{ "argv": [ "/workshop/bench-quant-ladder/coi_quant.py", "--arm-id", "q-q4_K_M", "--model", "mistral-nemo:12b-instruct-2407-q4_K_M", "--num-ctx", "4096", "--think", "off", "--tasks", "c,d", "--probes", "--out", "/workshop/bench-quant-ladder/results/tasks/q-q4_K_M" ], "arm_id": "q-q4_K_M", "baseline_rows": [ { "memory_mib": 1.0, "sm_mhz": 210.0, "temp_c": 27.0, "uuid": "", "watts": 23.17 }, { "memory_mib": 1.0, "sm_mhz": 210.0, "temp_c": 53.0, "uuid": "", "watts": 27.73 } ], "board": "", "box": "benchbox", "box_class": "gpu-ql3090-24g", "box_class_check": "gpu-ql3090-24g", "disk_floor": { "checked_utc": "2026-10-15T05:41:44Z", "filesystems": { "ollama model store": { "above_floor": true, "df_raw": "Avail Mounted on\n33732165632 /", "df_rc": 0, "free_gib": 31.42, "path": "/workshop/bench-quant-ladder/store", "readable": true }, "results": { "above_floor": true, "df_raw": "Avail Mounted on\n33732165632 /", "df_rc": 0, "free_gib": 31.42, "path": "/workshop/bench-quant-ladder/results/tasks/q-q4_K_M", "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 31.42 GiB on /workshop/bench-quant-ladder/results/tasks/q-q4_K_M (results); free 31.42 GiB on /workshop/bench-quant-ladder/store (ollama model store) -- all at or above the 20.0 GiB floor" }, "ended": "2026-10-15T05:44:55Z", "host": "http://127.0.0.1:withheld-2026-10-16", "idle_preload": { "board": { "cards": [ "" ], "joules": 392.94713947534564, "mean_watts": 28.02761341012339, "peak_watts": 28.28, "per_card": { "": { "joules": 392.94713947534564, "mean_watts": 28.02761341012339, "min_watts": 27.75, "peak_watts": 28.28, "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": 1352.6240043640137, "mean_watts": 104.0, "median_s_between_changes": null, "min_watts": 104.0, "peak_watts": 104.0, "samples": 14, "seconds": 13.00600004196167, "value_changes": 0 }, "window_epoch": [ 1792042910.26, 1792042924.28 ] }, "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-q4_K_M", "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": "Q4_K_M" }, "digest": "daf6737417121831e572a9c482e92a221ee0c33537f35f1f857c7b4f7191df55", "expires_at": "2319-01-25T05:29:26.384368592Z", "model": "mistral-nemo:12b-instruct-2407-q4_K_M", "name": "mistral-nemo:12b-instruct-2407-q4_K_M", "size": 7884568001, "size_vram": 7884568001 }, "card": "", "cards": [ "" ], "fit": { "growth_bytes": 8207204352.0, "ps_size": 7884568001, "ps_size_vram": 7884568001, "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": 7477217229, "whole": true }, "floor_mib": 200.0, "growth_mib": { "": 0.0, "": 7827.0 }, "memory_after_mib": { "": 1.0, "": 7828.0 }, "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": 60.0, "uuid": "", "watts": 50.63 } ], "stamp": "2026-10-15T05:44:55Z" }, "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": "1635", "clocks_throttle_reasons.active": "0x0000000000000004", "driver_version": "595.84", "enforced.power.limit": "300.00", "memory.total": "24576", "memory.used": "7838", "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": "199.62", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P2", "temperature.gpu": "65", "utilization.gpu": "86", "uuid": "", "vbios_version": "94.02.42.C0.10" }, "compute_apps": [ ", 11628, /usr/local/lib/ollama/llama-server, 7828 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T05:44:49Z" } }, "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": "5001", "clocks.sm": "270", "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": "2", "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": "79.58", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P5", "temperature.gpu": "59", "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": "2026-10-15T05:41:44Z" }, "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": 2928.4092185664176, "mean_watts": 265.97722262515657, "peak_watts": 301.27, "per_card": { "": { "joules": 2928.4092185664176, "mean_watts": 265.97722262515657, "min_watts": 156.9, "peak_watts": 301.27, "samples": 12, "seconds": 11.009999990463257 } }, "samples": 12, "seconds": 11.009999990463257 }, "num_predict": 256, "prompt": "Describe, in about three hundred words, a lighthouse on a rocky coast during a winter storm.", "runs": [ { "decode_tok_s": 87.29898641102614, "done_reason": "length", "eval_count": 256, "eval_s": 2.932451, "load_s": 0.430624332, "prefill_tok_s": 1209.2507683780925, "prompt_eval_count": 24, "prompt_eval_s": 0.019847, "text_head": "On a jagged, icy coast, a