{ "argv": [ "/workshop/bench-quant-ladder/coi_quant.py", "--arm-id", "q-q3_K_M", "--model", "mistral-nemo:12b-instruct-2407-q3_K_M", "--num-ctx", "4096", "--think", "off", "--tasks", "c,d", "--probes", "--out", "/workshop/bench-quant-ladder/results/tasks/q-q3_K_M" ], "arm_id": "q-q3_K_M", "baseline_rows": [ { "memory_mib": 1.0, "sm_mhz": 210.0, "temp_c": 29.0, "uuid": "", "watts": 23.24 }, { "memory_mib": 1.0, "sm_mhz": 210.0, "temp_c": 53.0, "uuid": "", "watts": 27.98 } ], "board": "", "box": "benchbox", "box_class": "gpu-ql3090-24g", "box_class_check": "gpu-ql3090-24g", "disk_floor": { "checked_utc": "2026-10-15T10:58:41Z", "filesystems": { "ollama model store": { "above_floor": true, "df_raw": "Avail Mounted on\n35096117248 /", "df_rc": 0, "free_gib": 32.69, "path": "/workshop/bench-quant-ladder/store", "readable": true }, "results": { "above_floor": true, "df_raw": "Avail Mounted on\n35096117248 /", "df_rc": 0, "free_gib": 32.69, "path": "/workshop/bench-quant-ladder/results/tasks/q-q3_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 32.69 GiB on /workshop/bench-quant-ladder/results/tasks/q-q3_K_M (results); free 32.69 GiB on /workshop/bench-quant-ladder/store (ollama model store) -- all at or above the 20.0 GiB floor" }, "ended": "2026-10-15T11:01:53Z", "host": "http://127.0.0.1:withheld-2026-10-16", "idle_preload": { "board": { "cards": [ "" ], "joules": 395.0088771033287, "mean_watts": 28.170651476808573, "peak_watts": 28.5, "per_card": { "": { "joules": 395.0088771033287, "mean_watts": 28.170651476808573, "min_watts": 27.99, "peak_watts": 28.5, "samples": 15, "seconds": 14.022000074386597 } }, "samples": 15, "seconds": 14.022000074386597 }, "wall": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 1371.6340050697327, "mean_watts": 105.4616331419642, "median_s_between_changes": null, "min_watts": 99.0, "peak_watts": 107.0, "samples": 14, "seconds": 13.00600004196167, "value_changes": 1 }, "window_epoch": [ 1792061927.12, 1792061941.142 ] }, "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-q3_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": "Q3_K_M" }, "digest": "0549a17f596fcb2387d26e5eec0b167b8d912517398a88bff2d7745e275996d1", "expires_at": "2319-01-25T10:46:22.931490548Z", "model": "mistral-nemo:12b-instruct-2407-q3_K_M", "name": "mistral-nemo:12b-instruct-2407-q3_K_M", "size": 6579583712, "size_vram": 6579583712 }, "card": "", "cards": [ "" ], "fit": { "growth_bytes": 6902775808.0, "ps_size": 6579583712, "ps_size_vram": 6579583712, "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": 6083102669, "whole": true }, "floor_mib": 200.0, "growth_mib": { "": 0.0, "": 6583.0 }, "memory_after_mib": { "": 1.0, "": 6584.0 }, "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": 29.0, "uuid": "", "watts": 23.32 }, { "memory_mib": 1.0, "sm_mhz": 210.0, "temp_c": 60.0, "uuid": "", "watts": 46.95 } ], "stamp": "2026-10-15T11:01:53Z" }, "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": "1545", "clocks_throttle_reasons.active": "0x0000000000000004", "driver_version": "595.84", "enforced.power.limit": "300.00", "memory.total": "24576", "memory.used": "6594", "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": "207.03", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P2", "temperature.gpu": "66", "utilization.gpu": "93", "uuid": "", "vbios_version": "94.02.42.C0.10" }, "compute_apps": [ ", 42912, /usr/local/lib/ollama/llama-server, 6584 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T11:01:46Z" } }, "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": "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": "79.53", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P3", "temperature.gpu": "58", "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": "2026-10-15T10:58:41Z" }, "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": 3278.3485098934175, "mean_watts": 273.0364385963045, "peak_watts": 300.43, "per_card": { "": { "joules": 3278.3485098934175, "mean_watts": 273.0364385963045, "min_watts": 160.89, "peak_watts": 300.43, "samples": 13, "seconds": 12.006999969482422 } }, "samples": 13, "seconds": 12.006999969482422 }, "num_predict": 256, "prompt": "Describe, in about three hundred words, a lighthouse on a rocky coast during a winter