1{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:44:36", "prompt_tokens": 25, "gen_tokens": 64, "ttft_ms": 190.09033299516886, "prompt_tps": 248.33388555616366, "gen_tps": 121.36794961117941, "decode_tps_wall": 119.49292325368565, "itl_ms_p50": 8.360582985915244, "itl_ms_p95": 8.444542065262794, "itl_ms_p99": 8.543666917830706, "itl_ms_mean": 8.3686964279632, "total_s": 0.7173182079568505, "peak_mem_gb": 17.262479236, "text": "import csv\nfrom datetime import datetime\nimport statistics\n\ndef median_inter_arrival_time(csv_file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time "}2{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:44:36", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 47.851667040959, "prompt_tps": 338.8367836252268, "gen_tps": 121.09034285180219, "decode_tps_wall": 119.20153883118209, "itl_ms_p50": 8.380957995541394, "itl_ms_p95": 8.534958004020154, "itl_ms_p99": 8.731834008358419, "itl_ms_mean": 8.389153443868198, "total_s": 0.5763683340046555, "peak_mem_gb": 17.226143856, "text": "1. **Create a Weekly Schedule:** Use a digital calendar or planner to block out time for classes, study sessions, assignments, and personal activities. Having a visual representation of your week help"}3{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:44:37", "prompt_tokens": 19, "gen_tokens": 64, "ttft_ms": 51.94775003474206, "prompt_tps": 370.8173313765438, "gen_tps": 120.87331925022238, "decode_tps_wall": 118.99180893614097, "itl_ms_p50": 8.380333078093827, "itl_ms_p95": 8.60504200682044, "itl_ms_p99": 8.897583931684494, "itl_ms_mean": 8.40393980846755, "total_s": 0.5813959579681978, "peak_mem_gb": 17.249301776, "text": "Include details about temperature, precipitation, and weather patterns.\n\nGatineau, Qu\u00e9bec, experiences a humid continental climate with distinct seasonal variations. In winter (December\u2013February), tem"}4{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:44:37", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 49.24083407968283, "prompt_tps": 349.63177556877565, "gen_tps": 121.55569446717989, "decode_tps_wall": 119.65929399141673, "itl_ms_p50": 8.350541931577027, "itl_ms_p95": 8.482167031615973, "itl_ms_p99": 8.48712504375726, "itl_ms_mean": 8.35706084035337, "total_s": 0.5757356670219451, "peak_mem_gb": 17.226964956, "text": "Mixture-of-Experts (MoE) models are memory-bound during decoding because they rely on storing and accessing a large number of expert parameters, which significantly increases memory requirements. Duri"}5{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:44:38", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 53.90341696329415, "prompt_tps": 394.2674782039338, "gen_tps": 121.6063822332558, "decode_tps_wall": 119.70997873886853, "itl_ms_p50": 8.343624998815358, "itl_ms_p95": 8.49016709253192, "itl_ms_p99": 8.694582968018949, "itl_ms_mean": 8.353522492735276, "total_s": 0.5801753340056166, "peak_mem_gb": 17.25364684, "text": "class LRUCache:\n def __init__(self, capacity):\n self.capacity = capacity\n self.cache = {}\n self.order = []\n\n def get(self, key):\n if key in self.cache:\n # "}6{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:44:38", "prompt_tokens": 13, "gen_tokens": 64, "ttft_ms": 43.613624991849065, "prompt_tps": 302.6094225075362, "gen_tps": 121.57031801287954, "decode_tps_wall": 119.67484525660637, "itl_ms_p50": 8.331499993801117, "itl_ms_p95": 8.527584024704993, "itl_ms_p99": 8.608166011981666, "itl_ms_mean": 8.355974873882674, "total_s": 0.5700400420464575, "peak_mem_gb": 17.221640784, "text": " if n < 0:\n raise ValueError(\"n must be non-negative\")\n if n == 0:\n return 0\n if n == 1:\n return 1\n return fibonacci(n - 1) + fibonacci(n - 2)\n\n# Example usage:\nprint(f"}7{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:44:39", "prompt_tokens": 25, "gen_tokens": 64, "ttft_ms": 58.112749946303666, "prompt_tps": 434.565325733222, "gen_tps": 121.39687992141418, "decode_tps_wall": 119.50260375382749, "itl_ms_p50": 8.366625057533383, "itl_ms_p95": 8.409874979406595, "itl_ms_p99": 8.545209071598947, "itl_ms_mean": 8.368018508282683, "total_s": 0.5852979159681126, "peak_mem_gb": 17.26254994, "text": "import csv\nfrom datetime import datetime\nimport statistics\n\ndef median_inter_arrival_time(csv_file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time "}8{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:44:40", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 49.202000023797154, "prompt_tps": 349.57276418716367, "gen_tps": 121.89223393922533, "decode_tps_wall": 119.98965804328934, "itl_ms_p50": 8.334666024893522, "itl_ms_p95": 8.4102499531582, "itl_ms_p99": 8.483083103783429, "itl_ms_mean": 8.334051586672782, "total_s": 0.5742472499841824, "peak_mem_gb": 17.226956788, "text": "Mixture-of-Experts (MoE) models are memory-bound during decoding because they rely on storing and accessing a large number of expert parameters, which significantly increases memory requirements. Duri"}9{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:44:40", "prompt_tokens": 19, "gen_tokens": 64, "ttft_ms": 51.78712494671345, "prompt_tps": 370.81039172716237, "gen_tps": 121.22391645529108, "decode_tps_wall": 119.33175172953106, "itl_ms_p50": 8.359083090908825, "itl_ms_p95": 8.605082985013723, "itl_ms_p99": 8.688666974194348, "itl_ms_mean": 8.379999333844772, "total_s": 0.579727082978934, "peak_mem_gb": 17.249301772, "text": "Include details about temperature, precipitation, and weather patterns.\n\nGatineau, Qu\u00e9bec, experiences a humid continental climate with distinct seasonal variations. In winter (December\u2013February), tem"}10{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:44:41", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 53.91687504015863, "prompt_tps": 393.6823358259812, "gen_tps": 121.8112669918118, "decode_tps_wall": 119.91010556436554, "itl_ms_p50": 8.336875005625188, "itl_ms_p95": 8.415000047534704, "itl_ms_p99": 8.565249969251454, "itl_ms_mean": 8.339580682490672, "total_s": 0.5793104580370709, "peak_mem_gb": 17.25364682, "text": "class LRUCache:\n def __init__(self, capacity):\n self.capacity = capacity\n self.cache = {}\n self.order = []\n\n def get(self, key):\n if key in self.cache:\n # "}11{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:44:41", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 47.72816691547632, "prompt_tps": 339.2244104269285, "gen_tps": 121.67545188769039, "decode_tps_wall": 119.7779166732931, "itl_ms_p50": 8.335041929967701, "itl_ms_p95": 8.467959007248282, "itl_ms_p99": 8.659999934025109, "itl_ms_mean": 8.348784381745471, "total_s": 0.573701582965441, "peak_mem_gb": 17.22614392, "text": "1. **Create a Weekly Schedule:** Use a digital calendar or planner to block out time for classes, study sessions, assignments, and personal activities. Having a visual representation of your week help"}12{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:44:42", "prompt_tokens": 13, "gen_tokens": 64, "ttft_ms": 43.5285420389846, "prompt_tps": 302.6079571144651, "gen_tps": 121.40749208551493, "decode_tps_wall": 119.5127483416421, "itl_ms_p50": 8.359082974493504, "itl_ms_p95": 8.43141705263406, "itl_ms_p99": 8.60049994662404, "itl_ms_mean": 8.367308206664074, "total_s": 0.5706689590588212, "peak_mem_gb": 17.221640784, "text": " if n < 0:\n raise ValueError(\"n must be non-negative\")\n if n == 0:\n return 0\n if n == 1:\n return 1\n return fibonacci(n - 1) + fibonacci(n - 2)\n\n# Example usage:\nprint(f"}13{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:44:42", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 53.48683404736221, "prompt_tps": 396.8369599269693, "gen_tps": 121.25802358542144, "decode_tps_wall": 119.36565680352243, "itl_ms_p50": 8.361416985280812, "itl_ms_p95": 8.532540989108384, "itl_ms_p99": 8.71666707098484, "itl_ms_mean": 8.377619047043106, "total_s": 0.5812768340110779, "peak_mem_gb": 17.253646828, "text": "class LRUCache:\n def __init__(self, capacity):\n self.capacity = capacity\n self.cache = {}\n self.order = []\n\n def get(self, key):\n if key in self.cache:\n # "}14{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:44:43", "prompt_tokens": 13, "gen_tokens": 64, "ttft_ms": 43.41387492604554, "prompt_tps": 303.4981079859437, "gen_tps": 121.88605345378718, "decode_tps_wall": 119.9840021389397, "itl_ms_p50": 8.336500031873584, "itl_ms_p95": 8.409750065766275, "itl_ms_p99": 8.488874998874962, "itl_ms_mean": 8.334444444035254, "total_s": 0.5684838749002665, "peak_mem_gb": 17.221640784, "text": " if n < 0:\n raise ValueError(\"n must be non-negative\")\n if n == 0:\n return 0\n if n == 1:\n return 1\n return fibonacci(n - 1) + fibonacci(n - 2)\n\n# Example