solitary lighthouse stands sentinel against the winter storm's fury. Its towering form, a beac", "thinking_chars": 0, "total_s": 3.429181279 }, { "decode_tok_s": 87.21725511626299, "done_reason": "length", "eval_count": 256, "eval_s": 2.935199, "load_s": 0.431465345, "prefill_tok_s": 2093.3275185346706, "prompt_eval_count": 24, "prompt_eval_s": 0.011465, "text_head": "On a jagged, icy coast, a solitary lighthouse stands sentinel against the winter storm's fury. Its towering form, a beac", "thinking_chars": 0, "total_s": 3.406480398 }, { "decode_tok_s": 86.99579395725857, "done_reason": "length", "eval_count": 256, "eval_s": 2.942671, "load_s": 0.442538288, "prefill_tok_s": 2074.868159419037, "prompt_eval_count": 24, "prompt_eval_s": 0.011567, "text_head": "On a jagged, icy coast, a solitary lighthouse stands sentinel against the winter storm's fury. Its towering form, a beac", "thinking_chars": 0, "total_s": 3.401151334 } ], "wall": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 2913.7849793434143, "mean_watts": 291.34936510950877, "median_s_between_changes": null, "min_watts": 285.0, "peak_watts": 412.0, "samples": 11, "seconds": 10.000999927520752, "value_changes": 1 } }, "prefill": { "board": { "cards": [ "" ], "joules": 3145.0530601346495, "mean_watts": 224.4702789625867, "peak_watts": 250.96, "per_card": { "": { "joules": 3145.0530601346495, "mean_watts": 224.4702789625867, "min_watts": 152.02, "peak_watts": 250.96, "samples": 15, "seconds": 14.010999917984009 } }, "samples": 15, "seconds": 14.010999917984009 }, "runs": [ { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.418061132, "prefill_tok_s": 3215.7016930863133, "prompt_eval_count": 2241, "prompt_eval_s": 0.696893, "text_head": "Under", "thinking_chars": 0, "total_s": 1.145549436 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.418145102, "prefill_tok_s": 3224.5010232282166, "prompt_eval_count": 2239, "prompt_eval_s": 0.694371, "text_head": "Under", "thinking_chars": 0, "total_s": 1.355973197 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.399022319, "prefill_tok_s": 3222.576449871672, "prompt_eval_count": 2240, "prompt_eval_s": 0.695096, "text_head": "Under", "thinking_chars": 0, "total_s": 1.335426647 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.383015149, "prefill_tok_s": 3212.621011115095, "prompt_eval_count": 2240, "prompt_eval_s": 0.69725, "text_head": "Under", "thinking_chars": 0, "total_s": 1.316308794 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.410451249, "prefill_tok_s": 3228.277522433793, "prompt_eval_count": 2242, "prompt_eval_s": 0.694488, "text_head": "Under", "thinking_chars": 0, "total_s": 1.343985402 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.406918009, "prefill_tok_s": 3217.3394026995734, "prompt_eval_count": 2237, "prompt_eval_s": 0.695295, "text_head": "Under", "thinking_chars": 0, "total_s": 1.343095089 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.406384775, "prefill_tok_s": 3219.9478716158937, "prompt_eval_count": 2241, "prompt_eval_s": 0.695974, "text_head": "Under", "thinking_chars": 0, "total_s": 1.343582565 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.435129043, "prefill_tok_s": 3222.629250760299, "prompt_eval_count": 2238, "prompt_eval_s": 0.694464, "text_head": "Under", "thinking_chars": 0, "total_s": 1.373058658 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.445231625, "prefill_tok_s": 3225.8946783533975, "prompt_eval_count": 2241, "prompt_eval_s": 0.694691, "text_head": "Under", "thinking_chars": 0, "total_s": 1.377210701 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.427272902, "prefill_tok_s": 3223.6199236980674, "prompt_eval_count": 2240, "prompt_eval_s": 0.694871, "text_head": "Under", "thinking_chars": 0, "total_s": 1.362118013 } ], "target_tokens_approx": 3000.0, "wall": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 5105.912068367004, "mean_watts": 392.7322516112051, "median_s_between_changes": null, "min_watts": 245.0, "peak_watts": 412.0, "samples": 14, "seconds": 13.001000165939331, "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": 0.5628708500025823, "p95_s": 0.5982682950016169, "runs": [ { "eval_count": 8, "prompt_eval_count": 216, "server_load_ms": 438.102066, "server_prompt_eval_ms": 62.174, "total_s": 0.8008645020017866, "ttft_s": 0.7317075470018608 }, { "eval_count": 8, "prompt_eval_count": 315, "server_load_ms": 399.146767, "server_prompt_eval_ms": 94.517, "total_s": 