477274779714, "prompt_eval_count": 2239, "prompt_eval_s": 0.77513, "text_head": "Under", "thinking_chars": 0, "total_s": 1.448725849 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.43077868, "prefill_tok_s": 2891.4045255644046, "prompt_eval_count": 2240, "prompt_eval_s": 0.77471, "text_head": "Under", "thinking_chars": 0, "total_s": 1.45204006 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.408298366, "prefill_tok_s": 2878.326458758979, "prompt_eval_count": 2240, "prompt_eval_s": 0.77823, "text_head": "Under", "thinking_chars": 0, "total_s": 1.426141276 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.451698158, "prefill_tok_s": 2870.4323700783675, "prompt_eval_count": 2242, "prompt_eval_s": 0.781067, "text_head": "Under", "thinking_chars": 0, "total_s": 1.47384917 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.43881949, "prefill_tok_s": 2870.244913879822, "prompt_eval_count": 2237, "prompt_eval_s": 0.779376, "text_head": "Under", "thinking_chars": 0, "total_s": 1.460307019 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.396368409, "prefill_tok_s": 2867.9110064563192, "prompt_eval_count": 2241, "prompt_eval_s": 0.781405, "text_head": "Under", "thinking_chars": 0, "total_s": 1.418798879 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.381524548, "prefill_tok_s": 2868.8887094231072, "prompt_eval_count": 2238, "prompt_eval_s": 0.780093, "text_head": "Under", "thinking_chars": 0, "total_s": 1.399757305 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.459428031, "prefill_tok_s": 2876.938659970499, "prompt_eval_count": 2241, "prompt_eval_s": 0.778953, "text_head": "Under", "thinking_chars": 0, "total_s": 1.481863941 } ], "target_tokens_approx": 3000.0, "wall": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 5029.920069694519, "mean_watts": 419.1250729226885, "median_s_between_changes": null, "min_watts": 413.0, "peak_watts": 420.0, "samples": 13, "seconds": 12.001000165939331, "value_changes": 1 } }, "ttft": { "how": "the first 12 task_c prompts, run.py's body with stream true, in task order (so the prompt-cache condition is the task's own); TTFT = client clock on localhost from the request to the first chunk carrying content or thinking", "median_s": 0.5519888050002919, "p95_s": 0.5782116369955475, "runs": [ { "eval_count": 8, "prompt_eval_count": 216, "server_load_ms": 418.675545, "server_prompt_eval_ms": 71.134, "total_s": 0.8194441369996639, "ttft_s": 0.7293168409960344 }, { "eval_count": 8, "prompt_eval_count": 315, "server_load_ms": 396.248171, "server_prompt_eval_ms": 107.859, "total_s": 0.636843886000861, "ttft_s": 0.5300537160001113 }, { "eval_count": 8, "prompt_eval_count": 234, "server_load_ms": 454.317865, "server_prompt_eval_ms": 71.169, "total_s": 0.6686325999980909, "ttft_s": 0.5782116369955475 }, { "eval_count": 3, "prompt_eval_count": 334, "server_load_ms": 355.591245, "server_prompt_eval_ms": 109.061, "total_s": 0.5249350599988247, "ttft_s": 0.49271866100025363 }, { "eval_count": 6, "prompt_eval_count": 343, "server_load_ms": 391.795179, "server_prompt_eval_ms": 117.619, "total_s": 0.6260962000014842, "ttft_s": 0.5648884330003057 }, { "eval_count": 3, "prompt_eval_count": 345, "server_load_ms": 414.364213, "server_prompt_eval_ms": 119.889, "total_s": 0.6100218999999925, "ttft_s": 0.5779167500004405 }, { "eval_count": 8, "prompt_eval_count": 254, "server_load_ms": 419.328649, "server_prompt_eval_ms": 76.738, "total_s": 0.6446423589950427, "ttft_s": 0.5520860440010438 }, { "eval_count": 6, "prompt_eval_count": 232, "server_load_ms": 367.989616, "server_prompt_eval_ms": 71.633, "total_s": 0.5472041259999969, "ttft_s": 0.486934991000453 }, { "eval_count": 8, "prompt_eval_count": 305, "server_load_ms": 440.973341, "server_prompt_eval_ms": 87.936, "total_s": 0.6671187020037905, "ttft_s": 0.5606449390033958 }, { "eval_count": 3, "prompt_eval_count": 340, "server_load_ms": 408.432224, "server_prompt_eval_ms": 109.791, "total_s": 0.5839740669980529, "ttft_s": 0.55189156599954 }, { "eval_count": 8, "prompt_eval_count": 336, "server_load_ms": 382.049175, "server_prompt_eval_ms": 110.67, "total_s": 0.6413549370045075, "ttft_s": 0.5503581390003092 }, { "eval_count": 