{ "argv": [ "/workshop/bench-quant-ladder/coi_quant.py", "--arm-id", "q-q5_K_M", "--model", "mistral-nemo:12b-instruct-2407-q5_K_M", "--num-ctx", "4096", "--think", "off", "--tasks", "c,d", "--probes", "--out", "/workshop/bench-quant-ladder/results/tasks/q-q5_K_M" ], "arm_id": "q-q5_K_M", "baseline_rows": [ { "memory_mib": 1.0, "sm_mhz": 210.0, "temp_c": 29.0, "uuid": "", "watts": 23.19 }, { "memory_mib": 1.0, "sm_mhz": 210.0, "temp_c": 54.0, "uuid": "", "watts": 27.68 } ], "board": "", "box": "benchbox", "box_class": "gpu-ql3090-24g", "box_class_check": "gpu-ql3090-24g", "disk_floor": { "checked_utc": "2026-10-15T10:46:19Z", "filesystems": { "ollama model store": { "above_floor": true, "df_raw": "Avail Mounted on\n32462303232 /", "df_rc": 0, "free_gib": 30.23, "path": "/workshop/bench-quant-ladder/store", "readable": true }, "results": { "above_floor": true, "df_raw": "Avail Mounted on\n32462303232 /", "df_rc": 0, "free_gib": 30.23, "path": "/workshop/bench-quant-ladder/results/tasks/q-q5_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 30.23 GiB on /workshop/bench-quant-ladder/results/tasks/q-q5_K_M (results); free 30.23 GiB on /workshop/bench-quant-ladder/store (ollama model store) -- all at or above the 20.0 GiB floor" }, "ended": "2026-10-15T10:49:32Z", "host": "http://127.0.0.1:withheld-2026-10-16", "idle_preload": { "board": { "cards": [ "" ], "joules": 390.77918148756027, "mean_watts": 27.874968257284852, "peak_watts": 28.03, "per_card": { "": { "joules": 390.77918148756027, "mean_watts": 27.874968257284852, "min_watts": 27.75, "peak_watts": 28.03, "samples": 15, "seconds": 14.019000053405762 } }, "samples": 15, "seconds": 14.019000053405762 }, "wall": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 1313.4029948711395, "mean_watts": 100.99992310615774, "median_s_between_changes": 10.003000020980835, "min_watts": 100.0, "peak_watts": 102.0, "samples": 14, "seconds": 13.003999948501587, "value_changes": 2 }, "window_epoch": [ 1792061185.806, 1792061199.825 ] }, "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-q5_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": "Q5_K_M" }, "digest": "cf4d8cc50281aec50958585a1e89425a7de658700bb354890f2a2bdad78ff18e", "expires_at": "2319-01-25T10:34:02.120363012Z", "model": "mistral-nemo:12b-instruct-2407-q5_K_M", "name": "mistral-nemo:12b-instruct-2407-q5_K_M", "size": 9051108801, "size_vram": 9051108801 }, "card": "", "cards": [ "" ], "fit": { "growth_bytes": 9373220864.0, "ps_size": 9051108801, "ps_size_vram": 9051108801, "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": 8727644109, "whole": true }, "floor_mib": 200.0, "growth_mib": { "": 0.0, "": 8939.0 }, "memory_after_mib": { "": 1.0, "": 8940.0 }, "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": 23.07 }, { "memory_mib": 1.0, "sm_mhz": 210.0, "temp_c": 60.0, "uuid": "", "watts": 51.25 } ], "stamp": "2026-10-15T10:49:32Z" }, "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": "1440", "clocks_throttle_reasons.active": "0x0000000000000004", "driver_version": "595.84", "enforced.power.limit": "300.00", "memory.total": "24576", "memory.used": "8950", "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": "199.79", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P2", "temperature.gpu": "66", "utilization.gpu": "94", "uuid": "", "vbios_version": "94.02.42.C0.10" }, "compute_apps": [ ", 36280, /usr/local/lib/ollama/llama-server, 8940 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T10:49:25Z" } }, "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": "810", "clocks.sm": "225", "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": "34.74", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P8", "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": "2026-10-15T10:46:19Z" }, "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": 3219.3730484330654, "mean_watts": 268.12468198680614, "peak_watts": 301.84, "per_card": { "": { "joules": 3219.3730484330654, "mean_watts": 268.12468198680614, "min_watts": 147.58, "peak_watts": 301.84, "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": 