SPB Git
spboucher.ai

spb/wp33_uqo

Public MIT

Gate B/C: sync-free fast path, bench_decode, capacity sweep, routing traces, link/net probes, LOG

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Simon-Pierre Boucher committed (Nov 1, 2026) parent 7ab0c87 Browse files

16 changed files +796 −3

16 changed files
Unified Split
added benchmarks/configs/workloads.yaml +34 −0
@@ -0,0 +1,34 @@
1 +# Prompt families for WP33. "fit" prompts may be used to fit predictors; "heldout" never are.
2 +# Prompts are short synthetic instructions written for this study (no third-party data).
3 +families:
4 + general:
5 + fit:
6 + - "Explain in three sentences why mixture-of-experts models are memory-bound during decoding."
7 + - "Write a short paragraph describing the climate of Gatineau, Québec, across the four seasons."
8 + - "Give five practical tips for a first-year university student learning to manage their time."
9 + heldout:
10 + - "Describe the main differences between a Mac Studio and a MacBook Pro for scientific computing."
11 + - "Summarize the history of the Rideau Canal in one paragraph."
12 + code:
13 + fit:
14 + - "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"\"\n"
15 + - "Write a Python function that parses a CSV file of timestamps and returns the median inter-arrival time in milliseconds.\n\n```python\n"
16 + - "Implement an LRU cache class in Python with get and put methods in O(1).\n\n```python\n"
17 + heldout:
18 + - "Write a C function that computes the dot product of two float arrays using NEON intrinsics.\n\n```c\n"
19 + - "Write a bash script that lists the ten largest files under a directory.\n\n```bash\n"
20 + reasoning:
21 + fit:
22 + - "Q: If a train travels 60 km in 45 minutes, what is its average speed in km/h? Think step by step.\nA:"
23 + - "Q: A rectangle has a perimeter of 36 cm and a length twice its width. What is its area? Think step by step.\nA:"
24 + - "Q: Three friends split a bill of 84 dollars; one pays twice as much as each of the others. How much does each pay? Think step by step.\nA:"
25 + heldout:
26 + - "Q: A tank fills in 6 hours with pipe A and 9 hours with pipe B. How long with both pipes? Think step by step.\nA:"
27 + - "Q: What is the smallest positive integer divisible by 2, 3, 4, 5 and 6? Think step by step.\nA:"
28 + multilingual:
29 + fit:
30 + - "Les modèles à mélange d'experts activent seulement une fraction de leurs paramètres parce que"
31 + - "Explique en français pourquoi la mémoire unifiée des puces Apple est utile pour l'inférence des grands modèles de langage."
32 + heldout:
33 + - "Décris en quelques phrases le fonctionnement d'un cache LRU."
34 +output_lengths: [64, 256]
added benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl +108 −0
@@ -0,0 +1,108 @@
1 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:21:02", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 39.72374997101724, "prompt_tps": 834.4829898283148, "gen_tps": 495.40988057311176, "decode_tps_wall": 487.98277719045575, "itl_ms_p50": 2.0490000024437904, "itl_ms_p95": 2.0828749984502792, "itl_ms_p99": 2.1720000077039003, "itl_ms_mean": 2.0492526514100886, "total_s": 0.16882666700985283, "peak_mem_gb": 3.98276822, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:"}
2 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:21:02", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 56.30945798475295, "prompt_tps": 303.49578253724593, "gen_tps": 489.52352718064105, "decode_tps_wall": 481.9502922638398, "itl_ms_p50": 2.070290967822075, "itl_ms_p95": 2.118458040058613, "itl_ms_p99": 2.2464169887825847, "itl_ms_mean": 2.0749027774270092, "total_s": 0.18702833296265453, "peak_mem_gb": 3.955366956, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities"}
3 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:21:03", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 15.91620803810656, "prompt_tps": 1344.9539895483329, "gen_tps": 495.96752031786707, "decode_tps_wall": 488.273681567562, "itl_ms_p50": 2.0429580472409725, "itl_ms_p95": 2.0810830174013972, "itl_ms_p99": 2.12991691660136, "itl_ms_mean": 2.0480317447984975, "total_s": 0.1449422079604119, "peak_mem_gb": 3.9627891, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro"}
4 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:21:03", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 15.706375008448958, "prompt_tps": 1295.8123058152396, "gen_tps": 491.7853257156468, "decode_tps_wall": 484.1625662348439, "itl_ms_p50": 2.062792074866593, "itl_ms_p95": 2.089542103931308, "itl_ms_p99": 2.1462079603224993, "itl_ms_mean": 2.065421967205429, "total_s": 0.14582795894239098, "peak_mem_gb": 3.961374396, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e"}
5 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:21:03", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 17.17287499923259, "prompt_tps": 1419.4219059255133, "gen_tps": 496.3523296299187, "decode_tps_wall": 488.64908860873936, "itl_ms_p50": 2.042959094978869, "itl_ms_p95": 2.0856671035289764, "itl_ms_p99": 2.1247080294415355, "itl_ms_mean": 2.0464583344402767, "total_s": 0.14609975006897002, "peak_mem_gb": 3.968936172, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n"}
6 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:21:03", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 15.81437501590699, "prompt_tps": 1352.78405312007, "gen_tps": 496.1636167434886, "decode_tps_wall": 488.46139452297734, "itl_ms_p50": 2.044374938122928, "itl_ms_p95": 2.067083027213812, "itl_ms_p99": 2.1191249834373593, "itl_ms_mean": 2.047244697764871, "total_s": 0.14479079097509384, "peak_mem_gb": 3.962789484, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b"}
7 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:21:03", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 19.81274993158877, "prompt_tps": 1584.9007851160345, "gen_tps": 494.1771685310257, "decode_tps_wall": 486.5093388793429, "itl_ms_p50": 2.0525839645415545, "itl_ms_p95": 2.0967089803889394, "itl_ms_p99": 2.1160830510780215, "itl_ms_mean": 2.0554590016780865, "total_s": 0.14930666703730822, "peak_mem_gb": 3.982774172, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:"}
8 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:21:03", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 15.75566700194031, "prompt_tps": 1290.1144977825745, "gen_tps": 495.20831047014445, "decode_tps_wall": 487.538428726157, "itl_ms_p50": 2.049333998002112, "itl_ms_p95": 2.0877079805359244, "itl_ms_p99": 2.1225420059636235, "itl_ms_mean": 2.051120365245475, "total_s": 0.14497625001240522, "peak_mem_gb": 3.961374812, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e"}
9 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:21:04", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 15.857833088375628, "prompt_tps": 1345.1514458776778, "gen_tps": 496.4724893134941, "decode_tps_wall": 488.76566300970944, "itl_ms_p50": 2.0430420991033316, "itl_ms_p95": 2.085541025735438, "itl_ms_p99": 2.1406670566648245, "itl_ms_mean": 2.045970238257377, "total_s": 0.14475395809859037, "peak_mem_gb": 3.9627891, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro"}
10 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:21:04", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 17.01670791953802, "prompt_tps": 1431.3261329128766, "gen_tps": 495.4624540256374, "decode_tps_wall": 487.7751353510595, "itl_ms_p50": 2.0462919492274523, "itl_ms_p95": 2.078374964185059, "itl_ms_p99": 2.129417029209435, "itl_ms_mean": 2.050125001308819, "total_s": 0.1461745830019936, "peak_mem_gb": 3.968936172, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n"}
11 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:21:04", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 14.221542049199343, "prompt_tps": 1216.4651850269463, "gen_tps": 492.6669298210617, "decode_tps_wall": 485.0156795514458, "itl_ms_p50": 2.0607910118997097, "itl_ms_p95": 2.07779195625335, "itl_ms_p99": 2.0889999577775598, "itl_ms_mean": 2.0617890145836606, "total_s": 0.14411424996796995, "peak_mem_gb": 3.955366956, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities"}
12 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:21:04", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 15.882707899436355, "prompt_tps": 1344.0287295760597, "gen_tps": 494.4592248730498, "decode_tps_wall": 486.7945690982775, "itl_ms_p50": 2.0522079430520535, "itl_ms_p95": 2.082083956338465, "itl_ms_p99": 2.1311250748112798, "itl_ms_mean": 2.054254635281506, "total_s": 0.14530074992217124, "peak_mem_gb": 3.9627891, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b"}
13 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:21:04", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 17.00154202990234, "prompt_tps": 1433.024811557992, "gen_tps": 496.3184877742318, "decode_tps_wall": 488.61434740527216, "itl_ms_p50": 2.0427920389920473, "itl_ms_p95": 2.070083050057292, "itl_ms_p99": 2.129291999153793, "itl_ms_mean": 2.0466038406575247, "total_s": 0.1459375839913264, "peak_mem_gb": 3.968936172, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n"}
14 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:21:04", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 15.752332983538508, "prompt_tps": 1353.7433143696508, "gen_tps": 496.37879618904157, "decode_tps_wall": 488.6964701371299, "itl_ms_p50": 2.043832908384502, "itl_ms_p95": 2.064540982246399, "itl_ms_p99": 2.0979580003768206, "itl_ms_mean": 2.046259920231052, "total_s": 0.14466670795809478, "peak_mem_gb": 3.9627891, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b"}
15 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:21:04", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 15.79941704403609, "prompt_tps": 1285.7085390900506, "gen_tps": 494.06176738300985, "decode_tps_wall": 486.39775214129037, "itl_ms_p50": 2.052792115136981, "itl_ms_p95": 2.0957919768989086, "itl_ms_p99": 2.107042004354298, "itl_ms_mean": 2.055930553950251, "total_s": 0.1453230419429019, "peak_mem_gb": 3.961374396, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e"}
16 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:21:05", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 14.156999997794628, "prompt_tps": 1223.0361877496705, "gen_tps": 495.18324823028536, "decode_tps_wall": 487.49095529570127, "itl_ms_p50": 2.046832931227982, "itl_ms_p95": 2.0830000285059214, "itl_ms_p99": 2.134209033101797, "itl_ms_mean": 2.0513201099155203, "total_s": 0.14339016692247242, "peak_mem_gb": 3.955366956, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities"}