storm.", "runs": [ { "decode_tok_s": 74.36315654170654, "done_reason": "length", "eval_count": 256, "eval_s": 3.442565, "load_s": 0.399897787, "prefill_tok_s": 1141.389641889, "prompt_eval_count": 24, "prompt_eval_s": 0.021027, "text_head": "In the grip of a winter storm, the lighthouse stands sentinel against the relentless onslaught of the sea. Its towering ", "thinking_chars": 0, "total_s": 3.910172819 }, { "decode_tok_s": 74.17395995649234, "done_reason": "length", "eval_count": 256, "eval_s": 3.451346, "load_s": 0.371299494, "prefill_tok_s": 1931.58953722334, "prompt_eval_count": 24, "prompt_eval_s": 0.012425, "text_head": "In the grip of a winter storm, the lighthouse stands sentinel against the relentless onslaught of the sea. Its towering ", "thinking_chars": 0, "total_s": 3.863217285 }, { "decode_tok_s": 74.1799780125909, "done_reason": "length", "eval_count": 256, "eval_s": 3.451066, "load_s": 0.414057025, "prefill_tok_s": 1887.8313537323997, "prompt_eval_count": 24, "prompt_eval_s": 0.012713, "text_head": "In the grip of a winter storm, the lighthouse stands sentinel against the relentless onslaught of the sea. Its towering ", "thinking_chars": 0, "total_s": 3.910642325 } ], "wall": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 3767.9089703559875, "mean_watts": 342.5060444669183, "median_s_between_changes": null, "min_watts": 276.0, "peak_watts": 409.0, "samples": 12, "seconds": 11.000999927520752, "value_changes": 1 } }, "prefill": { "board": { "cards": [ "" ], "joules": 2957.631635453701, "mean_watts": 227.38768264623468, "peak_watts": 257.83, "per_card": { "": { "joules": 2957.631635453701, "mean_watts": 227.38768264623468, "min_watts": 165.53, "peak_watts": 257.83, "samples": 14, "seconds": 13.007000207901001 } }, "samples": 14, "seconds": 13.007000207901001 }, "runs": [ { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.412050801, "prefill_tok_s": 2975.4357267378946, "prompt_eval_count": 2241, "prompt_eval_s": 0.753167, "text_head": "Under", "thinking_chars": 0, "total_s": 1.195758225 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.437133816, "prefill_tok_s": 2982.8516008014667, "prompt_eval_count": 2239, "prompt_eval_s": 0.750624, "text_head": "Under", "thinking_chars": 0, "total_s": 1.434307706 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.433304359, "prefill_tok_s": 2987.550981757267, "prompt_eval_count": 2240, "prompt_eval_s": 0.749778, "text_head": "Under", "thinking_chars": 0, "total_s": 1.420093569 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.439498045, "prefill_tok_s": 2981.44714652881, "prompt_eval_count": 2240, "prompt_eval_s": 0.751313, "text_head": "Under", "thinking_chars": 0, "total_s": 1.427449085 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.370725729, "prefill_tok_s": 2984.6653167778927, "prompt_eval_count": 2242, "prompt_eval_s": 0.751173, "text_head": "Under", "thinking_chars": 0, "total_s": 1.36090451 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.417892802, "prefill_tok_s": 2982.3763820321487, "prompt_eval_count": 2237, "prompt_eval_s": 0.750073, "text_head": "Under", "thinking_chars": 0, "total_s": 1.407840545 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.441081932, "prefill_tok_s": 2986.116755699406, "prompt_eval_count": 2241, "prompt_eval_s": 0.750473, "text_head": "Under", "thinking_chars": 0, "total_s": 1.431084024 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.384612409, "prefill_tok_s": 2978.602771777539, "prompt_eval_count": 2238, "prompt_eval_s": 0.751359, "text_head": "Under", "thinking_chars": 0, "total_s": 1.374477702 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.41107481, "prefill_tok_s": 2983.6994944619823, "prompt_eval_count": 2241, "prompt_eval_s": 0.751081, "text_head": "Under", "thinking_chars": 0, "total_s": 1.401938615 } ], "target_tokens_approx": 3000.0, "wall": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 4349.707019805908, "mean_watts": 362.4151796312744, "median_s_between_changes": 3.0005000829696655, "min_watts": 298.0, "peak_watts": 415.0, "samples": 13, "seconds": 12.002000093460083, "value_changes": 3 } }, "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": 0.5551731219966314, "p95_s": 0.611813447001623, "runs": [ { "eval_count": 