8, "prompt_eval_count": 333, "server_load_ms": 376.091486, "server_prompt_eval_ms": 109.165, "total_s": 0.631479572999524, "ttft_s": 0.524306245999469 } ] } }, "show": { "capabilities": [ "completion", "tools" ], "details": { "families": [ "llama" ], "family": "llama", "format": "gguf", "parameter_size": "12.2B", "parent_model": "", "quantization_level": "Q6_K" }, "model_info": { "general.architecture": "llama", "general.basename": "Mistral-Nemo", "general.file_type": 18, "general.finetune": "Instruct", "general.languages": null, "general.license": "apache-2.0", "general.parameter_count": 12247782400, "general.quantization_version": 2, "general.size_label": "12B", "general.type": "model", "general.version": "2407", "llama.attention.head_count": 32, "llama.attention.head_count_kv": 8, "llama.attention.key_length": 128, "llama.attention.layer_norm_rms_epsilon": 1e-05, "llama.attention.value_length": 128, "llama.block_count": 40, "llama.context_length": 1024000, "llama.embedding_length": 5120, "llama.feed_forward_length": 14336, "llama.rope.dimension_count": 128, "llama.rope.freq_base": 1000000, "llama.vocab_size": 131072, "tokenizer.ggml.add_bos_token": true, "tokenizer.ggml.add_eos_token": false, "tokenizer.ggml.add_space_prefix": false, "tokenizer.ggml.bos_token_id": 1, "tokenizer.ggml.eos_token_id": 2, "tokenizer.ggml.merges": null, "tokenizer.ggml.model": "gpt2", "tokenizer.ggml.pre": "tekken", "tokenizer.ggml.token_type": null, "tokenizer.ggml.tokens": null, "tokenizer.ggml.unknown_token_id": 0 }, "parameters": "stop \"[INST]\"\nstop \"[/INST]\"", "template_head": "{{- range $i, $_ := .Messages }}\n{{- if eq .Role \"user\" }}\n{{- if and $.Tools (le (len (slice $.Messages $i)) 2) }}[AVAILABLE_TOOLS]{{ $.Tools }}[/AVAILABLE_TOOLS]\n{{- end }}[INST]{{ if and $.System (eq (len (slice $.Messages $i)) 1) }}{{ $.System }}\n\n{{ end }}{{ .Content }}[/INST]\n{{- else if eq .Role \"assistant\" }}\n{{- if .Content }} {{ .Content }}{{ if not (eq (len (slice $.Messages $i)) 1) }}{{ end }}\n{{- else if .ToolCalls }}[TOOL_CALLS][\n{{- range .ToolCalls }}{\"name\": \"{{ .Function.Name }}\", \"arguments\": {{ .Function.Arguments }}}\n{{- end }}]\n{{- end }}\n{{- else if eq .Role \"too" }, "started": "2026-10-15T10:33:20Z", "tag": { "details": { "context_length": 1024000, "embedding_length": 5120, "families": [ "llama" ], "family": "llama", "format": "gguf", "parameter_size": "12.2B", "parent_model": "", "quantization_level": "Q6_K" }, "digest": "b6435edbd7917b5503544812b0d283185985367f9c9a5c4dba1ad1cf737cf99c", "modified_at": "2026-10-15T06:00:50.908003928Z", "name": "mistral-nemo:12b-instruct-2407-q6_K", "size": 10056222667 }, "tasks": [ "c", "d" ], "tasks_run": { "c": { "board_idle_window_epoch": [ 1792060429.107, 1792060438.116 ], "board_task_window_epoch": [ 1792060439.174, 1792060507.191 ], "energy": { "cards": [ "" ], "decisions": 108, "joules": 12675.796804749965, "joules_per_decision": 117.36848893287005, "mean_watts": 186.36218610368843, "net_joules_per_decision": 48.85906667619171, "peak_watts": 204.88, "per_card": { "": { "joules": 12675.796804749965, "mean_watts": 186.36218610368843, "min_watts": 108.63, "peak_watts": 204.88, "samples": 69, "seconds": 68.01699995994568 } }, "samples": 69, "seconds": 68.01699995994568 }, "idle": { "cards": [ "" ], "joules": 980.016000894308, "mean_watts": 108.78188699999184, "peak_watts": 109.44, "per_card": { "": { "joules": 980.016000894308, "mean_watts": 108.78188699999184, "min_watts": 108.62, "peak_watts": 109.44, "samples": 10, "seconds": 9.008999824523926 } }, "samples": 10, "seconds": 9.008999824523926, "what": "the draw of these cards immediately before the arm, with the model resident and no work in flight" }, "posture_end": { "boards_on_bus": [ "0, , NVIDIA GeForce RTX 3090, withheld-2026-10-16", "1, , NVIDIA GeForce RTX 3090, withheld-2026-10-16" ], "card": { "clocks.max.sm": "2130", "clocks.mem": "9501", "clocks.sm": "1950", "clocks_throttle_reasons.active": "0x0000000000000000", "driver_version": "595.84", "enforced.power.limit": "300.00", "memory.total": "24576", "memory.used": "10130", "name": "NVIDIA GeForce RTX 