usage:\nprint(f"}15{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:44:44", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 49.159541027620435, "prompt_tps": 350.5781123763205, "gen_tps": 121.71373900696982, "decode_tps_wall": 119.81625321442641, "itl_ms_p50": 8.33791692275554, "itl_ms_p95": 8.414624957367778, "itl_ms_p99": 8.695708005689085, "itl_ms_mean": 8.346113095444347, "total_s": 0.5749646660406142, "peak_mem_gb": 17.226960852, "text": "Mixture-of-Experts (MoE) models are memory-bound during decoding because they rely on storing and accessing a large number of expert parameters, which significantly increases memory requirements. Duri"}16{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:44:44", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 47.806749935261905, "prompt_tps": 338.7148422536113, "gen_tps": 121.40541932367935, "decode_tps_wall": 119.51012231876976, "itl_ms_p50": 8.363333996385336, "itl_ms_p95": 8.515625027939677, "itl_ms_p99": 8.595749968662858, "itl_ms_mean": 8.367492063414483, "total_s": 0.5749587499303743, "peak_mem_gb": 17.226143856, "text": "1. **Create a Weekly Schedule:** Use a digital calendar or planner to block out time for classes, study sessions, assignments, and personal activities. Having a visual representation of your week help"}17{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:44:45", "prompt_tokens": 19, "gen_tokens": 64, "ttft_ms": 51.81904206983745, "prompt_tps": 370.7121119944423, "gen_tps": 121.67745683114906, "decode_tps_wall": 119.77846688130617, "itl_ms_p50": 8.337999926880002, "itl_ms_p95": 8.509374922141433, "itl_ms_p99": 8.653583005070686, "itl_ms_mean": 8.34874603121231, "total_s": 0.577790042036213, "peak_mem_gb": 17.249301776, "text": "Include details about temperature, precipitation, and weather patterns.\n\nGatineau, Qu\u00e9bec, experiences a humid continental climate with distinct seasonal variations. In winter (December\u2013February), tem"}18{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:44:45", "prompt_tokens": 25, "gen_tokens": 64, "ttft_ms": 58.237458928488195, "prompt_tps": 433.6403767509204, "gen_tps": 121.9397082106601, "decode_tps_wall": 120.03663974946353, "itl_ms_p50": 8.328375057317317, "itl_ms_p95": 8.410041918978095, "itl_ms_p99": 8.536917041055858, "itl_ms_mean": 8.330789682943196, "total_s": 0.5830772089539096, "peak_mem_gb": 17.262549952, "text": "import csv\nfrom datetime import datetime\nimport statistics\n\ndef median_inter_arrival_time(csv_file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time "}19{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 2, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:44:46", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 53.03687497507781, "prompt_tps": 400.2061405267773, "gen_tps": 122.84042573142314, "decode_tps_wall": 120.92373271713936, "itl_ms_p50": 8.24875000398606, "itl_ms_p95": 8.425749954767525, "itl_ms_p99": 8.515499997884035, "itl_ms_mean": 8.26967525340262, "total_s": 0.5740264159394428, "peak_mem_gb": 17.253640676, "text": "class LRUCache:\n def __init__(self, capacity):\n self.capacity = capacity\n self.cache = {}\n self.order = []\n\n def get(self, key):\n if key in self.cache:\n # "}20{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 2, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:44:47", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 48.9684579661116, "prompt_tps": 351.7829711443551, "gen_tps": 121.93942754370109, "decode_tps_wall": 120.03831691113533, "itl_ms_p50": 8.33029206842184, "itl_ms_p95": 8.438290911726654, "itl_ms_p99": 8.636665996164083, "itl_ms_mean": 8.330673286099993, "total_s": 0.5738008749904111, "peak_mem_gb": 17.226956788, "text": "Mixture-of-Experts (MoE) models are memory-bound during decoding because they rely on storing and accessing a large number of expert parameters, which significantly increases memory requirements. Duri"}21{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 2, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:44:47", "prompt_tokens": 13, "gen_tokens": 64, "ttft_ms": 43.77604101318866, "prompt_tps": 300.7414438468318, "gen_tps": 121.05787559700309, "decode_tps_wall": 119.16836550102333, "itl_ms_p50": 8.376457961276174, "itl_ms_p95": 8.59345798380673, "itl_ms_p99": 8.693208917975426, "itl_ms_mean": 8.39148876294198, "total_s": 0.5724398330785334, "peak_mem_gb": 17.221638736, "text": " if n < 0:\n raise ValueError(\"n