0.602228630999889, "ttft_s": 0.5196954129969527 }, { "eval_count": 8, "prompt_eval_count": 234, "server_load_ms": 365.133405, "server_prompt_eval_ms": 63.881, "total_s": 0.5487745579994225, "ttft_s": 0.47953836400120053 }, { "eval_count": 8, "prompt_eval_count": 334, "server_load_ms": 424.427767, "server_prompt_eval_ms": 95.203, "total_s": 0.6329637120034022, "ttft_s": 0.5631044530018698 }, { "eval_count": 8, "prompt_eval_count": 343, "server_load_ms": 439.323026, "server_prompt_eval_ms": 104.996, "total_s": 0.6681355040018389, "ttft_s": 0.5982682950016169 }, { "eval_count": 8, "prompt_eval_count": 345, "server_load_ms": 416.0667, "server_prompt_eval_ms": 105.941, "total_s": 0.6458658800002013, "ttft_s": 0.5634694910004328 }, { "eval_count": 8, "prompt_eval_count": 254, "server_load_ms": 442.199182, "server_prompt_eval_ms": 68.489, "total_s": 0.6346829370013438, "ttft_s": 0.5626372470032948 }, { "eval_count": 8, "prompt_eval_count": 232, "server_load_ms": 445.385803, "server_prompt_eval_ms": 62.474, "total_s": 0.6226579929971194, "ttft_s": 0.5531143909975071 }, { "eval_count": 8, "prompt_eval_count": 305, "server_load_ms": 427.330122, "server_prompt_eval_ms": 77.079, "total_s": 0.617259919999924, "ttft_s": 0.5352540370004135 }, { "eval_count": 8, "prompt_eval_count": 340, "server_load_ms": 410.864916, "server_prompt_eval_ms": 95.074, "total_s": 0.62653724899792, "ttft_s": 0.5439082319971931 }, { "eval_count": 8, "prompt_eval_count": 336, "server_load_ms": 420.979387, "server_prompt_eval_ms": 95.375, "total_s": 0.6387818809998862, "ttft_s": 0.569019702001242 }, { "eval_count": 8, "prompt_eval_count": 333, "server_load_ms": 444.61756, "server_prompt_eval_ms": 94.801, "total_s": 0.6605035719985608, "ttft_s": 0.5779171639987908 } ] } }, "show": { "capabilities": [ "completion", "tools" ], "details": { "families": [ "llama" ], "family": "llama", "format": "gguf", "parameter_size": "12.2B", "parent_model": "", "quantization_level": "Q4_K_M" }, "model_info": { "general.architecture": "llama", "general.basename": "Mistral-Nemo", "general.file_type": 15, "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": "2026-10-15T05:41:44Z", "tag": { "details": { "context_length": 1024000, "embedding_length": 5120, "families": [ "llama" ], "family": "llama", "format": "gguf", "parameter_size": "12.2B", "parent_model": "", "quantization_level": "Q4_K_M" }, "digest": "daf6737417121831e572a9c482e92a221ee0c33537f35f1f857c7b4f7191df55", "modified_at": "2026-10-15T05:21:06.358180485Z", "name": "mistral-nemo:12b-instruct-2407-q4_K_M", "size": 7477217229 }, "tasks": [ "c", "d" ], "tasks_run": { "c": { "board_idle_window_epoch": [ 1792042932.788, 1792042941.795 ], "board_task_window_epoch": [ 1792042942.857, 1792043008.873 ], "energy": { "cards": [ "" ], "decisions": 108, "joules": 12360.668367311955, "joules_per_decision": 114.45063303066625, "mean_watts": 187.23746373427176, "net_joules_per_decision": 44.033027744154836, "peak_watts": 202.77, "per_card": { "": { "joules": 12360.668367311955, "mean_watts": 187.23746373427176, "min_watts": 114.89, "peak_watts": 202.77, "samples": 67, "seconds": 66.01599979400635 } }, "samples": 67, "seconds": 66.01599979400635 }, "idle": { "cards": [ "" ], "joules": 1037.6143363690376, "mean_watts": 115.20088152378034, "peak_watts": 117.55, "per_card": { "": { "joules": 1037.6143363690376, "mean_watts": 115.20088152378034, "min_watts": 114.44, "peak_watts": 117.55, "samples": 10, "seconds": 9.006999969482422 } }, "samples": 10, "seconds": 9.006999969482422, "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": "1965", "clocks_throttle_reasons.active": "0x0000000000000000", "driver_version": "595.84", "enforced.power.limit": "300.00", "memory.total": "24576", "memory.used": "7838", "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": "172.00", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P2", "temperature.gpu": "63", "utilization.gpu": "0", "uuid": "", "vbios_version": "94.02.42.C0.10" }, "compute_apps": [ ", 11628, /usr/local/lib/ollama/llama-server, 7828 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T05:43:30Z" }, "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": "7828", "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": "117.55", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P2", "temperature.gpu": "56", "utilization.gpu": "0", "uuid": "", "vbios_version": "94.02.42.C0.10" }, "compute_apps": [ ", 