3090", "pci.bus_id": "withheld-2026-10-16", "pcie.link.gen.current": "4", "pcie.link.gen.max": "4", "pcie.link.width.current": "16", "pcie.link.width.max": "16", "persistence_mode": "Enabled", "power.default_limit": "350.00", "power.draw": "156.54", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P2", "temperature.gpu": "50", "utilization.gpu": "0", "uuid": "", "vbios_version": "94.02.42.C0.10" }, "compute_apps": [ ", 29550, /usr/local/lib/ollama/llama-server, 10120 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T10:35:09Z" }, "posture_start": { "boards_on_bus": [ "0, , NVIDIA GeForce RTX 3090, withheld-2026-10-16", "1, , NVIDIA GeForce RTX 3090, withheld-2026-10-16" ], "card": { "clocks.max.sm": "2130", "clocks.mem": "9501", "clocks.sm": "1725", "clocks_throttle_reasons.active": "0x0000000000000000", "driver_version": "595.84", "enforced.power.limit": "300.00", "memory.total": "24576", "memory.used": "10122", "name": "NVIDIA GeForce RTX 3090", "pci.bus_id": "withheld-2026-10-16", "pcie.link.gen.current": "4", "pcie.link.gen.max": "4", "pcie.link.width.current": "16", "pcie.link.width.max": "16", "persistence_mode": "Enabled", "power.default_limit": "350.00", "power.draw": "109.44", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P2", "temperature.gpu": "31", "utilization.gpu": "0", "uuid": "", "vbios_version": "94.02.42.C0.10" }, "compute_apps": [ ", 29550, /usr/local/lib/ollama/llama-server, 10112 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T10:33:48Z" }, "summary": { "accuracy": 0.5092592592592593, "brier": 0.9814814814814815, "correct": 55, "errors": 0, "eval_tokens_total": 831, "floored_rows": 0, "letters_floored_total": 0, "median_seconds": 0.6374609209997288, "n": 108, "n_labelled": 108, "n_reached": 108, "p95_seconds": 0.6837606979970587, "prompt_tokens_total": 30074, "refused": 0, "refused_by_reason": {}, "reliability": [ { "accuracy": null, "correct": 0, "gap": null, "hi": 0.1, "lo": 0.0, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.2, "lo": 0.1, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.3, "lo": 0.2, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.4, "lo": 0.3, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.5, "lo": 0.4, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.6, "lo": 0.5, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.7, "lo": 0.6, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.8, "lo": 0.7, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.9, "lo": 0.8, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": 0.5092592592592593, "correct": 55, "gap": 0.4907407407407407, "hi": 1.0, "lo": 0.9, "mean_p": 1.0, "n": 108, "sum_p": 108.0 } ], "unparsed": 0 }, "wall_idle_resident": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 1456.7279906272888, "mean_watts": 182.0, "median_s_between_changes": null, "min_watts": 182.0, "peak_watts": 182.0, "samples": 9, "seconds": 8.003999948501587, "value_changes": 0 }, "wall_seconds": 68.8320137979972, "wall_task": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 18523.429985046387, "mean_watts": 276.44022256604046, "median_s_between_changes": 13.501500010490417, "min_watts": 182.0, "peak_watts": 305.0, "samples": 68, "seconds": 67.00699996948242, "value_changes": 5 } }, "d": { "board_idle_window_epoch": [ 1792060509.447, 1792060518.456 ], "board_task_window_epoch": [ 1792060519.512, 1792060550.521 ], "energy": { "cards": [ "" ], "decisions": 48, "joules": 5837.286772757769, "joules_per_decision": 121.61014109912018, "mean_watts": 188.2449226285997, "net_joules_per_decision": 45.71348045268641, "peak_watts": 216.43, "per_card": { "": { "joules": 5837.286772757769, "mean_watts": 188.2449226285997, "min_watts": 110.78, "peak_watts": 216.43, "samples": 32, "seconds": 31.008999824523926 } }, "samples": 32, "seconds": 31.008999824523926 }, "idle": { "cards": [ "" ], "joules": 1058.407080901861, "mean_watts": 117.48330264259826, "peak_watts": 156.54, "per_card": { "": { "joules": 1058.407080901861, "mean_watts": 117.48330264259826, "min_watts": 110.72, "peak_watts": 156.54, "samples": 10, "seconds": 9.009000062942505 } }, "samples": 10, "seconds": 9.009000062942505, "what": "the draw of these cards immediately before the arm, with the model resident