must be non-negative\")\n if n == 0:\n return 0\n if n == 1:\n return 1\n return fibonacci(n - 1) + fibonacci(n - 2)\n\n# Example usage:\nprint(f"}22{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 2, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:44:48", "prompt_tokens": 19, "gen_tokens": 64, "ttft_ms": 51.64116702508181, "prompt_tps": 372.6351792877575, "gen_tps": 121.64076273747433, "decode_tps_wall": 119.74421314955649, "itl_ms_p50": 8.346999995410442, "itl_ms_p95": 8.42333398759365, "itl_ms_p99": 8.572999970056117, "itl_ms_mean": 8.351134252734482, "total_s": 0.5777626249473542, "peak_mem_gb": 17.249301776, "text": "Include details about temperature, precipitation, and weather patterns.\n\nGatineau, Qu\u00e9bec, experiences a humid continental climate with distinct seasonal variations. In winter (December\u2013February), tem"}23{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 2, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:44:48", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 47.79970902018249, "prompt_tps": 338.6786920728613, "gen_tps": 121.35481265622165, "decode_tps_wall": 119.46146588396937, "itl_ms_p50": 8.368958020582795, "itl_ms_p95": 8.484416990540922, "itl_ms_p99": 8.614916005171835, "itl_ms_mean": 8.370900127504553, "total_s": 0.5751664170529693, "peak_mem_gb": 17.226143856, "text": "1. **Create a Weekly Schedule:** Use a digital calendar or planner to block out time for classes, study sessions, assignments, and personal activities. Having a visual representation of your week help"}24{"backend": "mlx_lm", "capacity": null, "policy": null, "predictor": null, "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend mlx_lm --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 2, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:44:49", "prompt_tokens": 25, "gen_tokens": 64, "ttft_ms": 58.15908405929804, "prompt_tps": 434.2721836914425, "gen_tps": 121.31440349292212, "decode_tps_wall": 119.42142230634539, "itl_ms_p50": 8.356499951332808, "itl_ms_p95": 8.525499957613647, "itl_ms_p99": 8.59662494622171, "itl_ms_mean": 8.373707000698362, "total_s": 0.5857026251032948, "peak_mem_gb": 17.26254994, "text": "import csv\nfrom datetime import datetime\nimport statistics\n\ndef median_inter_arrival_time(csv_file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time "}25{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:44:54", "prompt_tokens": 25, "gen_tokens": 64, "ttft_ms": 1220.5764999380335, "prompt_tps": 27.00275251582446, "gen_tps": 119.63154427799184, "decode_tps_wall": 117.78116172783089, "itl_ms_p50": 8.376332931220531, "itl_ms_p95": 8.598791086114943, "itl_ms_p99": 14.143792097456753, "itl_ms_mean": 8.490322096761139, "total_s": 1.7554667920339853, "peak_mem_gb": 33.489018536, "text": "import csv\nfrom datetime import datetime\nimport statistics\n\ndef median_inter_arrival_time(csv_file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time ", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 34176, "stall_s": 0.0, "hits_device": 34176, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}26{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:44:55", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 48.77100000157952, "prompt_tps": 332.56485232655496, "gen_tps": 121.4452169164887, "decode_tps_wall": 119.55078327663087, "itl_ms_p50": 8.360707899555564, "itl_ms_p95": 8.474875008687377, "itl_ms_p99": 8.506624959409237, "itl_ms_mean": 8.364646157826341, "total_s": 0.575743707944639, "peak_mem_gb": 17.218282524, "text": "1. **Create a Weekly Schedule:** Use a digital calendar or planner to block out time for classes, study sessions, assignments, and personal activities. Having a visual representation of your week help", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 30720, "stall_s": 0.0, "hits_device": 30720, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}27{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:44:55", "prompt_tokens": 19, "gen_tokens": 64, "ttft_ms": 53.07695805095136, "prompt_tps": 361.7012308418058, "gen_tps": 121.23609680837703, "decode_tps_wall": 119.34598402398768, "itl_ms_p50": 8.373667020350695, "itl_ms_p95": 8.46533291041851, "itl_ms_p99": 8.867958909831941, "itl_ms_mean": 8.378999998852137, "total_s": 0.580953957978636, "peak_mem_gb": 17.239856048, "text": "Include details about temperature, precipitation, and weather patterns.