8, "prompt_eval_count": 216, "server_load_ms": 384.905896, "server_prompt_eval_ms": 67.393, "total_s": 0.7738114090025192, "ttft_s": 0.6929719470062992 }, { "eval_count": 8, "prompt_eval_count": 315, "server_load_ms": 390.434269, "server_prompt_eval_ms": 99.566, "total_s": 0.6133644849978737, "ttft_s": 0.5315593310006079 }, { "eval_count": 8, "prompt_eval_count": 234, "server_load_ms": 414.759309, "server_prompt_eval_ms": 67.244, "total_s": 0.6158652520025498, "ttft_s": 0.534927881002659 }, { "eval_count": 8, "prompt_eval_count": 334, "server_load_ms": 389.603806, "server_prompt_eval_ms": 100.657, "total_s": 0.6146815730025992, "ttft_s": 0.5325536220043432 }, { "eval_count": 8, "prompt_eval_count": 343, "server_load_ms": 442.42648, "server_prompt_eval_ms": 111.113, "total_s": 0.692649539996637, "ttft_s": 0.610884684996563 }, { "eval_count": 8, "prompt_eval_count": 345, "server_load_ms": 442.742717, "server_prompt_eval_ms": 113.293, "total_s": 0.6935123820003355, "ttft_s": 0.611813447001623 }, { "eval_count": 8, "prompt_eval_count": 254, "server_load_ms": 399.064525, "server_prompt_eval_ms": 74.785, "total_s": 0.6114469000021927, "ttft_s": 0.5281612489998224 }, { "eval_count": 8, "prompt_eval_count": 232, "server_load_ms": 419.547697, "server_prompt_eval_ms": 65.971, "total_s": 0.6137319909976213, "ttft_s": 0.5325474920027773 }, { "eval_count": 8, "prompt_eval_count": 305, "server_load_ms": 425.791173, "server_prompt_eval_ms": 84.086, "total_s": 0.6384074779998627, "ttft_s": 0.5429461029998492 }, { "eval_count": 8, "prompt_eval_count": 340, "server_load_ms": 428.415541, "server_prompt_eval_ms": 102.101, "total_s": 0.6639756569929887, "ttft_s": 0.5674001409934135 }, { "eval_count": 8, "prompt_eval_count": 336, "server_load_ms": 421.353372, "server_prompt_eval_ms": 101.511, "total_s": 0.6591095690018847, "ttft_s": 0.5770658070032368 }, { "eval_count": 8, "prompt_eval_count": 333, "server_load_ms": 452.735481, "server_prompt_eval_ms": 99.646, "total_s": 0.68794827000238, "ttft_s": 0.6059599279979011 } ] } }, "show": { "capabilities": [ "completion", "tools" ], "details": { "families": [ "llama" ], "family": "llama", "format": "gguf", "parameter_size": "12.2B", "parent_model": "", "quantization_level": "Q3_K_M" }, "model_info": { "general.architecture": "llama", "general.basename": "Mistral-Nemo", "general.file_type": 12, "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": "2026-10-15T10:58:41Z", "tag": { "details": { "context_length": 1024000, "embedding_length": 5120, "families": [ "llama" ], "family": "llama", "format": "gguf", "parameter_size": "12.2B", "parent_model": "", "quantization_level": "Q3_K_M" }, "digest": "0549a17f596fcb2387d26e5eec0b167b8d912517398a88bff2d7745e275996d1", "modified_at": "2026-10-15T10:50:34.863312983Z", "name": "mistral-nemo:12b-instruct-2407-q3_K_M", "size": 6083102669 }, "tasks": [ "c", "d" ], "tasks_run": { "c": { "board_idle_window_epoch": [ 1792061949.332, 1792061958.341 ], "board_task_window_epoch": [ 1792061959.4, 1792062026.415 ], "energy": { "cards": [ "" ], "decisions": 108, "joules": 12828.076293174028, "joules_per_decision": 118.77848419605581, "mean_watts": 191.42097021161658, "net_joules_per_decision": 47.48011355105751, "peak_watts": 208.55, "per_card": { "": { "joules": 12828.076293174028, "mean_watts": 191.42097021161658, "min_watts": 114.61, "peak_watts": 208.55, "samples": 68, "seconds": 67.0149998664856 } }, "samples": 68, "seconds": 67.0149998664856 }, "idle": { "cards": [ "" ], "joules": 1035.1610670161247, "mean_watts": 114.90299253899906, "peak_watts": 116.82, "per_card": { "": { "joules": 1035.1610670161247, "mean_watts": 114.90299253899906, "min_watts": 114.37, "peak_watts": 116.82, "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": "1950", "clocks_throttle_reasons.active": "0x0000000000000000", "driver_version": "595.84", "enforced.power.limit": "300.00", "memory.total": "24576", "memory.used": "6594", "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": "161.65", "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": [ ", 42912, /usr/local/lib/ollama/llama-server, 6584 