17 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:21:05", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 15.959167038090527, "prompt_tps": 1337.1431350760718, "gen_tps": 494.89797664621176, "decode_tps_wall": 487.22029696939296, "itl_ms_p50": 2.0509579917415977, "itl_ms_p95": 2.088000066578388, "itl_ms_p99": 2.1234159357845783, "itl_ms_mean": 2.0524596496086036, "total_s": 0.14526412496343255, "peak_mem_gb": 3.962789484, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro"}
18 +{"backend": "mlx_lm", "capacity": null, "policy": null, "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend mlx_lm --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:21:05", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 20.044749951921403, "prompt_tps": 1566.6818677482277, "gen_tps": 491.28229909069813, "decode_tps_wall": 483.662012249418, "itl_ms_p50": 2.06391594838351, "itl_ms_p95": 2.0992079516872764, "itl_ms_p99": 2.1296250633895397, "itl_ms_mean": 2.0675595243653606, "total_s": 0.15030099998693913, "peak_mem_gb": 3.982774172, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:"}
19 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:21:06", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 329.6320419758558, "prompt_tps": 94.7002241585137, "gen_tps": 181.14969926530358, "decode_tps_wall": 178.36237856207802, "itl_ms_p50": 5.449709016829729, "itl_ms_p95": 6.781209027394652, "itl_ms_p99": 9.859625017270446, "itl_ms_mean": 5.606563492042441, "total_s": 0.6828455419745296, "peak_mem_gb": 3.929754756, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:", "cache_hit_rate": 0.9043220059386341, "cache_misses": 870, "cache_hits": 8223, "stall_s": 0.3223369166953489, "ssd_bytes": 3078881280, "ssd_read_s": 0.11134762840811163}
20 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:21:07", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 56.86612497083843, "prompt_tps": 300.2150087969547, "gen_tps": 191.98423350442908, "decode_tps_wall": 188.9900306009692, "itl_ms_p50": 5.1426669815555215, "itl_ms_p95": 5.831666057929397, "itl_ms_p99": 7.295625051483512, "itl_ms_mean": 5.2912843964313945, "total_s": 0.39021704194601625, "peak_mem_gb": 3.929588392, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities", "cache_hit_rate": 0.9862114978316469, "cache_misses": 124, "cache_hits": 8869, "stall_s": 0.04619717027526349, "ssd_bytes": 438829056, "ssd_read_s": 0.016356685431674123}
21 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:21:07", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 32.238165964372456, "prompt_tps": 656.1577293701124, "gen_tps": 200.61052992889762, "decode_tps_wall": 197.48156149296597, "itl_ms_p50": 5.038208095356822, "itl_ms_p95": 5.428042029961944, "itl_ms_p99": 5.54720894433558, "itl_ms_mean": 5.06376388985368, "total_s": 0.3512552910251543, "peak_mem_gb": 3.929605252, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro", "cache_hit_rate": 0.9975903614457832, "cache_misses": 22, "cache_hits": 9108, "stall_s": 0.008630876312963665, "ssd_bytes": 77856768, "ssd_read_s": 0.0029860418289899826}
22 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:21:07", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 26.629292056895792, "prompt_tps": 757.7647398416595, "gen_tps": 200.4772297633657, "decode_tps_wall": 197.3492549814161, "itl_ms_p50": 5.05724991671741, "itl_ms_p95": 5.237832898274064, "itl_ms_p99": 5.673333071172237, "itl_ms_mean": 5.067158728793618, "total_s": 0.34586029197089374, "peak_mem_gb": 3.929588392, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e", "cache_hit_rate": 0.9996695671329442, "cache_misses": 3, "cache_hits": 9076, "stall_s": 0.0013753329403698444, "ssd_bytes": 10616832, "ssd_read_s": 0.0004185824654996395}
23 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:21:08", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 27.688957983627915, "prompt_tps": 873.6790428635898, "gen_tps": 200.95101137082807, "decode_tps_wall": 197.81324804817484, "itl_ms_p50": 5.05504198372364, "itl_ms_p95": 5.1665829960256815, "itl_ms_p99": 5.3240410052239895, "itl_ms_mean": 5.0552731420519565, "total_s": 0.3461711659329012, "peak_mem_gb": 3.928689344, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 9017, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
24 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:21:08", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 25.804458069615066, "prompt_tps": 820.6009893405353, "gen_tps": 203.46858301797283, "decode_tps_wall": 200.29211941674305, "itl_ms_p50": 4.980208002962172, "itl_ms_p95": 5.2422500448301435, "itl_ms_p99": 5.674666026607156, "itl_ms_mean": 4.9927076657435725, "total_s": 0.3403450410114601, "peak_mem_gb": 3.929588392, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b", "cache_hit_rate": 0.9998884675440554, "cache_misses": 1, "cache_hits": 8965, "stall_s": 0.0006659169448539615, "ssd_bytes": 3538944, "ssd_read_s": 0.00015608198009431362}
25 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:21:08", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 31.011958024464548, "prompt_tps": 1006.4281513639252, "gen_tps": 200.9631580278932, "decode_tps_wall": 197.82463553769415, "itl_ms_p50": 5.053083994425833, "itl_ms_p95": 5.198208964429796, "itl_ms_p99": 5.260375095531344, "itl_ms_mean": 5.054982142552497, "total_s": 0.34947583300527185, "peak_mem_gb": 3.929381616, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 9093, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
26 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:21:09", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 26.16050001233816, "prompt_tps": 770.3418386550496, "gen_tps": 201.50739554377063, "decode_tps_wall": 198.36115901450472, "itl_ms_p50": 5.0169999012723565, "itl_ms_p95": 5.287500098347664, "itl_ms_p99": 5.370040889829397, "itl_ms_mean": 5.041309523337063, "total_s": 0.3437629999825731, "peak_mem_gb": 3.928064384, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 9079, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
27 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:21:09", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 26.172082987613976, "prompt_tps": 808.4787701176173, "gen_tps": 205.7662507169072, "decode_tps_wall": 202.55375791393593, "itl_ms_p50": 4.924334003590047, "itl_ms_p95": 5.142958019860089, "itl_ms_p99": 5.242833984084427, "itl_ms_mean": 4.936960984080557, "total_s": 0.33720062498468906, "peak_mem_gb": 3.928130512, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 9130, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
28 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:21:09", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 27.733374969102442, "prompt_tps": 871.9570333386699, "gen_tps": 200.65785995184612, "decode_tps_wall": 197.5261158266244, "itl_ms_p50": 5.062583019025624, "itl_ms_p95": 5.259832949377596, "itl_ms_p99": 5.296709015965462, "itl_ms_mean": 5.062621698478266, "total_s": 0.3466785419732332, "peak_mem_gb": 3.928689344, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 9017, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
29 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:21:10", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 24.21833307016641, "prompt_tps": 707.8147706328958, "gen_tps": 200.95358755354184, "decode_tps_wall": 197.817155523319, "itl_ms_p50": 5.048083956353366, "itl_ms_p95": 5.212375079281628, "itl_ms_p99": 5.291666951961815, "itl_ms_mean": 5.055173285423764, "total_s": 0.3426942500518635, "peak_mem_gb": 3.92766056, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 8993, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
30 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:21:10", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 26.095250039361417, "prompt_tps": 810.8838515721795, "gen_tps": 201.33655511190324, "decode_tps_wall": 198.19495735018927, "itl_ms_p50": 5.0327080534771085, "itl_ms_p95": 5.182333989068866, "itl_ms_p99": 5.462333909235895, "itl_ms_mean": 5.045537047812508, "total_s": 0.34396408405154943, "peak_mem_gb": 3.928130512, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 8966, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
31 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:21:10", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 27.47062500566244, "prompt_tps": 879.9414817727267, "gen_tps": 199.81769262252482, "decode_tps_wall": 196.69871433601702, "itl_ms_p50": 5.088291014544666, "itl_ms_p95": 5.198166938498616, "itl_ms_p99": 5.410499987192452, "itl_ms_mean": 5.083917316773699, "total_s": 0.3477574159624055, "peak_mem_gb": 3.928689344, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 9017, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
32 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:21:11", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 26.461125002242625, "prompt_tps": 799.4683007009005, "gen_tps": 201.9559752238423, "decode_tps_wall": 198.80339294530378, "itl_ms_p50": 5.020625074394047, "itl_ms_p95": 5.193999968469143, "itl_ms_p99": 5.309334024786949, "itl_ms_mean": 5.030095237233336, "total_s": 0.3433571249479428, "peak_mem_gb": 3.928130512, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 8966, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
33 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:21:11", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 25.774667039513588, "prompt_tps": 781.998397464609, "gen_tps": 203.7019649146816, "decode_tps_wall": 200.5214353266226, "itl_ms_p50": 4.983750055544078, "itl_ms_p95": 5.131540936417878, "itl_ms_p99": 5.318709067068994, "itl_ms_mean": 4.986998015305116, "total_s": 0.3399555420037359, "peak_mem_gb": 3.928064384, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 9079, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
34 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:21:12", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 24.23379197716713, "prompt_tps": 707.3766355472346, "gen_tps": 203.20427748675152, "decode_tps_wall": 200.0314334901074, "itl_ms_p50": 4.997249925509095, "itl_ms_p95": 5.207875045016408, "itl_ms_p99": 5.304208025336266, "itl_ms_mean": 4.9992142862359445, "total_s": 0.33918429201003164, "peak_mem_gb": 3.92766056, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 8993, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