78.7046203301413, "done_reason": "length", "eval_count": 256, "eval_s": 3.252668, "load_s": 0.433409695, "prefill_tok_s": 1148.6551163013305, "prompt_eval_count": 24, "prompt_eval_s": 0.020894, "text_head": "In the heart of a winter storm, the lighthouse stands sentinel, a solitary beacon against the relentless fury of the sea", "thinking_chars": 0, "total_s": 3.753002741 }, { "decode_tok_s": 78.59127593857478, "done_reason": "length", "eval_count": 256, "eval_s": 3.257359, "load_s": 0.420054022, "prefill_tok_s": 1827.875095201828, "prompt_eval_count": 24, "prompt_eval_s": 0.01313, "text_head": "In the heart of a winter storm, the lighthouse stands sentinel, a solitary beacon against the relentless fury of the sea", "thinking_chars": 0, "total_s": 3.718981888 }, { "decode_tok_s": 78.51233839799254, "done_reason": "length", "eval_count": 256, "eval_s": 3.260634, "load_s": 0.439816759, "prefill_tok_s": 1879.1105543376134, "prompt_eval_count": 24, "prompt_eval_s": 0.012772, "text_head": "In the heart of a winter storm, the lighthouse stands sentinel, a solitary beacon against the relentless fury of the sea", "thinking_chars": 0, "total_s": 3.745664615 } ], "wall": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 4020.833975791931, "mean_watts": 365.4971368314598, "median_s_between_changes": null, "min_watts": 334.0, "peak_watts": 411.0, "samples": 12, "seconds": 11.000999927520752, "value_changes": 1 } }, "prefill": { "board": { "cards": [ "" ], "joules": 2911.8432003688813, "mean_watts": 223.7985718794822, "peak_watts": 253.5, "per_card": { "": { "joules": 2911.8432003688813, "mean_watts": 223.7985718794822, "min_watts": 152.04, "peak_watts": 253.5, "samples": 14, "seconds": 13.010999917984009 } }, "samples": 14, "seconds": 13.010999917984009 }, "runs": [ { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.434541762, "prefill_tok_s": 3126.0505608338367, "prompt_eval_count": 2241, "prompt_eval_s": 0.716879, "text_head": "Under", "thinking_chars": 0, "total_s": 1.182249181 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.430909937, "prefill_tok_s": 3126.2697416464093, "prompt_eval_count": 2239, "prompt_eval_s": 0.716189, "text_head": "Under", "thinking_chars": 0, "total_s": 1.396277345 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.429861455, "prefill_tok_s": 3124.5902462567688, "prompt_eval_count": 2240, "prompt_eval_s": 0.716894, "text_head": "Under", "thinking_chars": 0, "total_s": 1.389282896 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.41440042, "prefill_tok_s": 3130.3934653036413, "prompt_eval_count": 2240, "prompt_eval_s": 0.715565, "text_head": "Under", "thinking_chars": 0, "total_s": 1.369170801 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.439594811, "prefill_tok_s": 3127.641825025041, "prompt_eval_count": 2242, "prompt_eval_s": 0.716834, "text_head": "Under", "thinking_chars": 0, "total_s": 1.3956647 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.432174165, "prefill_tok_s": 3124.6857155628068, "prompt_eval_count": 2237, "prompt_eval_s": 0.715912, "text_head": "Under", "thinking_chars": 0, "total_s": 1.387664129 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.452772916, "prefill_tok_s": 3121.4003262079236, "prompt_eval_count": 2241, "prompt_eval_s": 0.717947, "text_head": "Under", "thinking_chars": 0, "total_s": 1.40941615 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.434805092, "prefill_tok_s": 3124.4590134639334, "prompt_eval_count": 2238, "prompt_eval_s": 0.716284, "text_head": "Under", "thinking_chars": 0, "total_s": 1.386341161 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.400795159, "prefill_tok_s": 3126.7004311246915, "prompt_eval_count": 2241, "prompt_eval_s": 0.71673, "text_head": "Under", "thinking_chars": 0, "total_s": 1.354061563 } ], "target_tokens_approx": 3000.0, "wall": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 4964.329037904739, "mean_watts": 413.6251457463172, "median_s_between_changes": null, "min_watts": 411.0, "peak_watts": 418.0, "samples": 13, "seconds": 