{ "argv": [ "/workshop/bench-quant-ladder/coi_quant.py", "--arm-id", "q-q8_0", "--model", "mistral-nemo:12b-instruct-2407-q8_0", "--num-ctx", "4096", "--think", "off", "--tasks", "c,d", "--probes", "--out", "/workshop/bench-quant-ladder/results/tasks/q-q8_0" ], "arm_id": "q-q8_0", "baseline_rows": [ { "memory_mib": 1.0, "sm_mhz": 210.0, "temp_c": 28.0, "uuid": "", "watts": 23.17 }, { "memory_mib": 1.0, "sm_mhz": 210.0, "temp_c": 52.0, "uuid": "", "watts": 27.75 } ], "board": "", "box": "benchbox", "box_class": "gpu-ql3090-24g", "box_class_check": "gpu-ql3090-24g", "disk_floor": { "checked_utc": "2026-10-15T05:55:59Z", "filesystems": { "ollama model store": { "above_floor": true, "df_raw": "Avail Mounted on\n28182585344 /", "df_rc": 0, "free_gib": 26.25, "path": "/workshop/bench-quant-ladder/store", "readable": true }, "results": { "above_floor": true, "df_raw": "Avail Mounted on\n28182585344 /", "df_rc": 0, "free_gib": 26.25, "path": "/workshop/bench-quant-ladder/results/tasks/q-q8_0", "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 26.25 GiB on /workshop/bench-quant-ladder/results/tasks/q-q8_0 (results); free 26.25 GiB on /workshop/bench-quant-ladder/store (ollama model store) -- all at or above the 20.0 GiB floor" }, "ended": "2026-10-15T05:59:14Z", "host": "http://127.0.0.1:withheld-2026-10-16", "idle_preload": { "board": { "cards": [ "" ], "joules": 391.70502854824065, "mean_watts": 27.943003640364914, "peak_watts": 28.23, "per_card": { "": { "joules": 391.70502854824065, "mean_watts": 27.943003640364914, "min_watts": 27.69, "peak_watts": 28.23, "samples": 15, "seconds": 14.01800012588501 } }, "samples": 15, "seconds": 14.01800012588501 }, "wall": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 1324.0095117092133, "mean_watts": 101.8077278014797, "median_s_between_changes": null, "min_watts": 101.0, "peak_watts": 102.0, "samples": 14, "seconds": 13.005000114440918, "value_changes": 1 }, "window_epoch": [ 1792043765.972, 1792043779.99 ] }, "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-q8_0", "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": "Q8_0" }, "digest": "e9b8ba00d2a1194bf87d936f64cdfb20b7d25d9776c8dae6baebde266728528c", "expires_at": "2319-01-25T05:43:43.027434992Z", "model": "mistral-nemo:12b-instruct-2407-q8_0", "name": "mistral-nemo:12b-instruct-2407-q8_0", "size": 13094187171, "size_vram": 13094187171 }, "card": "", "cards": [ "" ], "fit": { "growth_bytes": 13416529920.0, "ps_size": 13094187171, "ps_size_vram": 13094187171, "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": 13022382027, "whole": true }, "floor_mib": 200.0, "growth_mib": { "": 0.0, "": 12795.0 }, "memory_after_mib": { "": 1.0, "": 12796.0 }, "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": 23.14 }, { "memory_mib": 1.0, "sm_mhz": 210.0, "temp_c": 60.0, "uuid": "", "watts": 54.82 } ], "stamp": "2026-10-15T05:59:14Z" }, "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": "1680", "clocks_throttle_reasons.active": "0x0000000000000004", "driver_version": "595.84", "enforced.power.limit": "300.00", "memory.total": "24576", "memory.used": "12804", "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": "188.05", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P2", "temperature.gpu": "65", "utilization.gpu": "70", "uuid": "", "vbios_version": "94.02.42.C0.10" }, "compute_apps": [ ", 18280, /usr/local/lib/ollama/llama-server, 12794 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T05:59:08Z" } }, "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": "255", "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": "50.68", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P5", "temperature.gpu": "56", "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": "2026-10-15T05:55:59Z" }, "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": 4113.784297969341, "mean_watts": 274.0513127111456, "peak_watts": 300.07, "per_card": { "": { "joules": 4113.784297969341, "mean_watts": 274.0513127111456, "min_watts": 142.93, "peak_watts": 300.07, "samples": 16, "seconds": 15.011000156402588 } }, "samples": 16, "seconds": 15.011000156402588 }, "num_predict": 