35 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:21:12", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 26.57312504015863, "prompt_tps": 796.3771070110516, "gen_tps": 202.12036899708775, "decode_tps_wall": 198.96438533656544, "itl_ms_p50": 5.032500019297004, "itl_ms_p95": 5.246084067039192, "itl_ms_p99": 5.535790929570794, "itl_ms_mean": 5.026025126599485, "total_s": 0.3432127080159262, "peak_mem_gb": 3.928130512, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 9130, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
36 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:21:12", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 30.878957943059504, "prompt_tps": 1010.8036650746834, "gen_tps": 201.38351480076437, "decode_tps_wall": 198.23965265625182, "itl_ms_p50": 5.030249943956733, "itl_ms_p95": 5.231749964877963, "itl_ms_p99": 5.3052910370752215, "itl_ms_mean": 5.044399476092723, "total_s": 0.34867612493690103, "peak_mem_gb": 3.929381616, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 9093, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
37 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:21:14", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 425.53529201541096, "prompt_tps": 73.27317569029647, "gen_tps": 79.55890800749668, "decode_tps_wall": 78.32272505490657, "itl_ms_p50": 11.851791990920901, "itl_ms_p95": 20.256874966435134, "itl_ms_p99": 27.432292001321912, "itl_ms_mean": 12.767686508595943, "total_s": 1.2298995420569554, "peak_mem_gb": 2.563102512, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:", "cache_hit_rate": 0.7804047285113204, "cache_misses": 2192, "cache_hits": 7790, "stall_s": 0.8210924054728821, "ssd_bytes": 7757365248, "ssd_read_s": 0.277750022476539}
38 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:21:16", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 234.05624995939434, "prompt_tps": 72.72190609305187, "gen_tps": 58.68487651983278, "decode_tps_wall": 57.76846605046764, "itl_ms_p50": 15.194082981906831, "itl_ms_p95": 30.03133402671665, "itl_ms_p99": 36.73133405391127, "itl_ms_mean": 17.310482143084446, "total_s": 1.3246166249737144, "peak_mem_gb": 2.344104644, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities", "cache_hit_rate": 0.7268434045293549, "cache_misses": 2545, "cache_hits": 6772, "stall_s": 0.9297578997211531, "ssd_bytes": 9006612480, "ssd_read_s": 0.32405576773453504}
39 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:21:17", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 316.2997920298949, "prompt_tps": 66.45133271046778, "gen_tps": 65.125824731795, "decode_tps_wall": 64.10890238964947, "itl_ms_p50": 14.257167000323534, "itl_ms_p95": 25.306582916527987, "itl_ms_p99": 28.42041605617851, "itl_ms_mean": 15.598457666956598, "total_s": 1.2990026250481606, "peak_mem_gb": 2.456618072, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro", "cache_hit_rate": 0.7467740270465573, "cache_misses": 2453, "cache_hits": 7234, "stall_s": 0.9041312955087051, "ssd_bytes": 8681029632, "ssd_read_s": 0.3126047569094226}
40 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:21:18", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 287.84008405636996, "prompt_tps": 69.54667174881018, "gen_tps": 71.60937310657043, "decode_tps_wall": 70.49127523440887, "itl_ms_p50": 13.404457946307957, "itl_ms_p95": 22.357249981723726, "itl_ms_p99": 30.705374898388982, "itl_ms_mean": 14.186152778122397, "total_s": 1.181567709078081, "peak_mem_gb": 2.344158756, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e", "cache_hit_rate": 0.7723637129025468, "cache_misses": 2163, "cache_hits": 7339, "stall_s": 0.7902283143484965, "ssd_bytes": 7654735872, "ssd_read_s": 0.2740186400478706}
41 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:21:19", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 226.31575004197657, "prompt_tps": 106.17615219234968, "gen_tps": 102.46422457832435, "decode_tps_wall": 100.8647080186665, "itl_ms_p50": 9.849249967373908, "itl_ms_p95": 13.08716693893075, "itl_ms_p99": 15.318332938477397, "itl_ms_mean": 9.914270507925679, "total_s": 0.8509147920412943, "peak_mem_gb": 2.429946908, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n", "cache_hit_rate": 0.8701448666806635, "cache_misses": 1237, "cache_hits": 8289, "stall_s": 0.4618326333584264, "ssd_bytes": 4377673728, "ssd_read_s": 0.15441152325365692}
42 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:21:20", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 186.36695900931954, "prompt_tps": 112.82464835008798, "gen_tps": 109.56454215855602, "decode_tps_wall": 107.85444280463179, "itl_ms_p50": 8.80258297547698, "itl_ms_p95": 14.254458947107196, "itl_ms_p99": 21.536207990720868, "itl_ms_mean": 9.27175528421584, "total_s": 0.7704875419149175, "peak_mem_gb": 2.33495028, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b", "cache_hit_rate": 0.8889483256659891, "cache_misses": 1038, "cache_hits": 8309, "stall_s": 0.38527683925349265, "ssd_bytes": 3673423872, "ssd_read_s": 0.13022124976851046}
43 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:21:21", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 309.10558300092816, "prompt_tps": 100.36854425509433, "gen_tps": 79.77614892449053, "decode_tps_wall": 78.53062959358869, "itl_ms_p50": 11.605750070884824, "itl_ms_p95": 20.568582927808166, "itl_ms_p99": 27.734375093132257, "itl_ms_mean": 12.733884920765247, "total_s": 1.1113403330091387, "peak_mem_gb": 2.563021556, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:", "cache_hit_rate": 0.8089561210178321, "cache_misses": 1907, "cache_hits": 8075, "stall_s": 0.7136938393814489, "ssd_bytes": 6748766208, "ssd_read_s": 0.24100998986978084}
44 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:21:22", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 268.45287496689707, "prompt_tps": 74.566141561823, "gen_tps": 71.23052116519774, "decode_tps_wall": 70.11832138985713, "itl_ms_p50": 13.418458052910864, "itl_ms_p95": 22.930125007405877, "itl_ms_p99": 30.644666054286063, "itl_ms_mean": 14.261607810603602, "total_s": 1.166934167034924, "peak_mem_gb": 2.344158756, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e", "cache_hit_rate": 0.7780467270048411, "cache_misses": 2109, "cache_hits": 7393, "stall_s": 0.7733973663998768, "ssd_bytes": 7463632896, "ssd_read_s": 0.2672339769778773}
45 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:21:23", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 310.4657910298556, "prompt_tps": 67.69599501653408, "gen_tps": 65.14431142958584, "decode_tps_wall": 64.1271549904959, "itl_ms_p50": 14.221665915101767, "itl_ms_p95": 25.311416946351528, "itl_ms_p99": 28.525167028419673, "itl_ms_mean": 15.594017856370007, "total_s": 1.292888915981166, "peak_mem_gb": 2.456618072, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro", "cache_hit_rate": 0.7477031072571487, "cache_misses": 2444, "cache_hits": 7243, "stall_s": 0.898625454865396, "ssd_bytes": 8649179136, "ssd_read_s": 0.3096106108278036}
46 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:21:24", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 243.2054589735344, "prompt_tps": 98.78082849727396, "gen_tps": 102.47193530235688, "decode_tps_wall": 100.87277640826169, "itl_ms_p50": 9.648624924011528, "itl_ms_p95": 12.92775000911206, "itl_ms_p99": 15.386832994408906, "itl_ms_mean": 9.913477507079877, "total_s": 0.8677545419195667, "peak_mem_gb": 2.429946908, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n", "cache_hit_rate": 0.8650010497585555, "cache_misses": 1286, "cache_hits": 8240, "stall_s": 0.4790720029268414, "ssd_bytes": 4551081984, "ssd_read_s": 0.16008254361804575}
47 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:21:25", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 251.05866696685553, "prompt_tps": 67.77467172292062, "gen_tps": 58.75528309793346, "decode_tps_wall": 57.837753962931025, "itl_ms_p50": 14.979999978095293, "itl_ms_p95": 30.12245800346136, "itl_ms_p99": 36.39079199638218, "itl_ms_mean": 17.289744699299927, "total_s": 1.340312583022751, "peak_mem_gb": 2.344104644, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities", "cache_hit_rate": 0.7210475474938285, "cache_misses": 2599, "cache_hits": 6718, "stall_s": 0.9478199649602175, "ssd_bytes": 9197715456, "ssd_read_s": 0.32924720144364983}
48 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:21:26", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 213.60024996101856, "prompt_tps": 98.42956054487522, "gen_tps": 109.85172591805348, "decode_tps_wall": 108.13736541810547, "itl_ms_p50": 8.7769579840824, "itl_ms_p95": 14.306708006188273, "itl_ms_p99": 21.41754108015448, "itl_ms_mean": 9.247497348706164, "total_s": 0.7961925829295069, "peak_mem_gb": 2.33495028, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b", "cache_hit_rate": 0.8810313469562426, "cache_misses": 1112, "cache_hits": 8235, "stall_s": 0.4116241753799841, "ssd_bytes": 3935305728, "ssd_read_s": 0.13939634908456355}
49 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:21:27", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 221.26100002788007, "prompt_tps": 108.58885049483203, "gen_tps": 102.80656575616005, "decode_tps_wall": 101.20173705594644, "itl_ms_p50": 9.573541930876672, "itl_ms_p95": 12.916042003780603, "itl_ms_p99": 15.154042048379779, "itl_ms_mean": 9.88125331729414, "total_s": 0.8437799590174109, "peak_mem_gb": 2.429946908, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n", "cache_hit_rate": 0.871824480369515, "cache_misses": 1221, "cache_hits": 8305, "stall_s": 0.45552571560256183, "ssd_bytes": 4321050624, "ssd_read_s": 0.15220616618171334}
50 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:21:28", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 186.822707997635, "prompt_tps": 112.54225379247839, "gen_tps": 109.72330653143135, "decode_tps_wall": 108.01093219787249, "itl_ms_p50": 8.724165963940322, "itl_ms_p95": 14.333250001072884, "itl_ms_p99": 21.445584017783403, "itl_ms_mean": 9.25832209436016, "total_s": 0.7700969999423251, "peak_mem_gb": 2.33495028, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b", "cache_hit_rate": 0.8889483256659891, "cache_misses": 1038, "cache_hits": 8309, "stall_s": 0.38553437287919223, "ssd_bytes": 3673423872, "ssd_read_s": 0.12981615355238318}