\n\nGatineau, Qu\u00e9bec, experiences a humid continental climate with distinct seasonal variations. In winter (December\u2013February), tem", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 31872, "stall_s": 0.0, "hits_device": 31872, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}28{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:44:56", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 50.045416108332574, "prompt_tps": 343.29477135842853, "gen_tps": 121.54620061293649, "decode_tps_wall": 119.64870752061906, "itl_ms_p50": 8.346708957105875, "itl_ms_p95": 8.540916023775935, "itl_ms_p99": 8.558625006116927, "itl_ms_mean": 8.357800269824645, "total_s": 0.5765868331072852, "peak_mem_gb": 17.21856966, "text": "Mixture-of-Experts (MoE) models are memory-bound during decoding because they rely on storing and accessing a large number of expert parameters, which significantly increases memory requirements. Duri", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 31104, "stall_s": 0.0, "hits_device": 31104, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}29{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:44:57", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 55.06129201967269, "prompt_tps": 385.90881583195323, "gen_tps": 117.97438897158018, "decode_tps_wall": 116.13470519492213, "itl_ms_p50": 8.571584010496736, "itl_ms_p95": 8.948875009082258, "itl_ms_p99": 9.198999963700771, "itl_ms_mean": 8.61069047638762, "total_s": 0.5975347920320928, "peak_mem_gb": 17.243156824, "text": "class LRUCache:\n def __init__(self, capacity):\n self.capacity = capacity\n self.cache = {}\n self.order = []\n\n def get(self, key):\n if key in self.cache:\n # ", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 32640, "stall_s": 0.0, "hits_device": 32640, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}30{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:44:57", "prompt_tokens": 13, "gen_tokens": 64, "ttft_ms": 44.345750007778406, "prompt_tps": 296.6154511429461, "gen_tps": 120.9117500982621, "decode_tps_wall": 119.02583968272118, "itl_ms_p50": 8.385625085793436, "itl_ms_p95": 8.570082951337099, "itl_ms_p99": 8.722792030312121, "itl_ms_mean": 8.401537033182288, "total_s": 0.5736425830982625, "peak_mem_gb": 17.215356644, "text": " if n < 0:\n raise ValueError(\"n must be non-negative\")\n if n == 0:\n return 0\n if n == 1:\n return 1\n return fibonacci(n - 1) + fibonacci(n - 2)\n\n# Example usage:\nprint(f", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 29568, "stall_s": 0.0, "hits_device": 29568, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}31{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:44:58", "prompt_tokens": 25, "gen_tokens": 64, "ttft_ms": 60.059874900616705, "prompt_tps": 420.1707153604413, "gen_tps": 121.58243342111969, "decode_tps_wall": 119.68656369037812, "itl_ms_p50": 8.351624943315983, "itl_ms_p95": 8.393040974624455, "itl_ms_p99": 8.665792061947286, "itl_ms_mean": 8.355156745805981, "total_s": 0.5864347498863935, "peak_mem_gb": 17.250497668, "text": "import csv\nfrom datetime import datetime\nimport statistics\n\ndef median_inter_arrival_time(csv_file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time ", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 34176, "stall_s": 0.0, "hits_device": 34176, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}32{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:44:58", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 49.783666036091745, "prompt_tps": 344.82525692735203, "gen_tps": 122.00261589794333, "decode_tps_wall": 120.09875733097927, "itl_ms_p50": 8.330084034241736, "itl_ms_p95": 8.390332921408117, "itl_ms_p99": 8.473166031762958, "itl_ms_mean": 8.326480824810764, "total_s": 0.5743519579991698, "peak_mem_gb": 17.218569684, "text": "Mixture-of-Experts (MoE) models are memory-bound during decoding because they rely on storing and accessing a large number of expert parameters, which significantly increases memory requirements. Duri", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 31104, "stall_s": 0.0, "hits_device": 31104, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}33{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:44:59", "prompt_tokens": 19, "gen_tokens": 64, "ttft_ms": 52.693875040858984, "prompt_tps": 363.86588147938994, "gen_tps": 121.39097933124957, "decode_tps_wall": 119.49697464663882, "itl_ms_p50": 8.35495803039521, "itl_ms_p95": 8.45249998383224, "itl_ms_p99": 8.801499963738024, "itl_ms_mean": 8.368412697953838, "total_s": 0.5799038750119507, "peak_mem_gb": 17.239863212, "text": "Include details about temperature, precipitation, and weather patterns.