12.002000093460083, "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": 0.5649458310035698, "p95_s": 0.5863656680012355, "runs": [ { "eval_count": 8, "prompt_eval_count": 216, "server_load_ms": 439.159165, "server_prompt_eval_ms": 64.843999, "total_s": 0.8196262810015469, "ttft_s": 0.742970821003837 }, { "eval_count": 8, "prompt_eval_count": 315, "server_load_ms": 433.052908, "server_prompt_eval_ms": 97.362, "total_s": 0.6475314950002939, "ttft_s": 0.5564868200017372 }, { "eval_count": 8, "prompt_eval_count": 234, "server_load_ms": 431.896708, "server_prompt_eval_ms": 63.038, "total_s": 0.6213435030003893, "ttft_s": 0.5447574420031742 }, { "eval_count": 8, "prompt_eval_count": 334, "server_load_ms": 412.10433, "server_prompt_eval_ms": 97.244, "total_s": 0.6287096379965078, "ttft_s": 0.5374932049962808 }, { "eval_count": 6, "prompt_eval_count": 343, "server_load_ms": 394.99606, "server_prompt_eval_ms": 108.321, "total_s": 0.6130338560033124, "ttft_s": 0.5609728210038156 }, { "eval_count": 8, "prompt_eval_count": 345, "server_load_ms": 431.033815, "server_prompt_eval_ms": 109.935, "total_s": 0.6727459140020073, "ttft_s": 0.581873078001081 }, { "eval_count": 8, "prompt_eval_count": 254, "server_load_ms": 447.753499, "server_prompt_eval_ms": 70.283, "total_s": 0.6512100319960155, "ttft_s": 0.5720853629973135 }, { "eval_count": 6, "prompt_eval_count": 232, "server_load_ms": 402.386275, "server_prompt_eval_ms": 63.099, "total_s": 0.5612681870043161, "ttft_s": 0.5099507460035966 }, { "eval_count": 8, "prompt_eval_count": 305, "server_load_ms": 424.379583, "server_prompt_eval_ms": 78.979, "total_s": 0.6251159499952337, "ttft_s": 0.5345967209941591 }, { "eval_count": 3, "prompt_eval_count": 340, "server_load_ms": 435.995332, "server_prompt_eval_ms": 97.598, "total_s": 0.5969998959990335, "ttft_s": 0.5689188410033239 }, { "eval_count": 6, "prompt_eval_count": 336, "server_load_ms": 436.043731, "server_prompt_eval_ms": 97.518, "total_s": 0.6382438549990184, "ttft_s": 0.5863656680012355 }, { "eval_count": 8, "prompt_eval_count": 333, "server_load_ms": 438.400901, "server_prompt_eval_ms": 97.744, "total_s": 0.6683433130019694, "ttft_s": 0.5764580450049834 } ] } }, "show": { "capabilities": [ "completion", "tools" ], "details": { "families": [ "llama" ], "family": "llama", "format": "gguf", "parameter_size": "12.2B", "parent_model": "", "quantization_level": "Q5_K_M" }, "model_info": { "general.architecture": "llama", "general.basename": "Mistral-Nemo", "general.file_type": 17, "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": "2026-10-15T10:46:19Z", "tag": { "details": { "context_length": 1024000, "embedding_length": 5120, "families": [ "llama" ], "family": "llama", "format": "gguf", "parameter_size": "12.2B", "parent_model": "", "quantization_level": "Q5_K_M" }, "digest": "cf4d8cc50281aec50958585a1e89425a7de658700bb354890f2a2bdad78ff18e", "modified_at": "2026-10-15T10:38:00.498318762Z", "name": "mistral-nemo:12b-instruct-2407-q5_K_M", "size": 8727644109 }, "tasks": [ "c", "d" ], "tasks_run": { "c": { "board_idle_window_epoch": [ 1792061208.522, 1792061217.532 ], "board_task_window_epoch": [ 1792061218.591, 1792061285.607 ], "energy": { "cards": [ "" ], "decisions": 108, "joules": 12757.96750985384, "joules_per_decision": 118.12932879494295, "mean_watts": 190.37196346665039, "net_joules_per_decision": 46.713962328297185, "peak_watts": 207.04, "per_card": { "": { "joules": 12757.96750985384, "mean_watts": 190.37196346665039, "min_watts": 114.79, "peak_watts": 207.04, "samples": 68, "seconds": 67.01600003242493 } }, "samples": 68, "seconds": 67.01600003242493 }, "idle": { "cards": [ "" ], "joules": 1036.959303661585, "mean_watts": 115.08982294774329, "peak_watts": 117.22, "per_card": { "": { "joules": 1036.959303661585, "mean_watts": 115.08982294774329, "min_watts": 114.39, "peak_watts": 117.22, "samples": 10, "seconds": 9.009999990463257 } }, "samples": 10, "seconds": 9.009999990463257, "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": "8950", "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": "160.69", "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": [ ", 36280, /usr/local/lib/ollama/llama-server, 8940 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T10:48:07Z" }, "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": "8940", "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": "117.22", "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": [ ", 36280, /usr/local/lib/ollama/llama-server, 8930 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T10:46:48Z" }, "summary": { "accuracy": 0.5277777777777778, "brier": 0.9444444444444444, "correct": 57, "errors": 0, "eval_tokens_total": 845, "floored_rows": 0, "letters_floored_total": 0, "median_seconds": 0.6297555099990859, "n": 108, "n_labelled": 108, "n_reached": 108, "p95_seconds": 0.6614327120041708, "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": 0.5277777777777778, "correct": 57, "gap": 0.4722222222222222, "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": 1703.0899815559387, "mean_watts": 212.75328019145445, "median_s_between_changes": null, "min_watts": 194.0, "peak_watts": 254.0, "samples": 9, "seconds": 8.004999876022339, "value_changes": 1 }, "wall_seconds": 67.54636706299789, "wall_task": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 18552.758006811142, "mean_watts": 281.0768414237598, "median_s_between_changes": 14.001500010490417, "min_watts": 194.0, "peak_watts": 304.0, "samples": 67, "seconds": 66.00600004196167, "value_changes": 5 } }, "d": { "board_idle_window_epoch": [ 1792061287.827, 1792061296.836 ], "board_task_window_epoch": [ 1792061297.892, 1792061328.901 ], "energy": { "cards": [ "" ], "decisions": 48, "joules": 5951.355899699926, "joules_per_decision": 123.98658124374846, "mean_watts": 191.9235024547641, "net_joules_per_decision": 41.663636602269, "peak_watts": 213.28, "per_card": { "": { "joules": 5951.355899699926, "mean_watts": 191.9235024547641, "min_watts": 120.68, "peak_watts": 213.28, "samples": 32, "seconds": 31.009000062942505 } }, "samples": 32, "seconds": 31.009000062942505 }, "idle": { "cards": [ "" ], "joules": 1148.023985735178, "mean_watts": 127.43078895708346, "peak_watts": 160.69, "per_card": { "": { "joules": 1148.023985735178, "mean_watts": 127.43078895708346, "min_watts": 120.8, "peak_watts": 160.69, "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": "8950", "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": "165.76", "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": [ ", 36280, /usr/local/lib/ollama/llama-server, 8940 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T10:48:50Z" }, "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": "8950", "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": "160.69", "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": [ ", 36280, /usr/local/lib/ollama/llama-server, 8940 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T10:48:07Z" }, "summary": { "accuracy": 0.6875, "brier": 0.625, "correct": 33, "errors": 0, "eval_tokens_total": 144, "floored_rows": 0, "letters_floored_total": 0, "median_seconds": 0.650465589002124, "n": 48, "n_labelled": 48, "n_reached": 48, "p95_seconds": 0.7383106409979519, "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": 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": 0.6875, "correct": 33, "gap": 0.3125, "hi": 1.0, "lo": 0.9, "mean_p": 1.0, "n": 48, "sum_p": 48.0 } ], "unparsed": 0 }, "wall_idle_resident": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 2170.603985786438, "mean_watts": 271.1899050165277, "median_s_between_changes": null, "min_watts": 199.0, "peak_watts": 276.0, "samples": 9, "seconds": 8.003999948501587, "value_changes": 1 }, "wall_seconds": 31.260591092999675, "wall_task": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 7838.299489498138, "mean_watts": 261.250524414788, "median_s_between_changes": 14.001999855041504, "min_watts": 199.0, "peak_watts": 334.0, "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": "2026-10-15T10:46:19Z", "ups.realpower": 267.0, "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": 4332.064163, "seconds": 4.365441251997254 } }