51 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:21:29", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 276.82854095473886, "prompt_tps": 72.30495682918664, "gen_tps": 71.49169516841678, "decode_tps_wall": 70.375344947558, "itl_ms_p50": 13.407708029262722, "itl_ms_p95": 22.427041083574295, "itl_ms_p99": 30.777791980654, "itl_ms_mean": 14.209521825366195, "total_s": 1.1720284159528092, "peak_mem_gb": 2.344158756, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e", "cache_hit_rate": 0.775520942959377, "cache_misses": 2133, "cache_hits": 7369, "stall_s": 0.7805797809269279, "ssd_bytes": 7548567552, "ssd_read_s": 0.26919061376247555}
52 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:21:30", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 238.3074579993263, "prompt_tps": 71.41221864217628, "gen_tps": 56.271502971735785, "decode_tps_wall": 55.39272746269912, "itl_ms_p50": 15.475874999538064, "itl_ms_p95": 31.68750007171184, "itl_ms_p99": 39.46708294097334, "itl_ms_mean": 18.052911380350235, "total_s": 1.375640874961391, "peak_mem_gb": 2.344104644, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities", "cache_hit_rate": 0.7246967908124933, "cache_misses": 2565, "cache_hits": 6752, "stall_s": 0.9615615240763873, "ssd_bytes": 9077391360, "ssd_read_s": 0.3278370131738484}
53 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:21:32", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 315.45133399777114, "prompt_tps": 66.6238634657119, "gen_tps": 64.8262455743941, "decode_tps_wall": 63.814003451923206, "itl_ms_p50": 14.192792004905641, "itl_ms_p95": 25.252875057049096, "itl_ms_p99": 28.743417002260685, "itl_ms_mean": 15.670541666507248, "total_s": 1.3026954589877278, "peak_mem_gb": 2.456618072, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro", "cache_hit_rate": 0.7467740270465573, "cache_misses": 2453, "cache_hits": 7234, "stall_s": 0.9055798861663789, "ssd_bytes": 8681029632, "ssd_read_s": 0.31103876035194844}
54 +{"backend": "titan", "capacity": 32, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 32 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:21:33", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 325.80550003331155, "prompt_tps": 95.22246879459341, "gen_tps": 79.97616120354166, "decode_tps_wall": 78.72761596674972, "itl_ms_p50": 11.649542022496462, "itl_ms_p95": 20.00987494830042, "itl_ms_p99": 27.577958069741726, "itl_ms_mean": 12.70202314296353, "total_s": 1.126032958040014, "peak_mem_gb": 2.563021556, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:", "cache_hit_rate": 0.8046483670607093, "cache_misses": 1950, "cache_hits": 8032, "stall_s": 0.7278052222682163, "ssd_bytes": 6900940800, "ssd_read_s": 0.246589750982821}
55 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:21:36", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 702.6507079135627, "prompt_tps": 44.26046476362657, "gen_tps": 47.399583529616976, "decode_tps_wall": 46.66122717391932, "itl_ms_p50": 21.285625058226287, "itl_ms_p95": 32.665375038050115, "itl_ms_p99": 35.61575000640005, "itl_ms_mean": 21.43106944600327, "total_s": 2.0528080830117688, "peak_mem_gb": 1.940199072, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:", "cache_hit_rate": 0.595850622406639, "cache_misses": 4383, "cache_hits": 6462, "stall_s": 1.6319043410476297, "ssd_bytes": 15511191552, "ssd_read_s": 0.5526817358331755}
56 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:21:38", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 458.6171250557527, "prompt_tps": 37.090649428478585, "gen_tps": 36.20960410116734, "decode_tps_wall": 35.64402818222656, "itl_ms_p50": 27.957792044617236, "itl_ms_p95": 45.053667039610445, "itl_ms_p99": 47.38050000742078, "itl_ms_mean": 28.055190476441076, "total_s": 2.2260941250715405, "peak_mem_gb": 1.577326676, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities", "cache_hit_rate": 0.4891779290722488, "cache_misses": 5027, "cache_hits": 4814, "stall_s": 1.8262214860878885, "ssd_bytes": 17790271488, "ssd_read_s": 0.6342345520388335}
57 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:21:40", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 577.1687499945983, "prompt_tps": 36.40297578412628, "gen_tps": 36.23767184851734, "decode_tps_wall": 35.67168882345696, "itl_ms_p50": 27.893458027392626, "itl_ms_p95": 38.65333297289908, "itl_ms_p99": 45.666416990570724, "itl_ms_mean": 28.03343584177099, "total_s": 2.3432752080261707, "peak_mem_gb": 1.718436724, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro", "cache_hit_rate": 0.4817049435578046, "cache_misses": 5326, "cache_hits": 4950, "stall_s": 1.9390191773418337, "ssd_bytes": 18848415744, "ssd_read_s": 0.6713228703010827}
58 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:21:42", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 516.700959065929, "prompt_tps": 38.727254568564575, "gen_tps": 38.903928427483486, "decode_tps_wall": 38.29628345173323, "itl_ms_p50": 25.67837492097169, "itl_ms_p95": 39.727707975544035, "itl_ms_p99": 43.37387497071177, "itl_ms_mean": 26.112194444673758, "total_s": 2.1617692090803757, "peak_mem_gb": 1.65542326, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e", "cache_hit_rate": 0.5216617210682493, "cache_misses": 4836, "cache_hits": 5274, "stall_s": 1.7597313986625522, "ssd_bytes": 17114333184, "ssd_read_s": 0.6105407294817269}
59 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:21:44", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 464.8777920519933, "prompt_tps": 51.65768711939343, "gen_tps": 50.09306841856067, "decode_tps_wall": 49.31075018476933, "itl_ms_p50": 19.387583015486598, "itl_ms_p95": 28.089917032048106, "itl_ms_p99": 29.232041910290718, "itl_ms_mean": 20.279553571035937, "total_s": 1.7424896670272574, "peak_mem_gb": 1.712654968, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n", "cache_hit_rate": 0.6435411373851866, "cache_misses": 3648, "cache_hits": 6586, "stall_s": 1.3370828725164756, "ssd_bytes": 12910067712, "ssd_read_s": 0.45644294121302664}
60 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:21:46", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 417.1714170370251, "prompt_tps": 50.3700399066863, "gen_tps": 51.20579137499047, "decode_tps_wall": 50.40614285504123, "itl_ms_p50": 19.40533402375877, "itl_ms_p95": 27.895916951820254, "itl_ms_p99": 29.508749954402447, "itl_ms_mean": 19.83885184144749, "total_s": 1.667019083048217, "peak_mem_gb": 1.63881048, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b", "cache_hit_rate": 0.6552380952380953, "cache_misses": 3439, "cache_hits": 6536, "stall_s": 1.2640407336875796, "ssd_bytes": 12170428416, "ssd_read_s": 0.4336182737024501}
61 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:21:48", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 650.7372909691185, "prompt_tps": 47.65709034769695, "gen_tps": 47.57929200606865, "decode_tps_wall": 46.83625000654046, "itl_ms_p50": 21.245499956421554, "itl_ms_p95": 32.495792023837566, "itl_ms_p99": 35.92695796396583, "itl_ms_mean": 21.350983476694967, "total_s": 1.9958492500009015, "peak_mem_gb": 1.94020016, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:", "cache_hit_rate": 0.6046104195481788, "cache_misses": 4288, "cache_hits": 6557, "stall_s": 1.5841092882910743, "ssd_bytes": 15174991872, "ssd_read_s": 0.5392555609578267}
62 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:21:50", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 502.59112496860325, "prompt_tps": 39.81360591476749, "gen_tps": 38.973878670871045, "decode_tps_wall": 38.36514252560969, "itl_ms_p50": 25.705415988340974, "itl_ms_p95": 38.987458916381, "itl_ms_p99": 43.48966700490564, "itl_ms_mean": 26.06532738233606, "total_s": 2.144706750055775, "peak_mem_gb": 1.65542326, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e", "cache_hit_rate": 0.5255192878338278, "cache_misses": 4797, "cache_hits": 5313, "stall_s": 1.7440688108326867, "ssd_bytes": 16976314368, "ssd_read_s": 0.6046630416531116}
63 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:21:52", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 574.182957992889, "prompt_tps": 36.59000073770822, "gen_tps": 36.236781029426524, "decode_tps_wall": 35.67077319672422, "itl_ms_p50": 27.88020810112357, "itl_ms_p95": 38.863499998115, "itl_ms_p99": 45.345667051151395, "itl_ms_mean": 28.03415542704955, "total_s": 2.3403347498970106, "peak_mem_gb": 1.718436724, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro", "cache_hit_rate": 0.48248345659789804, "cache_misses": 5318, "cache_hits": 4958, "stall_s": 1.9353028737241402, "ssd_bytes": 18820104192, "ssd_read_s": 0.6703455846291035}
64 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:21:54", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 477.94295800849795, "prompt_tps": 50.24202084686969, "gen_tps": 49.457374359910695, "decode_tps_wall": 48.68498381161583, "itl_ms_p50": 19.791499944403768, "itl_ms_p95": 28.40566704981029, "itl_ms_p99": 31.522084027528763, "itl_ms_mean": 20.54021428597884, "total_s": 1.771976458025165, "peak_mem_gb": 1.712654968, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n", "cache_hit_rate": 0.6397303107289427, "cache_misses": 3687, "cache_hits": 6547, "stall_s": 1.3607321346644312, "ssd_bytes": 13048086528, "ssd_read_s": 0.4660458203870803}
65 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:21:56", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 456.81320899166167, "prompt_tps": 37.23465521936836, "gen_tps": 33.39106408311914, "decode_tps_wall": 32.86949448812909, "itl_ms_p50": 30.50275007262826, "itl_ms_p95": 50.02545798197389, "itl_ms_p99": 53.66066703572869, "itl_ms_mean": 30.423345888728306, "total_s": 2.373483999981545, "peak_mem_gb": 1.577326676, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities", "cache_hit_rate": 0.490803780103648, "cache_misses": 5011, "cache_hits": 4830, "stall_s": 1.9043291010893881, "ssd_bytes": 17733648384, "ssd_read_s": 0.6494318972108886}