and no work in flight" }, "posture_end": { "boards_on_bus": [ "0, , NVIDIA GeForce RTX 3090, withheld-2026-10-16", "1, , NVIDIA GeForce RTX 3090, withheld-2026-10-16" ], "card": { "clocks.max.sm": "2130", "clocks.mem": "9501", "clocks.sm": "1980", "clocks_throttle_reasons.active": "0x0000000000000000", "driver_version": "595.84", "enforced.power.limit": "300.00", "memory.total": "24576", "memory.used": "10130", "name": "NVIDIA GeForce RTX 3090", "pci.bus_id": "withheld-2026-10-16", "pcie.link.gen.current": "4", "pcie.link.gen.max": "4", "pcie.link.width.current": "16", "pcie.link.width.max": "16", "persistence_mode": "Enabled", "power.default_limit": "350.00", "power.draw": "160.54", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P2", "temperature.gpu": "57", "utilization.gpu": "0", "uuid": "", "vbios_version": "94.02.42.C0.10" }, "compute_apps": [ ", 29550, /usr/local/lib/ollama/llama-server, 10120 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T10:35:52Z" }, "posture_start": { "boards_on_bus": [ "0, , NVIDIA GeForce RTX 3090, withheld-2026-10-16", "1, , NVIDIA GeForce RTX 3090, withheld-2026-10-16" ], "card": { "clocks.max.sm": "2130", "clocks.mem": "9501", "clocks.sm": "1950", "clocks_throttle_reasons.active": "0x0000000000000000", "driver_version": "595.84", "enforced.power.limit": "300.00", "memory.total": "24576", "memory.used": "10130", "name": "NVIDIA GeForce RTX 3090", "pci.bus_id": "withheld-2026-10-16", "pcie.link.gen.current": "4", "pcie.link.gen.max": "4", "pcie.link.width.current": "16", "pcie.link.width.max": "16", "persistence_mode": "Enabled", "power.default_limit": "350.00", "power.draw": "156.54", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P2", "temperature.gpu": "50", "utilization.gpu": "0", "uuid": "", "vbios_version": "94.02.42.C0.10" }, "compute_apps": [ ", 29550, /usr/local/lib/ollama/llama-server, 10120 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T10:35:09Z" }, "summary": { "accuracy": 0.7083333333333334, "brier": 0.5520833333333334, "correct": 34, "errors": 0, "eval_tokens_total": 126, "floored_rows": 0, "letters_floored_total": 0, "median_seconds": 0.6580668959977629, "n": 48, "n_labelled": 48, "n_reached": 48, "p95_seconds": 0.7559774590044981, "prompt_tokens_total": 23588, "refused": 0, "refused_by_reason": {}, "reliability": [ { "accuracy": null, "correct": 0, "gap": null, "hi": 0.1, "lo": 0.0, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.2, "lo": 0.1, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.3, "lo": 0.2, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.4, "lo": 0.3, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.5, "lo": 0.4, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": 0.0, "correct": 0, "gap": 0.5, "hi": 0.6, "lo": 0.5, "mean_p": 0.5, "n": 1, "sum_p": 0.5 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.7, "lo": 0.6, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.8, "lo": 0.7, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": null, "correct": 0, "gap": null, "hi": 0.9, "lo": 0.8, "mean_p": null, "n": 0, "sum_p": 0.0 }, { "accuracy": 0.723404255319149, "correct": 34, "gap": 0.276595744680851, "hi": 1.0, "lo": 0.9, "mean_p": 1.0, "n": 47, "sum_p": 47.0 } ], "unparsed": 1 }, "wall_idle_resident": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 1862.168951511383, "mean_watts": 232.68386881240016, "median_s_between_changes": null, "min_watts": 192.0, "peak_watts": 285.0, "samples": 9, "seconds": 8.002999782562256, "value_changes": 1 }, "wall_seconds": 31.88913925000088, "wall_task": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 7622.752991199493, "mean_watts": 254.06635955967633, "median_s_between_changes": 14.000999927520752, "min_watts": 192.0, "peak_watts": 304.0, "samples": 31, "seconds": 30.003000020980835, "value_changes": 2 } } }, "think": "off", "think_budget": 8, "ups_before": { "available": true, "error": null, "load_pct": 9.0, "read_utc": "2026-10-15T10:33:20Z", "ups.realpower": 98.0, "ups.status": "OL" }, "ups_stop": null, "util_before_pct": [ 0.0, 0.0, 0.0 ], "wall_unreadable_polls": 0, "warm": { "done_reason": "length", "load_ms": 4467.936207, "seconds": 4.50167119300022 } }