256, "prompt": "Describe, in about three hundred words, a lighthouse on a rocky coast during a winter storm.", "runs": [ { "decode_tok_s": 60.151238069906555, "done_reason": "length", "eval_count": 256, "eval_s": 4.255939, "load_s": 0.403607076, "prefill_tok_s": 911.0233829334953, "prompt_eval_count": 24, "prompt_eval_s": 0.026344, "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": 4.732515505 }, { "decode_tok_s": 60.162250068152545, "done_reason": "length", "eval_count": 256, "eval_s": 4.25516, "load_s": 0.387038602, "prefill_tok_s": 1398.8459520895262, "prompt_eval_count": 24, "prompt_eval_s": 0.017157, "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": 4.687314561 }, { "decode_tok_s": 60.14684287812042, "done_reason": "length", "eval_count": 256, "eval_s": 4.25625, "load_s": 0.338398526, "prefill_tok_s": 1404.4943820224719, "prompt_eval_count": 24, "prompt_eval_s": 0.017088, "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": 4.644184847 } ], "wall": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 5091.115018606186, "mean_watts": 363.5731636776489, "median_s_between_changes": null, "min_watts": 281.0, "peak_watts": 417.0, "samples": 15, "seconds": 14.003000020980835, "value_changes": 1 } }, "prefill": { "board": { "cards": [ "" ], "joules": 2911.268877413273, "mean_watts": 223.77162794368374, "peak_watts": 248.08, "per_card": { "": { "joules": 2911.268877413273, "mean_watts": 223.77162794368374, "min_watts": 145.76, "peak_watts": 248.08, "samples": 14, "seconds": 13.009999990463257 } }, "samples": 14, "seconds": 13.009999990463257 }, "runs": [ { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.399041776, "prefill_tok_s": 3387.3505658432323, "prompt_eval_count": 2241, "prompt_eval_s": 0.661579, "text_head": "Under", "thinking_chars": 0, "total_s": 1.092889009 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.414134945, "prefill_tok_s": 3374.4832398656235, "prompt_eval_count": 2239, "prompt_eval_s": 0.663509, "text_head": "Under", "thinking_chars": 0, "total_s": 1.317197181 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.398203374, "prefill_tok_s": 3382.689389077236, "prompt_eval_count": 2240, "prompt_eval_s": 0.662195, "text_head": "Under", "thinking_chars": 0, "total_s": 1.303001015 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.363882704, "prefill_tok_s": 3383.3433776339857, "prompt_eval_count": 2240, "prompt_eval_s": 0.662067, "text_head": "Under", "thinking_chars": 0, "total_s": 1.262078692 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.389447751, "prefill_tok_s": 3374.1532636128723, "prompt_eval_count": 2242, "prompt_eval_s": 0.664463, "text_head": "Under", "thinking_chars": 0, "total_s": 1.292997153 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.426029941, "prefill_tok_s": 3380.5328457225755, "prompt_eval_count": 2237, "prompt_eval_s": 0.66173, "text_head": "Under", "thinking_chars": 0, "total_s": 1.326797378 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.37572857, "prefill_tok_s": 3382.2177749806438, "prompt_eval_count": 2241, "prompt_eval_s": 0.662583, "text_head": "Under", "thinking_chars": 0, "total_s": 1.275153588 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.394787704, "prefill_tok_s": 3371.31460179924, "prompt_eval_count": 2238, "prompt_eval_s": 0.663836, "text_head": "Under", "thinking_chars": 0, "total_s": 1.295435831 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.413597599, "prefill_tok_s": 3371.9276863700993, "prompt_eval_count": 2241, "prompt_eval_s": 0.664605, "text_head": "Under", "thinking_chars": 0, "total_s": 1.317851565 }, { "decode_tok_s": 1000000.0, "done_reason": "length", "eval_count": 1, "eval_s": 1e-06, "load_s": 0.387660725, "prefill_tok_s": 3378.9491468153406, "prompt_eval_count": 2240, "prompt_eval_s": 0.662928, "text_head": "Under", "thinking_chars": 0, "total_s": 1.288443062 } ], "target_tokens_approx": 3000.0, "wall": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 4995.916969776154, "mean_watts": 416.2917256852483, "median_s_between_changes": null, "min_watts": 416.0, "peak_watts": 417.0, "samples": 13, "seconds": 12.000999927520752, "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": 