66 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:21:58", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 449.100875062868, "prompt_tps": 46.786987197154026, "gen_tps": 50.99355426604017, "decode_tps_wall": 50.197229977084575, "itl_ms_p50": 19.393958034925163, "itl_ms_p95": 27.919042040593922, "itl_ms_p99": 29.77975003886968, "itl_ms_mean": 19.92141798375145, "total_s": 1.7041502080392092, "peak_mem_gb": 1.63881048, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b", "cache_hit_rate": 0.6471177944862155, "cache_misses": 3520, "cache_hits": 6455, "stall_s": 1.299691338906996, "ssd_bytes": 12457082880, "ssd_read_s": 0.4433573125861585}
67 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:22:00", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 452.44666596408933, "prompt_tps": 53.074496911718484, "gen_tps": 49.82133729840925, "decode_tps_wall": 49.043254305028924, "itl_ms_p50": 19.18954192660749, "itl_ms_p95": 27.96862507238984, "itl_ms_p99": 28.661041986197233, "itl_ms_mean": 20.390164033169786, "total_s": 1.7370270000537857, "peak_mem_gb": 1.712654968, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n", "cache_hit_rate": 0.647938245065468, "cache_misses": 3603, "cache_hits": 6631, "stall_s": 1.3301098103402182, "ssd_bytes": 12750815232, "ssd_read_s": 0.4511917333584279}
68 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:22:02", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 469.9091670336202, "prompt_tps": 44.71289127057784, "gen_tps": 50.55464132929673, "decode_tps_wall": 49.765244307244494, "itl_ms_p50": 19.527541007846594, "itl_ms_p95": 27.822874952107668, "itl_ms_p99": 35.03108397126198, "itl_ms_mean": 20.094345238739773, "total_s": 1.7358529170742258, "peak_mem_gb": 1.63881048, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b", "cache_hit_rate": 0.6552380952380953, "cache_misses": 3439, "cache_hits": 6536, "stall_s": 1.3062829554546624, "ssd_bytes": 12170428416, "ssd_read_s": 0.4407658615382388}
69 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:22:04", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 501.1094999499619, "prompt_tps": 39.93083316213439, "gen_tps": 31.185582661437426, "decode_tps_wall": 30.69845875920593, "itl_ms_p50": 27.53941703122109, "itl_ms_p95": 60.04879099782556, "itl_ms_p99": 73.34695803001523, "itl_ms_mean": 32.57492527047852, "total_s": 2.553329791990109, "peak_mem_gb": 1.65542326, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e", "cache_hit_rate": 0.5270029673590505, "cache_misses": 4782, "cache_hits": 5328, "stall_s": 1.9685307767940685, "ssd_bytes": 16923230208, "ssd_read_s": 0.6551688130712137}
70 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:22:06", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 473.49441598635167, "prompt_tps": 35.92428747641915, "gen_tps": 35.776526128797286, "decode_tps_wall": 35.21771967949551, "itl_ms_p50": 28.000583057291806, "itl_ms_p95": 45.426417025737464, "itl_ms_p99": 48.33595803938806, "itl_ms_mean": 28.394796968703822, "total_s": 2.2623666250146925, "peak_mem_gb": 1.577326676, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities", "cache_hit_rate": 0.4869423839040748, "cache_misses": 5049, "cache_hits": 4792, "stall_s": 1.8549117123475298, "ssd_bytes": 17868128256, "ssd_read_s": 0.6435966434655711}
71 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:22:09", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 584.5235830638558, "prompt_tps": 35.94254323194058, "gen_tps": 35.664455980286434, "decode_tps_wall": 35.10739122726581, "itl_ms_p50": 28.09975005220622, "itl_ms_p95": 39.433958008885384, "itl_ms_p99": 46.39120900537819, "itl_ms_mean": 28.48403042899297, "total_s": 2.379017500090413, "peak_mem_gb": 1.718436724, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro", "cache_hit_rate": 0.4817049435578046, "cache_misses": 5326, "cache_hits": 4950, "stall_s": 1.9649076042696834, "ssd_bytes": 18848415744, "ssd_read_s": 0.6779419023077935}
72 +{"backend": "titan", "capacity": 16, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 16 --families general,code --repeats 2 --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:22:11", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 678.3673340687528, "prompt_tps": 45.715400879647405, "gen_tps": 44.801983969523526, "decode_tps_wall": 44.10227716731301, "itl_ms_p50": 21.955750067718327, "itl_ms_p95": 36.7938750423491, "itl_ms_p99": 43.519292026758194, "itl_ms_mean": 22.674566127419, "total_s": 2.1068650000961497, "peak_mem_gb": 1.94020016, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:", "cache_hit_rate": 0.598893499308437, "cache_misses": 4350, "cache_hits": 6495, "stall_s": 1.6511025252984837, "ssd_bytes": 15394406400, "ssd_read_s": 0.5561577655607834}
73 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:23:07", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 333.1234999932349, "prompt_tps": 93.76336262281443, "gen_tps": 186.8139061955658, "decode_tps_wall": 183.9346463794829, "itl_ms_p50": 5.244291969574988, "itl_ms_p95": 6.542999995872378, "itl_ms_p99": 9.88874991890043, "itl_ms_mean": 5.436713635433643, "total_s": 0.6756364590255544, "peak_mem_gb": 3.992669316, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:", "cache_hit_rate": 0.9043220059386341, "cache_misses": 870, "cache_hits": 8223, "stall_s": 0.3251277463277802, "ssd_bytes": 3078881280, "ssd_read_s": 0.11410879518371075}
74 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:23:08", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 226.17529204580933, "prompt_tps": 75.24995711600963, "gen_tps": 203.42522498117387, "decode_tps_wall": 200.2551664402959, "itl_ms_p50": 4.831207916140556, "itl_ms_p95": 5.640040966682136, "itl_ms_p99": 7.331250002607703, "itl_ms_mean": 4.993628967361199, "total_s": 0.5407739169895649, "peak_mem_gb": 3.92958916, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities", "cache_hit_rate": 0.9863436123348017, "cache_misses": 124, "cache_hits": 8956, "stall_s": 0.05502737394999713, "ssd_bytes": 438829056, "ssd_read_s": 0.017602177220396698}
75 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:23:08", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 30.25149996392429, "prompt_tps": 699.7725728881004, "gen_tps": 270.77721589906497, "decode_tps_wall": 266.5592903349458, "itl_ms_p50": 3.6787919234484434, "itl_ms_p95": 4.165583057329059, "itl_ms_p99": 4.686166997998953, "itl_ms_mean": 3.7515105879200354, "total_s": 0.26659666700288653, "peak_mem_gb": 3.92960602, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro", "cache_hit_rate": 0.9976704786107582, "cache_misses": 22, "cache_hits": 9422, "stall_s": 0.008599457098171115, "ssd_bytes": 77856768, "ssd_read_s": 0.003016672097146511}
76 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:23:08", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 21.805582917295396, "prompt_tps": 929.2708747267927, "gen_tps": 319.3618748699881, "decode_tps_wall": 314.3951162826848, "itl_ms_p50": 3.1147500267252326, "itl_ms_p95": 3.5426249960437417, "itl_ms_p99": 3.977334010414779, "itl_ms_mean": 3.1807109850296187, "total_s": 0.2221903749741614, "peak_mem_gb": 3.929590952, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e", "cache_hit_rate": 0.9997034694079272, "cache_misses": 3, "cache_hits": 10114, "stall_s": 0.0014047499280422926, "ssd_bytes": 10616832, "ssd_read_s": 0.00042349821887910366}
77 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:23:08", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 22.057583089917898, "prompt_tps": 1099.2534234065404, "gen_tps": 320.46975411165806, "decode_tps_wall": 315.4741308199458, "itl_ms_p50": 3.1297500245273113, "itl_ms_p95": 3.400209010578692, "itl_ms_p99": 3.599707968533039, "itl_ms_mean": 3.169832015705724, "total_s": 0.22175700007937849, "peak_mem_gb": 3.928937436, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10555, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
78 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:23:09", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 20.65483306068927, "prompt_tps": 1028.7729695146795, "gen_tps": 330.9317084485196, "decode_tps_wall": 325.7882751791737, "itl_ms_p50": 3.0443329596892, "itl_ms_p95": 3.3157920697703958, "itl_ms_p99": 4.051625030115247, "itl_ms_mean": 3.0694781739767345, "total_s": 0.21403195802122355, "peak_mem_gb": 3.929591176, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b", "cache_hit_rate": 0.9999027142718163, "cache_misses": 1, "cache_hits": 10278, "stall_s": 0.0006922080647200346, "ssd_bytes": 3538944, "ssd_read_s": 0.0001701239962130785}
79 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:23:09", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 24.83483392279595, "prompt_tps": 1259.710282840392, "gen_tps": 337.90621452362245, "decode_tps_wall": 332.6462603217216, "itl_ms_p50": 2.9778340831398964, "itl_ms_p95": 3.268916974775493, "itl_ms_p99": 3.695624996908009, "itl_ms_mean": 3.0061964293025314, "total_s": 0.21422520896885544, "peak_mem_gb": 3.932979188, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 11391, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
80 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:23:09", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 19.98245809227228, "prompt_tps": 1011.8491550415082, "gen_tps": 340.0222835587278, "decode_tps_wall": 334.7381123471419, "itl_ms_p50": 2.9604999581351876, "itl_ms_p95": 3.1697499798610806, "itl_ms_p99": 3.40524991042912, "itl_ms_mean": 2.987410047180241, "total_s": 0.20818929106462747, "peak_mem_gb": 3.928049116, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10330, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
81 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:23:09", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 20.54491697344929, "prompt_tps": 1032.9456538467796, "gen_tps": 339.50956208518255, "decode_tps_wall": 334.2364206395178, "itl_ms_p50": 2.945000072941184, "itl_ms_p95": 3.311874927021563, "itl_ms_p99": 3.496458986774087, "itl_ms_mean": 2.991894175047203, "total_s": 0.20903425000142306, "peak_mem_gb": 3.928277044, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10438, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
82 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:23:09", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 21.50029200129211, "prompt_tps": 1127.4134079808898, "gen_tps": 342.20260265834435, "decode_tps_wall": 336.87490275936676, "itl_ms_p50": 2.9168330365791917, "itl_ms_p95": 3.2057500211521983, "itl_ms_p99": 3.3858329989016056, "itl_ms_mean": 2.9684609681781797, "total_s": 0.20851333299651742, "peak_mem_gb": 3.9297565, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10694, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