11628, /usr/local/lib/ollama/llama-server, 7818 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T05:42:12Z" }, "summary": { "accuracy": 0.4351851851851852, "brier": 1.1296296296296295, "correct": 47, "errors": 0, "eval_tokens_total": 850, "floored_rows": 0, "letters_floored_total": 0, "median_seconds": 0.6148399154990329, "n": 108, "n_labelled": 108, "n_reached": 108, "p95_seconds": 0.6532226240015007, "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": 0.4351851851851852, "correct": 47, "gap": 0.5648148148148149, "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": 1138.693974018097, "mean_watts": 142.24784405417265, "median_s_between_changes": null, "min_watts": 102.0, "peak_watts": 194.0, "samples": 9, "seconds": 8.004999876022339, "value_changes": 1 }, "wall_seconds": 66.43111679899812, "wall_task": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 17834.643951654434, "mean_watts": 274.35381357956396, "median_s_between_changes": 14.001000046730042, "min_watts": 194.0, "peak_watts": 308.0, "samples": 66, "seconds": 65.00599980354309, "value_changes": 5 } }, "d": { "board_idle_window_epoch": [ 1792043011.133, 1792043020.142 ], "board_task_window_epoch": [ 1792043021.201, 1792043052.21 ], "energy": { "cards": [ "" ], "decisions": 48, "joules": 5950.453790158033, "joules_per_decision": 123.96778729495902, "mean_watts": 191.89441059304454, "net_joules_per_decision": 40.9222930215787, "peak_watts": 218.81, "per_card": { "": { "joules": 5950.453790158033, "mean_watts": 191.89441059304454, "min_watts": 120.95, "peak_watts": 218.81, "samples": 32, "seconds": 31.009000062942505 } }, "samples": 32, "seconds": 31.009000062942505 }, "idle": { "cards": [ "" ], "joules": 1158.1002082502841, "mean_watts": 128.5492507669078, "peak_watts": 172.0, "per_card": { "": { "joules": 1158.1002082502841, "mean_watts": 128.5492507669078, "min_watts": 121.04, "peak_watts": 172.0, "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": "1965", "clocks_throttle_reasons.active": "0x0000000000000000", "driver_version": "595.84", "enforced.power.limit": "300.00", "memory.total": "24576", "memory.used": "7838", "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": "158.37", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P2", "temperature.gpu": "62", "utilization.gpu": "0", "uuid": "", "vbios_version": "94.02.42.C0.10" }, "compute_apps": [ ", 11628, /usr/local/lib/ollama/llama-server, 7828 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T05:44:14Z" }, "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": "1965", "clocks_throttle_reasons.active": "0x0000000000000000", "driver_version": "595.84", "enforced.power.limit": "300.00", "memory.total": "24576", "memory.used": "7838", "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": "172.00", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P2", "temperature.gpu": "63", "utilization.gpu": "0", "uuid": "", "vbios_version": "94.02.42.C0.10" }, "compute_apps": [ ", 11628, /usr/local/lib/ollama/llama-server, 7828 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T05:43:30Z" }, "summary": { "accuracy": 0.6875, "brier": 0.5625, "correct": 33, "errors": 0, "eval_tokens_total": 222, "floored_rows": 0, "letters_floored_total": 0, "median_seconds": 0.6482231039990438, "n": 48, "n_labelled": 48, "n_reached": 48, "p95_seconds": 0.7292413700015459, "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": 2, "sum_p": 1.0 }, { "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": 0.717391304347826, "correct": 33, "gap": 0.28260869565217395, "hi": 1.0, "lo": 0.9, "mean_p": 1.0, "n": 46, "sum_p": 46.0 } ], "unparsed": 2 }, "wall_idle_resident": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 2465.5400352478027, "mean_watts": 308.0, "median_s_between_changes": null, "min_watts": 308.0, "peak_watts": 308.0, "samples": 9, "seconds": 8.005000114440918, "value_changes": 0 }, "wall_seconds": 31.6072231750004, "wall_task": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 8086.238946437836, "mean_watts": 269.51434873314093, "median_s_between_changes": 14.001499891281128, "min_watts": 203.0, "peak_watts": 333.0, "samples": 31, "seconds": 30.002999782562256, "value_changes": 3 } } }, "think": "off", "think_budget": 8, "ups_before": { "available": true, "error": null, "load_pct": 33.0, "read_utc": "2026-10-15T05:41:44Z", "ups.realpower": 337.0, "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": 4120.658297, "seconds": 4.169180521999806 } }