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T11:00:28Z" }, "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": "6584", "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": "116.82", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P2", "temperature.gpu": "55", "utilization.gpu": "0", "uuid": "", "vbios_version": "94.02.42.C0.10" }, "compute_apps": [ ", 42912, /usr/local/lib/ollama/llama-server, 6574 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T10:59:09Z" }, "summary": { "accuracy": 0.35185185185185186, "brier": 1.2962962962962963, "correct": 38, "errors": 0, "eval_tokens_total": 851, "floored_rows": 0, "letters_floored_total": 0, "median_seconds": 0.6208794944977853, "n": 108, "n_labelled": 108, "n_reached": 108, "p95_seconds": 0.6748884929984342, "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": 0.35185185185185186, "correct": 38, "gap": 0.6481481481481481, "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": 1315.1809742450714, "mean_watts": 164.31546510582922, "median_s_between_changes": null, "min_watts": 99.0, "peak_watts": 194.0, "samples": 9, "seconds": 8.003999948501587, "value_changes": 1 }, "wall_seconds": 67.40681108400167, "wall_task": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 18808.47054541111, "mean_watts": 284.94660393852814, "median_s_between_changes": 14.001500010490417, "min_watts": 194.0, "peak_watts": 312.0, "samples": 67, "seconds": 66.007000207901, "value_changes": 5 } }, "d": { "board_idle_window_epoch": [ 1792062028.643, 1792062037.651 ], "board_task_window_epoch": [ 1792062038.711, 1792062069.719 ], "energy": { "cards": [ "" ], "decisions": 48, "joules": 5955.945590348244, "joules_per_decision": 124.08219979892175, "mean_watts": 192.07770782819736, "net_joules_per_decision": 41.73105435164345, "peak_watts": 213.08, "per_card": { "": { "joules": 5955.945590348244, "mean_watts": 192.07770782819736, "min_watts": 120.94, "peak_watts": 213.08, "samples": 32, "seconds": 31.008000135421753 } }, "samples": 32, "seconds": 31.008000135421753 }, "idle": { "cards": [ "" ], "joules": 1148.326820590496, "mean_watts": 127.47855276722102, "peak_watts": 161.65, "per_card": { "": { "joules": 1148.326820590496, "mean_watts": 127.47855276722102, "min_watts": 121.13, "peak_watts": 161.65, "samples": 10, "seconds": 9.008000135421753 } }, "samples": 10, "seconds": 9.008000135421753, "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": "6594", "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": "166.46", "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": [ ", 42912, /usr/local/lib/ollama/llama-server, 6584 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T11:01:11Z" }, "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": "6594", "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": "161.65", "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": [ ", 42912, /usr/local/lib/ollama/llama-server, 6584 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T11:00:28Z" }, "summary": { "accuracy": 0.7083333333333334, "brier": 0.4270833333333333, "correct": 34, "errors": 0, "eval_tokens_total": 126, "floored_rows": 0, "letters_floored_total": 0, "median_seconds": 0.6515960639990226, "n": 48, "n_labelled": 48, "n_reached": 48, "p95_seconds": 0.7041152300007525, "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": 5, "sum_p": 2.5 }, { "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": 0.7906976744186046, "correct": 34, "gap": 0.2093023255813954, "hi": 1.0, "lo": 0.9, "mean_p": 1.0, "n": 43, "sum_p": 43.0 } ], "unparsed": 5 }, "wall_idle_resident": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 2178.110988378525, "mean_watts": 272.12781139338773, "median_s_between_changes": null, "min_watts": 199.0, "peak_watts": 289.0, "samples": 9, "seconds": 8.003999948501587, "value_changes": 1 }, "wall_seconds": 31.413001925997378, "wall_task": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 7906.313014268875, "mean_watts": 263.51741521648034, "median_s_between_changes": 14.002000093460083, "min_watts": 199.0, "peak_watts": 307.0, "samples": 31, "seconds": 30.003000020980835, "value_changes": 2 } } }, "think": "off", "think_budget": 8, "ups_before": { "available": true, "error": null, "load_pct": 20.0, "read_utc": "2026-10-15T10:58:41Z", "ups.realpower": 203.0, "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": 3823.693536, "seconds": 3.8570525879986235 } }