0.5442470734997187, "p95_s": 0.5861426230003417, "runs": [ { "eval_count": 8, "prompt_eval_count": 216, "server_load_ms": 425.797832, "server_prompt_eval_ms": 58.732, "total_s": 0.8287689739991038, "ttft_s": 0.7264223569982278 }, { "eval_count": 8, "prompt_eval_count": 315, "server_load_ms": 404.458109, "server_prompt_eval_ms": 93.391, "total_s": 0.6431951449994813, "ttft_s": 0.5409903170002508 }, { "eval_count": 8, "prompt_eval_count": 234, "server_load_ms": 416.090707, "server_prompt_eval_ms": 58.402, "total_s": 0.6307494430002407, "ttft_s": 0.52906717100268 }, { "eval_count": 3, "prompt_eval_count": 334, "server_load_ms": 380.983935, "server_prompt_eval_ms": 94.046, "total_s": 0.5429530309993424, "ttft_s": 0.5062495210004272 }, { "eval_count": 6, "prompt_eval_count": 343, "server_load_ms": 398.890078, "server_prompt_eval_ms": 101.105, "total_s": 0.6261049139975512, "ttft_s": 0.5573462639986246 }, { "eval_count": 3, "prompt_eval_count": 345, "server_load_ms": 410.785514, "server_prompt_eval_ms": 101.407, "total_s": 0.5890961570003128, "ttft_s": 0.5526726549978775 }, { "eval_count": 8, "prompt_eval_count": 254, "server_load_ms": 413.991946, "server_prompt_eval_ms": 66.01, "total_s": 0.644868219998898, "ttft_s": 0.5404062919988064 }, { "eval_count": 6, "prompt_eval_count": 232, "server_load_ms": 420.925995, "server_prompt_eval_ms": 59.223, "total_s": 0.5959624679999251, "ttft_s": 0.5274996250009281 }, { "eval_count": 3, "prompt_eval_count": 305, "server_load_ms": 436.638258, "server_prompt_eval_ms": 72.666, "total_s": 0.5765858259983361, "ttft_s": 0.5401527149988397 }, { "eval_count": 3, "prompt_eval_count": 340, "server_load_ms": 422.485071, "server_prompt_eval_ms": 94.833, "total_s": 0.5897798020014307, "ttft_s": 0.5532264710018353 }, { "eval_count": 6, "prompt_eval_count": 336, "server_load_ms": 432.718312, "server_prompt_eval_ms": 92.803, "total_s": 0.6548660330008715, "ttft_s": 0.5861426230003417 }, { "eval_count": 3, "prompt_eval_count": 333, "server_load_ms": 411.986733, "server_prompt_eval_ms": 94.634, "total_s": 0.5841727900005935, "ttft_s": 0.5475038299991866 } ] } }, "show": { "capabilities": [ "completion", "tools" ], "details": { "families": [ "llama" ], "family": "llama", "format": "gguf", "parameter_size": "12.2B", "parent_model": "", "quantization_level": "Q8_0" }, "model_info": { "general.architecture": "llama", "general.basename": "Mistral-Nemo", "general.file_type": 7, "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": "2026-10-15T05:55:59Z", "tag": { "details": { "context_length": 1024000, "embedding_length": 5120, "families": [ "llama" ], "family": "llama", "format": "gguf", "parameter_size": "12.2B", "parent_model": "", "quantization_level": "Q8_0" }, "digest": "e9b8ba00d2a1194bf87d936f64cdfb20b7d25d9776c8dae6baebde266728528c", "modified_at": "2026-10-15T05:47:53.425223113Z", "name": "mistral-nemo:12b-instruct-2407-q8_0", "size": 13022382027 }, "tasks": [ "c", "d" ], "tasks_run": { "c": { "board_idle_window_epoch": [ 1792043789.421, 1792043798.43 ], "board_task_window_epoch": [ 1792043799.487, 1792043866.502 ], "energy": { "cards": [ "" ], "decisions": 108, "joules": 12920.938952863216, "joules_per_decision": 119.63832363762238, "mean_watts": 192.80666914328123, "net_joules_per_decision": 49.04074902420315, "peak_watts": 209.78, "per_card": { "": { "joules": 12920.938952863216, "mean_watts": 192.80666914328123, "min_watts": 114.04, "peak_watts": 209.78, "samples": 68, "seconds": 67.01500010490417 } }, "samples": 68, "seconds": 67.01500010490417 }, "idle": { "cards": [ "" ], "joules": 1024.9864021360875, "mean_watts": 113.77360361581661, "peak_watts": 114.57, "per_card": { "": { "joules": 1024.9864021360875, "mean_watts": 113.77360361581661, "min_watts": 113.49, "peak_watts": 114.57, "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": "12804", "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": "159.65", "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": [ ", 18280, /usr/local/lib/ollama/llama-server, 12794 