\n\nGatineau, Qu\u00e9bec, experiences a humid continental climate with distinct seasonal variations. In winter (December\u2013February), tem", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 31872, "stall_s": 0.0, "hits_device": 31872, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}34{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:44:59", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 55.027499911375344, "prompt_tps": 385.0664480922441, "gen_tps": 121.53900657171316, "decode_tps_wall": 119.6423643508119, "itl_ms_p50": 8.351791999302804, "itl_ms_p95": 8.486833074130118, "itl_ms_p99": 8.650125004351139, "itl_ms_mean": 8.3582433816489, "total_s": 0.5815968329552561, "peak_mem_gb": 17.243156804, "text": "class LRUCache:\n def __init__(self, capacity):\n self.capacity = capacity\n self.cache = {}\n self.order = []\n\n def get(self, key):\n if key in self.cache:\n # ", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 32640, "stall_s": 0.0, "hits_device": 32640, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}35{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:45:00", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 48.99120796471834, "prompt_tps": 330.0295995197668, "gen_tps": 119.90046397486195, "decode_tps_wall": 118.03147842516533, "itl_ms_p50": 8.421833044849336, "itl_ms_p95": 8.914500009268522, "itl_ms_p99": 9.067249950021505, "itl_ms_mean": 8.472316142629891, "total_s": 0.5827471249504015, "peak_mem_gb": 17.218282524, "text": "1. **Create a Weekly Schedule:** Use a digital calendar or planner to block out time for classes, study sessions, assignments, and personal activities. Having a visual representation of your week help", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 30720, "stall_s": 0.0, "hits_device": 30720, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}36{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:45:01", "prompt_tokens": 13, "gen_tokens": 64, "ttft_ms": 44.39308401197195, "prompt_tps": 296.37510710304804, "gen_tps": 121.82079258300936, "decode_tps_wall": 119.92054796815864, "itl_ms_p50": 8.337458013556898, "itl_ms_p95": 8.392707910388708, "itl_ms_p99": 8.558874949812889, "itl_ms_mean": 8.338854491104565, "total_s": 0.5697409169515595, "peak_mem_gb": 17.215356624, "text": " if n < 0:\n raise ValueError(\"n must be non-negative\")\n if n == 0:\n return 0\n if n == 1:\n return 1\n return fibonacci(n - 1) + fibonacci(n - 2)\n\n# Example usage:\nprint(f", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 29568, "stall_s": 0.0, "hits_device": 29568, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}37{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:45:01", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 54.63941604830325, "prompt_tps": 387.86446016558983, "gen_tps": 121.27424188755138, "decode_tps_wall": 119.3818012321434, "itl_ms_p50": 8.351875003427267, "itl_ms_p95": 8.4734580013901, "itl_ms_p99": 8.981000049971044, "itl_ms_mean": 8.376486111609708, "total_s": 0.5823580410797149, "peak_mem_gb": 17.243156812, "text": "class LRUCache:\n def __init__(self, capacity):\n self.capacity = capacity\n self.cache = {}\n self.order = []\n\n def get(self, key):\n if key in self.cache:\n # ", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 32640, "stall_s": 0.0, "hits_device": 32640, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}38{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:45:02", "prompt_tokens": 13, "gen_tokens": 64, "ttft_ms": 44.173542060889304, "prompt_tps": 297.75367736190805, "gen_tps": 121.6031860206553, "decode_tps_wall": 119.70498422740722, "itl_ms_p50": 8.344833040609956, "itl_ms_p95": 8.480250020511448, "itl_ms_p99": 8.510250016115606, "itl_ms_mean": 8.353871030969517, "total_s": 0.5704674170119688, "peak_mem_gb": 17.215356644, "text": " if n < 0:\n raise ValueError(\"n must be non-negative\")\n if n == 0:\n return 0\n if n == 1:\n return 1\n return fibonacci(n - 1) + fibonacci(n - 2)\n\n# Example usage:\nprint(f", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 