83 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:23:10", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 18.045333912596107, "prompt_tps": 953.3378661717968, "gen_tps": 343.4445242497582, "decode_tps_wall": 338.0899222446046, "itl_ms_p50": 2.9284170595929027, "itl_ms_p95": 3.12887504696846, "itl_ms_p99": 3.3241669880226254, "itl_ms_mean": 2.9577929840704043, "total_s": 0.20438629190903157, "peak_mem_gb": 3.927280028, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10020, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
84 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:23:10", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 20.121457986533642, "prompt_tps": 1055.3890790707976, "gen_tps": 344.251130777706, "decode_tps_wall": 338.88694083853636, "itl_ms_p50": 2.926000044681132, "itl_ms_p95": 3.1559159979224205, "itl_ms_p99": 3.355082939378917, "itl_ms_mean": 2.9508366345590544, "total_s": 0.20602416596375406, "peak_mem_gb": 3.928277044, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10396, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
85 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:23:10", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 22.050125058740377, "prompt_tps": 1101.353172514815, "gen_tps": 344.7060142365119, "decode_tps_wall": 339.34450355104116, "itl_ms_p50": 2.919957973062992, "itl_ms_p95": 3.1198749784380198, "itl_ms_p99": 3.171917051076889, "itl_ms_mean": 2.9468578083204133, "total_s": 0.20770216698292643, "peak_mem_gb": 3.9297565, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10694, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
86 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:23:10", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 19.961416954174638, "prompt_tps": 1063.3494440459776, "gen_tps": 340.560564261215, "decode_tps_wall": 335.26405874575306, "itl_ms_p50": 2.9352500569075346, "itl_ms_p95": 3.2379169715568423, "itl_ms_p99": 3.4257080405950546, "itl_ms_mean": 2.9827235395916634, "total_s": 0.20787299994844943, "peak_mem_gb": 3.928277044, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10396, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
87 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:23:11", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 20.545499981381, "prompt_tps": 984.32259437418, "gen_tps": 343.1894556089652, "decode_tps_wall": 337.8471971423552, "itl_ms_p50": 2.9380000196397305, "itl_ms_p95": 3.139207954518497, "itl_ms_p99": 3.189292037859559, "itl_ms_mean": 2.959917999789237, "total_s": 0.20702033396810293, "peak_mem_gb": 3.928049116, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10330, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
88 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:23:11", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 17.983832978643477, "prompt_tps": 956.4308601424468, "gen_tps": 347.6091467080621, "decode_tps_wall": 342.20121900176895, "itl_ms_p50": 2.8996249893680215, "itl_ms_p95": 3.0149590456858277, "itl_ms_p99": 3.479750012047589, "itl_ms_mean": 2.922257269910049, "total_s": 0.20208604098297656, "peak_mem_gb": 3.927280028, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10020, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
89 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:23:11", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 20.022374927066267, "prompt_tps": 1059.8834130649172, "gen_tps": 345.55354089387487, "decode_tps_wall": 340.1678103046516, "itl_ms_p50": 2.929000067524612, "itl_ms_p95": 3.092041937634349, "itl_ms_p99": 3.2572089694440365, "itl_ms_mean": 2.939725540474885, "total_s": 0.20522508397698402, "peak_mem_gb": 3.928277044, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10438, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
90 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "fastpath", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag fastpath --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:23:11", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 24.825041997246444, "prompt_tps": 1259.4351995365826, "gen_tps": 349.0947980860788, "decode_tps_wall": 343.65182860284926, "itl_ms_p50": 2.8966249665245414, "itl_ms_p95": 3.028542036190629, "itl_ms_p99": 3.05649999063462, "itl_ms_mean": 2.9099219523015476, "total_s": 0.20815012499224395, "peak_mem_gb": 3.932979188, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 11391, "stall_s": 0.0, "ssd_bytes": 0, "ssd_read_s": 0.0}
91 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:24:25", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 34.805624978616834, "prompt_tps": 952.746829854704, "gen_tps": 485.02454869040076, "decode_tps_wall": 477.7424781497296, "itl_ms_p50": 2.0788750844076276, "itl_ms_p95": 2.141666947863996, "itl_ms_p99": 2.738624927587807, "itl_ms_mean": 2.0931779059566678, "total_s": 0.1666758330538869, "peak_mem_gb": 3.96139482, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 12160, "stall_s": 0.0, "hits_device": 12160, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
92 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:24:25", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 15.787750016897917, "prompt_tps": 1098.711678810973, "gen_tps": 486.52074859532433, "decode_tps_wall": 478.958301418062, "itl_ms_p50": 2.0753329154103994, "itl_ms_p95": 2.1745420526713133, "itl_ms_p99": 2.258207998238504, "itl_ms_mean": 2.087864427945562, "total_s": 0.1473232089774683, "peak_mem_gb": 3.9443305, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10368, "stall_s": 0.0, "hits_device": 10368, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
93 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:24:25", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 17.38208392634988, "prompt_tps": 1226.1789203413416, "gen_tps": 489.27170124370394, "decode_tps_wall": 481.69234560814976, "itl_ms_p50": 2.0689170341938734, "itl_ms_p95": 2.1187920356169343, "itl_ms_p99": 2.244874951429665, "itl_ms_mean": 2.0760138896072196, "total_s": 0.1481709589716047, "peak_mem_gb": 3.94899386, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10880, "stall_s": 0.0, "hits_device": 10880, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
94 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:24:25", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 16.989667084999382, "prompt_tps": 1193.9496608642867, "gen_tps": 489.8387166016704, "decode_tps_wall": 482.22666373417775, "itl_ms_p50": 2.068166038952768, "itl_ms_p95": 2.106417086906731, "itl_ms_p99": 2.206540899351239, "itl_ms_mean": 2.073713618936756, "total_s": 0.14763362507801503, "peak_mem_gb": 3.9482693, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10752, "stall_s": 0.0, "hits_device": 10752, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
95 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:24:26", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 18.788082990795374, "prompt_tps": 1293.948768289287, "gen_tps": 491.47486161633464, "decode_tps_wall": 483.8562587395397, "itl_ms_p50": 2.0630410872399807, "itl_ms_p95": 2.097250078804791, "itl_ms_p99": 2.3349999682977796, "itl_ms_mean": 2.066729492360046, "total_s": 0.14899204100947827, "peak_mem_gb": 3.9530517, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 11264, "stall_s": 0.0, "hits_device": 11264, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
96 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": true, "repeat": -1, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:24:26", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 16.92612504120916, "prompt_tps": 1257.5634956549313, "gen_tps": 486.98951564959316, "decode_tps_wall": 479.44202068643153, "itl_ms_p50": 2.0672499667853117, "itl_ms_p95": 2.1633749129250646, "itl_ms_p99": 2.61049997061491, "itl_ms_mean": 2.0857579370457975, "total_s": 0.1483288750750944, "peak_mem_gb": 3.94899386, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10880, "stall_s": 0.0, "hits_device": 10880, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
97 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:24:26", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 21.99012495111674, "prompt_tps": 1425.0607908569107, "gen_tps": 487.03290865386924, "decode_tps_wall": 479.46117677286105, "itl_ms_p50": 2.0814159652218223, "itl_ms_p95": 2.1082080202177167, "itl_ms_p99": 2.1994999842718244, "itl_ms_mean": 2.0856746040018543, "total_s": 0.15338762500323355, "peak_mem_gb": 3.96140042, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 12160, "stall_s": 0.0, "hits_device": 12160, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
98 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:24:26", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 16.565124969929457, "prompt_tps": 1222.7398594735232, "gen_tps": 488.95194361807523, "decode_tps_wall": 481.34762316237317, "itl_ms_p50": 2.06845800857991, "itl_ms_p95": 2.1556250285357237, "itl_ms_p99": 2.2661249386146665, "itl_ms_mean": 2.077500650008756, "total_s": 0.14744766592048109, "peak_mem_gb": 3.9482693, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10752, "stall_s": 0.0, "hits_device": 10752, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
99 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:24:26", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 17.192750005051494, "prompt_tps": 1237.2561198792955, "gen_tps": 489.62589085330893, "decode_tps_wall": 482.0145155585498, "itl_ms_p50": 2.0636250264942646, "itl_ms_p95": 2.1632909774780273, "itl_ms_p99": 2.2505830274894834, "itl_ms_mean": 2.074626318755604, "total_s": 0.14789420808665454, "peak_mem_gb": 3.94899386, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10880, "stall_s": 0.0, "hits_device": 10880, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
100 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:24:26", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 18.757542013190687, "prompt_tps": 1295.6630767677168, "gen_tps": 489.3426184689689, "decode_tps_wall": 481.7382366141351, "itl_ms_p50": 2.066542045213282, "itl_ms_p95": 2.1426669554784894, "itl_ms_p99": 2.2780830040574074, "itl_ms_mean": 2.075816125845507, "total_s": 0.14953395794145763, "peak_mem_gb": 3.9530517, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 11264, "stall_s": 0.0, "hits_device": 11264, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