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T05:57:48Z" }, "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": "12796", "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": "114.57", "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": [ ", 18280, /usr/local/lib/ollama/llama-server, 12786 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T05:56:29Z" }, "summary": { "accuracy": 0.5, "brier": 1.0, "correct": 54, "errors": 0, "eval_tokens_total": 809, "floored_rows": 0, "letters_floored_total": 0, "median_seconds": 0.6320349275010813, "n": 108, "n_labelled": 108, "n_reached": 108, "p95_seconds": 0.678643925999495, "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": 0.5, "correct": 54, "gap": 0.5, "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": 1478.3159894943237, "mean_watts": 184.6971513000917, "median_s_between_changes": null, "min_watts": 101.0, "peak_watts": 204.0, "samples": 9, "seconds": 8.003999948501587, "value_changes": 1 }, "wall_seconds": 67.86779876899891, "wall_task": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 19147.277016878128, "mean_watts": 290.08388638465783, "median_s_between_changes": 14.001500010490417, "min_watts": 204.0, "peak_watts": 313.0, "samples": 67, "seconds": 66.00600004196167, "value_changes": 5 } }, "d": { "board_idle_window_epoch": [ 1792043868.725, 1792043877.733 ], "board_task_window_epoch": [ 1792043878.789, 1792043908.799 ], "energy": { "cards": [ "" ], "decisions": 48, "joules": 5613.13340626359, "joules_per_decision": 116.94027929715811, "mean_watts": 187.04209956838926, "net_joules_per_decision": 34.42504861974106, "peak_watts": 197.68, "per_card": { "": { "joules": 5613.13340626359, "mean_watts": 187.04209956838926, "min_watts": 121.04, "peak_watts": 197.68, "samples": 31, "seconds": 30.009999990463257 } }, "samples": 31, "seconds": 30.009999990463257 }, "idle": { "cards": [ "" ], "joules": 1188.8792418837547, "mean_watts": 131.98037566726697, "peak_watts": 159.65, "per_card": { "": { "joules": 1188.8792418837547, "mean_watts": 131.98037566726697, "min_watts": 121.36, "peak_watts": 159.65, "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": "12804", "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": "142.93", "power.limit": "300.00", "power.max_limit": "366.00", "pstate": "P2", "temperature.gpu": "61", "utilization.gpu": "0", "uuid": "", "vbios_version": "94.02.42.C0.10" }, "compute_apps": [ ", 18280, /usr/local/lib/ollama/llama-server, 12794 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T05:58:30Z" }, "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": "1950", "clocks_throttle_reasons.active": "0x0000000000000000", "driver_version": "595.84", "enforced.power.limit": "300.00", "memory.total": "24576", "memory.used": "12804", "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": "159.65", "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": [ ", 18280, /usr/local/lib/ollama/llama-server, 12794 MiB" ], "how": "nvidia-smi -i ; enforced.power.limit and persistence_mode READ, never typed", "stamp": "2026-10-15T05:57:48Z" }, "summary": { "accuracy": 0.7083333333333334, "brier": 0.5520833333333334, "correct": 34, "errors": 0, "eval_tokens_total": 102, "floored_rows": 0, "letters_floored_total": 0, "median_seconds": 0.6226643595000496, "n": 48, "n_labelled": 48, "n_reached": 48, "p95_seconds": 0.6558643740027037, "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": 2222.6389651298523, "mean_watts": 277.69102691535477, "median_s_between_changes": null, "min_watts": 200.0, "peak_watts": 313.0, "samples": 9, "seconds": 8.003999948501587, "value_changes": 1 }, "wall_seconds": 30.01419336300023, "wall_task": { "instrument": "wall (UPS ups.realpower, whole box; integer watts; NUT pollinterval 2 s)", "joules": 7660.302013635635, "mean_watts": 264.1210222423251, "median_s_between_changes": 14.002000093460083, "min_watts": 200.0, "peak_watts": 301.0, "samples": 30, "seconds": 29.003000020980835, "value_changes": 2 } } }, "think": "off", "think_budget": 8, "ups_before": { "available": true, "error": null, "load_pct": 26.0, "read_utc": "2026-10-15T05:55:59Z", "ups.realpower": 261.0, "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": 5066.839858, "seconds": 5.101035069001227 } }