29568, "stall_s": 0.0, "hits_device": 29568, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}39{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:45:02", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 50.10233307257295, "prompt_tps": 342.8398592472836, "gen_tps": 121.59208703388818, "decode_tps_wall": 119.69455089682388, "itl_ms_p50": 8.344458998180926, "itl_ms_p95": 8.48733400925994, "itl_ms_p99": 8.629083051346242, "itl_ms_mean": 8.354599206959682, "total_s": 0.5764420831110328, "peak_mem_gb": 17.218569652, "text": "Mixture-of-Experts (MoE) models are memory-bound during decoding because they rely on storing and accessing a large number of expert parameters, which significantly increases memory requirements. Duri", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 31104, "stall_s": 0.0, "hits_device": 31104, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}40{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:45:03", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 48.36112505290657, "prompt_tps": 334.16526579787626, "gen_tps": 121.67285920481129, "decode_tps_wall": 119.77398856188454, "itl_ms_p50": 8.340667001903057, "itl_ms_p95": 8.455541916191578, "itl_ms_p99": 8.47962498664856, "itl_ms_mean": 8.349058188734547, "total_s": 0.574351790943183, "peak_mem_gb": 17.218290684, "text": "1. **Create a Weekly Schedule:** Use a digital calendar or planner to block out time for classes, study sessions, assignments, and personal activities. Having a visual representation of your week help", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 30720, "stall_s": 0.0, "hits_device": 30720, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}41{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:45:04", "prompt_tokens": 19, "gen_tokens": 64, "ttft_ms": 52.854583016596735, "prompt_tps": 362.84210416888635, "gen_tps": 121.97403551088667, "decode_tps_wall": 120.07064158084249, "itl_ms_p50": 8.331332937814295, "itl_ms_p95": 8.383834036067128, "itl_ms_p99": 8.41845793183893, "itl_ms_mean": 8.328430554164308, "total_s": 0.5775457079289481, "peak_mem_gb": 17.239863216, "text": "Include details about temperature, precipitation, and weather patterns.\n\nGatineau, Qu\u00e9bec, experiences a humid continental climate with distinct seasonal variations. In winter (December\u2013February), tem", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 31872, "stall_s": 0.0, "hits_device": 31872, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}42{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:45:04", "prompt_tokens": 25, "gen_tokens": 64, "ttft_ms": 59.81308396439999, "prompt_tps": 421.45989484963627, "gen_tps": 121.37909403443791, "decode_tps_wall": 119.48395266974292, "itl_ms_p50": 8.344041998498142, "itl_ms_p95": 8.558999979868531, "itl_ms_p99": 8.64954199641943, "itl_ms_mean": 8.369324730694412, "total_s": 0.587080541998148, "peak_mem_gb": 17.25049768, "text": "import csv\nfrom datetime import datetime\nimport statistics\n\ndef median_inter_arrival_time(csv_file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time ", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 34176, "stall_s": 0.0, "hits_device": 34176, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}43{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 2, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:45:05", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 55.06124999374151, "prompt_tps": 385.09145933521063, "gen_tps": 121.23431697523493, "decode_tps_wall": 119.34340303931936, "itl_ms_p50": 8.368332986719906, "itl_ms_p95": 8.470666012726724, "itl_ms_p99": 8.576125022955239, "itl_ms_mean": 8.379181207615941, "total_s": 0.5829496660735458, "peak_mem_gb": 17.243156804, "text": "class LRUCache:\n def __init__(self, capacity):\n self.capacity = capacity\n self.cache = {}\n self.order = []\n\n def get(self, key):\n if key in self.cache:\n # ", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 32640, "stall_s": 0.0, "hits_device": 32640, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}44{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 2, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:45:05", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 49.646333092823625, "prompt_tps": 345.721111045929, "gen_tps": 122.00395295728767, "decode_tps_wall": 120.09996893673218, "itl_ms_p50": 8.340833010151982, "itl_ms_p95": 