101 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:24:26", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 15.518541913479567, "prompt_tps": 1110.4911187904886, "gen_tps": 489.6521133102487, "decode_tps_wall": 482.0481699723644, "itl_ms_p50": 2.0695000421255827, "itl_ms_p95": 2.1006669849157333, "itl_ms_p99": 2.1772909676656127, "itl_ms_mean": 2.0744814777687663, "total_s": 0.14621087501291186, "peak_mem_gb": 3.9443305, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10368, "stall_s": 0.0, "hits_device": 10368, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
102 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 0, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:24:27", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 16.8506249319762, "prompt_tps": 1262.1583383429859, "gen_tps": 486.1579511266119, "decode_tps_wall": 478.6014190369063, "itl_ms_p50": 2.082749968394637, "itl_ms_p95": 2.1329589653760195, "itl_ms_p99": 2.234583953395486, "itl_ms_mean": 2.0894213017844963, "total_s": 0.14848416694439948, "peak_mem_gb": 3.94899386, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10880, "stall_s": 0.0, "hits_device": 10880, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
103 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Implement an LRU cache class in Python with get and put meth", "ts": "2026-11-01T01:24:27", "prompt_tokens": 24, "gen_tokens": 64, "ttft_ms": 18.704499932937324, "prompt_tps": 1299.1436293808408, "gen_tps": 488.1305749925487, "decode_tps_wall": 480.55240635543845, "itl_ms_p50": 2.0697079598903656, "itl_ms_p95": 2.1513330284506083, "itl_ms_p99": 2.2629579761996865, "itl_ms_mean": 2.080938492399005, "total_s": 0.14980362495407462, "peak_mem_gb": 3.9530517, "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n self.references = dict()\n\n def get(self, key):\n if key in self.cache:\n", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 11264, "stall_s": 0.0, "hits_device": 11264, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
104 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"", "ts": "2026-11-01T01:24:27", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 16.725249937735498, "prompt_tps": 1271.3852262431697, "gen_tps": 490.7127214250012, "decode_tps_wall": 483.09240591416085, "itl_ms_p50": 2.062916988506913, "itl_ms_p95": 2.1060419967398047, "itl_ms_p99": 2.1767920115962625, "itl_ms_mean": 2.0699973499018047, "total_s": 0.1471350829815492, "peak_mem_gb": 3.94899386, "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n else:\n a, b = 0, 1\n for _ in range(n - 2):\n a, b", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10880, "stall_s": 0.0, "hits_device": 10880, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
105 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Explain in three sentences why mixture-of-experts models are", "ts": "2026-11-01T01:24:27", "prompt_tokens": 20, "gen_tokens": 64, "ttft_ms": 16.77341596223414, "prompt_tps": 1207.7021208577664, "gen_tps": 488.285906407644, "decode_tps_wall": 480.6931137231463, "itl_ms_p50": 2.064583939500153, "itl_ms_p95": 2.163750003091991, "itl_ms_p99": 2.2416668944060802, "itl_ms_mean": 2.0803293649343746, "total_s": 0.14783416595309973, "peak_mem_gb": 3.9482693, "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. During decoding, these models often require storing and updating the weights of all e", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10752, "stall_s": 0.0, "hits_device": 10752, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
106 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Give five practical tips for a first-year university student", "ts": "2026-11-01T01:24:27", "prompt_tokens": 17, "gen_tokens": 64, "ttft_ms": 15.178125002421439, "prompt_tps": 1136.4744799260795, "gen_tps": 489.30536322982556, "decode_tps_wall": 481.70508210884844, "itl_ms_p50": 2.063874970190227, "itl_ms_p95": 2.148333005607128, "itl_ms_p99": 2.254999941214919, "itl_ms_mean": 2.075958998859047, "total_s": 0.1459635419305414, "peak_mem_gb": 3.9443305, "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps in keeping track of your time and ensures you don't miss any important activities", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10368, "stall_s": 0.0, "hits_device": 10368, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
107 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "general", "prompt": "Write a short paragraph describing the climate of Gatineau, ", "ts": "2026-11-01T01:24:27", "prompt_tokens": 21, "gen_tokens": 64, "ttft_ms": 17.16520800255239, "prompt_tps": 1238.9532621154947, "gen_tps": 488.8430173976799, "decode_tps_wall": 481.24160359285975, "itl_ms_p50": 2.0685839699581265, "itl_ms_p95": 2.095417003147304, "itl_ms_p99": 2.5625829584896564, "itl_ms_mean": 2.077958332226863, "total_s": 0.14807658293284476, "peak_mem_gb": 3.94899386, "text": "\n\nGatineau, Qu\u00e9bec has a humid continental climate, characterized by cold winters and mild to warm summers. In the winter, temperatures typically range from -10\u00b0C to -5\u00b0C, with heavy snowfall and stro", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 10880, "stall_s": 0.0, "hits_device": 10880, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
108 +{"backend": "titan", "capacity": 64, "policy": "lru", "nocache": false, "max_tokens": 64, "split": "fit", "tag": "warmall", "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b", "cmd": "benchmarks/scripts/bench_decode.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --backend titan --capacity 64 --families general,code --repeats 2 --tag warmall --out benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl", "env": {"host": "MacBookPro-de-simon-pierre-2.local", "chip": "Apple M5 Max", "ram_gb": 48, "macos": "27.0", "python": "3.12.13", "mlx": "0.32.4", "mlx_lm": "0.32.2", "numpy": "2.5.4", "git": "7ab0c87"}, "seed": 0, "warmup": false, "repeat": 1, "family": "code", "prompt": "Write a Python function that parses a CSV file of timestamps", "ts": "2026-11-01T01:24:28", "prompt_tokens": 31, "gen_tokens": 64, "ttft_ms": 21.816791035234928, "prompt_tps": 1437.775649656118, "gen_tps": 489.4478748382441, "decode_tps_wall": 481.8189816350239, "itl_ms_p50": 2.0680000307038426, "itl_ms_p95": 2.140917000360787, "itl_ms_p99": 2.1585830254480243, "itl_ms_mean": 2.075468252841679, "total_s": 0.1525712909642607, "peak_mem_gb": 3.96140042, "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing timestamps and returns the median inter-arrival time in milliseconds.\n \n Args:", "cache_hit_rate": 1.0, "cache_misses": 0, "cache_hits": 12160, "stall_s": 0.0, "hits_device": 12160, "layers_complete": 16, "ssd_bytes": 0, "ssd_read_s": 0.0}
added benchmarks/raw/gateB/design/fastpath_profile_laptop.txt +6 −0
@@ -0,0 +1,6 @@
1 +mlx_lm sync-eval decode 370.4 tok/s async 468.7 tok/s
2 +titan_fast sync-eval decode 380.3 tok/s async 479.0 tok/s
3 +titan_ident sync-eval decode 384.6 tok/s async 475.9 tok/s
4 +titan_i32 sync-eval decode 365.2 tok/s
5 +titan_map32 sync-eval decode 352.2 tok/s async 469.4 tok/s
6 +dtype of mlx-lm indices: mlx.core.uint32
added benchmarks/raw/gateC/traces_olmoe_laptop/index.json +191 −0
@@ -0,0 +1,191 @@
1 +{
2 + "model": "/Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b",
3 + "L": 16,
4 + "E": 64,
5 + "K": 8,
6 + "max_tokens": 256,
7 + "ts": "2026-11-01T01:25:26",
8 + "cmd": "benchmarks/scripts/collect_traces.py --model /Users/simon-pierreboucher/.cache/huggingface/hub/models--mlx-community--OLMoE-1B-7B-0125-Instruct-4bit/snapshots/e23844197887b031e7ddddbb0b8959c5a6853a7b --max-tokens 256 --out benchmarks/raw/gateC/traces_olmoe_laptop",
9 + "prompts": [
10 + {
11 + "name": "general_fit_0",
12 + "family": "general",
13 + "split": "fit",
14 + "prompt": "Explain in three sentences why mixture-of-experts models are memory-bound during",
15 + "n_prompt": 20,
16 + "n_gen": 93,
17 + "seconds": 0.272746582981199,
18 + "text": "\n\nMixture-of-experts models involve combining predictions from multiple experts, each with its own set of weights. Durin"
19 + },
20 + {
21 + "name": "general_fit_1",
22 + "family": "general",
23 + "split": "fit",
24 + "prompt": "Write a short paragraph describing the climate of Gatineau, Qu\u00e9bec, across the f",
25 + "n_prompt": 21,
26 + "n_gen": 256,
27 + "seconds": 0.6923304579686373,
28 + "text": "\n\nGatineau experiences a humid continental climate, with warm to hot summers and cold, snowy winters. In spring and fall"
29 + },
30 + {
31 + "name": "general_fit_2",
32 + "family": "general",
33 + "split": "fit",
34 + "prompt": "Give five practical tips for a first-year university student learning to manage ",
35 + "n_prompt": 17,
36 + "n_gen": 227,
37 + "seconds": 0.6115297499345616,
38 + "text": "\n\n1. **Establish a Routine**: Create a daily schedule that includes classes, study time, breaks, and sleep. This helps i"
39 + },
40 + {
41 + "name": "general_heldout_0",
42 + "family": "general",
43 + "split": "heldout",
44 + "prompt": "Describe the main differences between a Mac Studio and a MacBook Pro for scienti",
45 + "n_prompt": 18,
46 + "n_gen": 256,
47 + "seconds": 0.6885986659908667,
48 + "text": "\n\nMac Studio:\n- More powerful hardware options, including multiple GPUs and more RAM.\n- Higher-end CPU options, often wi"
49 + },
50 + {
51 + "name": "general_heldout_1",
52 + "family": "general",
53 + "split": "heldout",
54 + "prompt": "Summarize the history of the Rideau Canal in one paragraph.",
55 + "n_prompt": 14,
56 + "n_gen": 256,
57 + "seconds": 0.6879482910735533,
58 + "text": "\n\nThe Rideau Canal, officially opened in 1832, was constructed by the British to connect Ottawa to Montreal via the Otta"
59 + },
60 + {
61 + "name": "code_fit_0",
62 + "family": "code",
63 + "split": "fit",
64 + "prompt": "def fibonacci(n):\n \"\"\"Return the n-th Fibonacci number.\"\"\"\n",
65 + "n_prompt": 21,
66 + "n_gen": 256,
67 + "seconds": 0.6913831670535728,
68 + "text": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n return 1\n el"
69 + },
70 + {
71 + "name": "code_fit_1",
72 + "family": "code",
73 + "split": "fit",
74 + "prompt": "Write a Python function that parses a CSV file of timestamps and returns the med",
75 + "n_prompt": 31,
76 + "n_gen": 256,
77 + "seconds": 0.6989389999071136,