8.404958061873913, "itl_ms_p99": 8.497833972796798, "itl_ms_mean": 8.326396824688548, "total_s": 0.5742093330482021, "peak_mem_gb": 17.218569684, "text": "Mixture-of-Experts (MoE) models are memory-bound during decoding because they rely on storing and accessing a large number of expert parameters, which significantly increases memory requirements. Duri", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 31104, "stall_s": 0.0, "hits_device": 31104, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}45{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 2, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:45:06", "prompt_tokens": 13, "gen_tokens": 64, "ttft_ms": 43.86004095431417, "prompt_tps": 299.62950087647863, "gen_tps": 121.56949062605008, "decode_tps_wall": 119.67134054503336, "itl_ms_p50": 8.344084024429321, "itl_ms_p95": 8.46070796251297, "itl_ms_p99": 8.677207981236279, "itl_ms_mean": 8.356219588128466, "total_s": 0.5703018750064075, "peak_mem_gb": 17.215356608, "text": " if n < 0:\n raise ValueError(\"n must be non-negative\")\n if n == 0:\n return 0\n if n == 1:\n return 1\n return fibonacci(n - 1) + fibonacci(n - 2)\n\n# Example usage:\nprint(f", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 29568, "stall_s": 0.0, "hits_device": 29568, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}46{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 2, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:45:06", "prompt_tokens": 19, "gen_tokens": 64, "ttft_ms": 53.19029197562486, "prompt_tps": 360.89625262896436, "gen_tps": 121.45359072568283, "decode_tps_wall": 119.55926290013397, "itl_ms_p50": 8.34145792759955, "itl_ms_p95": 8.495041984133422, "itl_ms_p99": 8.691542083397508, "itl_ms_mean": 8.364052903498449, "total_s": 0.5801256248960271, "peak_mem_gb": 17.239863216, "text": "Include details about temperature, precipitation, and weather patterns.\n\nGatineau, Qu\u00e9bec, experiences a humid continental climate with distinct seasonal variations. In winter (December\u2013February), tem", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 31872, "stall_s": 0.0, "hits_device": 31872, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}47{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 2, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:45:07", "prompt_tokens": 16, "gen_tokens": 64, "ttft_ms": 48.36929100565612, "prompt_tps": 334.04317433962683, "gen_tps": 121.9046360404261, "decode_tps_wall": 120.0016190010754, "itl_ms_p50": 8.329583099111915, "itl_ms_p95": 8.395583950914443, "itl_ms_p99": 8.49329202901572, "itl_ms_mean": 8.333220904219955, "total_s": 0.5733622079715133, "peak_mem_gb": 17.2182825, "text": "1. **Create a Weekly Schedule:** Use a digital calendar or planner to block out time for classes, study sessions, assignments, and personal activities. Having a visual representation of your week help", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 30720, "stall_s": 0.0, "hits_device": 30720, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}48{"backend": "titan", "capacity": 128, "policy": "lru", "predictor": "none", "m": 8, "prefetch_workers": 2, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--Qwen3-30B-A3B-Instruct-2507-4bit/snapshots/e9675aa3ca5f900ccef55267914466d55ab325fa --backend titan --capacity 128 --families general,code --repeats 3 --out benchmarks/raw/gateC/qwen3_m3u96a/baseline.jsonl", "env": {"host": "MacStudnpierre2", "chip": "Apple M3 Ultra", "ram_gb": 96, "macos": "26.6", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": ""}, "seed": 0, "warmup": false, "repeat": 2, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:45:08", "prompt_tokens": 25, "gen_tokens": 64, "ttft_ms": 59.67658292502165, "prompt_tps": 422.439421145433, "gen_tps": 121.7417052208992, "decode_tps_wall": 119.84254969519033, "itl_ms_p50": 8.343541994690895, "itl_ms_p95": 8.473082911223173, "itl_ms_p99": 8.57179204467684, "itl_ms_mean": 8.344281747538064, "total_s": 0.5853663330199197, "peak_mem_gb": 17.250497668, "text": "import csv\nfrom datetime import datetime\nimport statistics\n\ndef median_inter_arrival_time(csv_file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time ", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 34176, "stall_s": 0.0, "hits_device": 34176, "layers_complete": 48, "prefetch": null, "ssd_bytes": 0, "ssd_read_s": 0.0}49