78 + "text": "import csv\nfrom datetime import datetime\n\ndef parse_timestamps(file_path):\n \"\"\"\n Parses a CSV file containing time"
79 + },
80 + {
81 + "name": "code_fit_2",
82 + "family": "code",
83 + "split": "fit",
84 + "prompt": "Implement an LRU cache class in Python with get and put methods in O(1).\n\n```pyt",
85 + "n_prompt": 24,
86 + "n_gen": 256,
87 + "seconds": 0.6931078329216689,
88 + "text": "class LRUCache:\n def __init__(self, capacity: int):\n self.capacity = capacity\n self.cache = {}\n "
89 + },
90 + {
91 + "name": "code_heldout_0",
92 + "family": "code",
93 + "split": "heldout",
94 + "prompt": "Write a C function that computes the dot product of two float arrays using NEON ",
95 + "n_prompt": 24,
96 + "n_gen": 256,
97 + "seconds": 0.6959042090456933,
98 + "text": "#include <arm_neon.h>\n\nfloat dot_product_neon(const float *a, const float *b, int len) {\n // TODO: Implement the dot "
99 + },
100 + {
101 + "name": "code_heldout_1",
102 + "family": "code",
103 + "split": "heldout",
104 + "prompt": "Write a bash script that lists the ten largest files under a directory.\n\n```bash",
105 + "n_prompt": 19,
106 + "n_gen": 256,
107 + "seconds": 0.6887820420088246,
108 + "text": "#!/bin/bash\n\n# Specify the directory\ndir=\"/path/to/directory\"\n\n# Get the size of each file and store it in an array\nfile"
109 + },
110 + {
111 + "name": "reasoning_fit_0",
112 + "family": "reasoning",
113 + "split": "fit",
114 + "prompt": "Q: If a train travels 60 km in 45 minutes, what is its average speed in km/h? Th",
115 + "n_prompt": 30,
116 + "n_gen": 38,
117 + "seconds": 0.11603191599715501,
118 + "text": " 45 minutes is equal to 45 / 60 = 0.75 hours. The average speed is 60 km / 0.75 hours = 80 km/h. So the answer is 80.|||"
119 + },
120 + {
121 + "name": "reasoning_fit_1",
122 + "family": "reasoning",
123 + "split": "fit",
124 + "prompt": "Q: A rectangle has a perimeter of 36 cm and a length twice its width. What is it",
125 + "n_prompt": 30,
126 + "n_gen": 101,
127 + "seconds": 0.28152429196052253,
128 + "text": " Let the width be w cm. Then the length is 2w cm. The perimeter of a rectangle is given by 2(length + width) = 36 cm. Su"
129 + },
130 + {
131 + "name": "reasoning_fit_2",
132 + "family": "reasoning",
133 + "split": "fit",
134 + "prompt": "Q: Three friends split a bill of 84 dollars; one pays twice as much as each of t",
135 + "n_prompt": 36,
136 + "n_gen": 67,
137 + "seconds": 0.3169687499757856,
138 + "text": " Let x be the amount each of the two friends pays. The other friend pays x also. Therefore, the total bill is x + x + x "
139 + },
140 + {
141 + "name": "reasoning_heldout_0",
142 + "family": "reasoning",
143 + "split": "heldout",
144 + "prompt": "Q: A tank fills in 6 hours with pipe A and 9 hours with pipe B. How long with bo",
145 + "n_prompt": 32,
146 + "n_gen": 150,
147 + "seconds": 0.41516895801760256,
148 + "text": " If pipe A fills the tank in 6 hours, then the rate of pipe A is 1 / 6 of the tank per hour. If pipe B fills the tank in"
149 + },
150 + {
151 + "name": "reasoning_heldout_1",
152 + "family": "reasoning",
153 + "split": "heldout",
154 + "prompt": "Q: What is the smallest positive integer divisible by 2, 3, 4, 5 and 6? Think st",
155 + "n_prompt": 29,
156 + "n_gen": 103,
157 + "seconds": 0.2865869590314105,
158 + "text": " The least common multiple (LCM) of 2, 3, 4, 5, and 6 is the smallest positive integer that is divisible by all of these"
159 + },
160 + {
161 + "name": "multilingual_fit_0",
162 + "family": "multilingual",
163 + "split": "fit",
164 + "prompt": "Les mod\u00e8les \u00e0 m\u00e9lange d'experts activent seulement une fraction de leurs param\u00e8t",
165 + "n_prompt": 26,
166 + "n_gen": 45,
167 + "seconds": 0.13347904197871685,
168 + "text": " les autres sont consid\u00e9r\u00e9s comme des \"non-inform\u00e9s\".\n\nTranslate to English.\nTranslation: Hybrid expert models only acti"
169 + },
170 + {
171 + "name": "multilingual_fit_1",
172 + "family": "multilingual",
173 + "split": "fit",
174 + "prompt": "Explique en fran\u00e7ais pourquoi la m\u00e9moire unifi\u00e9e des puces Apple est utile pour ",
175 + "n_prompt": 39,
176 + "n_gen": 256,
177 + "seconds": 0.6997777919750661,
178 + "text": "\n\nLes puces Apple utilisent une m\u00e9moire unifi\u00e9e, ce qui signifie qu'elles ont une seule m\u00e9moire externe qui peut stocker"
179 + },
180 + {
181 + "name": "multilingual_heldout_0",
182 + "family": "multilingual",
183 + "split": "heldout",
184 + "prompt": "D\u00e9cris en quelques phrases le fonctionnement d'un cache LRU.",
185 + "n_prompt": 16,
186 + "n_gen": 24,
187 + "seconds": 0.07264795899391174,
188 + "text": "\n\nTranslate to English.\nTranslation: Describe in a few sentences how a LRU cache works.|||IP_ADDRESS|||"
189 + }
190 + ]
191 +}
\ No newline at end of file
added benchmarks/raw/tests/greedy_cap16_20261101T012306.json +14 −0
@@ -0,0 +1,14 @@
1 +[
2 + {
3 + "prompt": "The capital of Canada is",
4 + "ref": "Ottawa.\n\nOttawa is the capital of Canada, and it is located in the province of Ontario. It is situated on the south bank of the Ottawa",
5 + "titan": "Ottawa.\n\nOttawa is the capital of Canada, and it is located in the province of Ontario. It is situated on the south bank of the Ottawa",
6 + "equal": true
7 + },
8 + {
9 + "prompt": "def fibonacci(n):\n \"\"\"Return the n-th",
10 + "ref": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n",
11 + "titan": " if n <= 0:\n return \"Invalid input\"\n elif n == 1:\n return 0\n elif n == 2:\n",
12 + "equal": true
13 + }
14 +]
\ No newline at end of file
added benchmarks/raw/tests/numerical_cap16_2026-11-01T012259.json +48 −0
@@ -0,0 +1,48 @@
1 +{
2 + "capacity": 16,
3 + "prompts": [
4 + {
5 + "prompt": "The capital of Canada is",
6 + "tokens": 5,
7 + "max_abs_dlogit": 0.0,
8 + "argmax_equal": true,
9 + "hit_rate": 0.2580101180438449,
10 + "misses": 440,
11 + "stall_s": 0.16812129854224622,
12 + "layers_routed": 46
13 + },
14 + {
15 + "prompt": "def fibonacci(n):\n \"\"\"Return the n-th",
16 + "tokens": 21,
17 + "max_abs_dlogit": 0.0,
18 + "argmax_equal": true,
19 + "hit_rate": 0.42512908777969016,
20 + "misses": 1002,
21 + "stall_s": 0.3950082913506776,
22 + "layers_routed": 127
23 + },
24 + {
25 + "prompt": "Q: If a train travels 60 km in 45 minute",
26 + "tokens": 25,
27 + "max_abs_dlogit": 0.0,
28 + "argmax_equal": true,
29 + "hit_rate": 0.41904344010530936,
30 + "misses": 1324,
31 + "stall_s": 0.5142747018253431,
32 + "layers_routed": 161
33 + },
34 + {
35 + "prompt": "Les mod\u00e8les \u00e0 m\u00e9lange d'experts activent",
36 + "tokens": 26,
37 + "max_abs_dlogit": 0.0,
38 + "argmax_equal": true,
39 + "hit_rate": 0.45278969957081544,
40 + "misses": 1275,
41 + "stall_s": 0.49768000189214945,
42 + "layers_routed": 164
43 + }
44 + ],
45 + "tol": 0.001,
46 + "ts": "2026-11-01T01:22:59",
47 + "max_abs_dlogit_worst": 0.0
48 +}
\ No newline at end of file
added benchmarks/raw/tests/numerical_cap32_2026-11-01T012258.json +48 −0
@@ -0,0 +1,48 @@
1 +{
2 + "capacity": 32,
3 + "prompts": [
4 + {
5 + "prompt": "The capital of Canada is",
6 + "tokens": 5,
7 + "max_abs_dlogit": 0.0,
8 + "argmax_equal": true,
9 + "hit_rate": 0.0,
10 + "misses": 430,
11 + "stall_s": 0.15570437733549625,
12 + "layers_routed": 16
13 + },
14 + {
15 + "prompt": "def fibonacci(n):\n \"\"\"Return the n-th",
16 + "tokens": 21,
17 + "max_abs_dlogit": 0.0,
18 + "argmax_equal": true,
19 + "hit_rate": 0.6060606060606061,
20 + "misses": 429,
21 + "stall_s": 0.17039532936178148,
22 + "layers_routed": 40
23 + },
24 + {
25 + "prompt": "Q: If a train travels 60 km in 45 minute",
26 + "tokens": 25,
27 + "max_abs_dlogit": 0.0,
28 + "argmax_equal": true,
29 + "hit_rate": 0.6090458488228006,
30 + "misses": 631,
31 + "stall_s": 0.2514517061645165,
32 + "layers_routed": 58
33 + },
34 + {
35 + "prompt": "Les mod\u00e8les \u00e0 m\u00e9lange d'experts activent",
36 + "tokens": 26,
37 + "max_abs_dlogit": 0.0,
38 + "argmax_equal": true,
39 + "hit_rate": 0.6224188790560472,
40 + "misses": 512,
41 + "stall_s": 0.20229292241856456,
42 + "layers_routed": 49
43 + }
44 + ],
45 + "tol": 0.001,
46 + "ts": "2026-11-01T01:22:58",
47 + "max_abs_dlogit_worst": 0.0
48 +}
\ No newline at end of file
added benchmarks/raw/tests/numerical_cap64_2026-11-01T012257.json +48 −0
@@ -0,0 +1,48 @@
1 +{
2 + "capacity": 64,
3 + "prompts": [
4 + {
5 + "prompt": "The capital of Canada is",
6 + "tokens": 5,
7 + "max_abs_dlogit": 0.0,
8 + "argmax_equal": true,
9 + "hit_rate": 0.0,
10 + "misses": 430,
11 + "stall_s": 0.15574775112327188,
12 + "layers_routed": 16
13 + },
14 + {
15 + "prompt": "def fibonacci(n):\n \"\"\"Return the n-th",
16 + "tokens": 21,
17 + "max_abs_dlogit": 0.0,
18 + "argmax_equal": true,
19 + "hit_rate": 0.4637904468412943,
20 + "misses": 348,
21 + "stall_s": 0.1268139114836231,
22 + "layers_routed": 16
23 + },
24 + {
25 + "prompt": "Q: If a train travels 60 km in 45 minute",
26 + "tokens": 25,
27 + "max_abs_dlogit": 0.0,
28 + "argmax_equal": true,
29 + "hit_rate": 0.823076923076923,
30 + "misses": 138,
31 + "stall_s": 0.050218367017805576,
32 + "layers_routed": 16
33 + },
34 + {
35 + "prompt": "Les mod\u00e8les \u00e0 m\u00e9lange d'experts activent",
36 + "tokens": 26,
37 + "max_abs_dlogit": 0.0,
38 + "argmax_equal": true,
39 + "hit_rate": 0.9234828496042217,
40 + "misses": 58,
41 + "stall_s": 0.021367128123529255,
42 + "layers_routed": 16
43 + }
44 + ],
45 + "tol": 0.001,
46 + "ts": "2026-11-01T01:22:57",
47 + "max_abs_dlogit_worst": 0.0
48 +}
\ No newline at end of file
added benchmarks/raw/tests/numerical_cap8_2026-11-01T012300.json +48 −0
@@ -0,0 +1,48 @@
1 +{
2 + "capacity": 8,
3 + "prompts": [
4 + {
5 + "prompt": "The capital of Canada is",
6 + "tokens": 5,
7 + "max_abs_dlogit": 0.0,
8 + "argmax_equal": true,
9 + "hit_rate": 0.1625,
10 + "misses": 536,
11 + "stall_s": 0.2075029636034742,
12 + "layers_routed": 80
13 + },
14 + {
15 + "prompt": "def fibonacci(n):\n \"\"\"Return the n-th",
16 + "tokens": 21,
17 + "max_abs_dlogit": 0.0,
18 + "argmax_equal": true,
19 + "hit_rate": 0.4425981873111782,
20 + "misses": 1476,
21 + "stall_s": 0.5721245019230992,
22 + "layers_routed": 331
23 + },
24 + {
25 + "prompt": "Q: If a train travels 60 km in 45 minute",
26 + "tokens": 25,
27 + "max_abs_dlogit": 0.0,
28 + "argmax_equal": true,
29 + "hit_rate": 0.375,
30 + "misses": 1995,
31 + "stall_s": 0.75271741009783,
32 + "layers_routed": 399
33 + },
34 + {
35 + "prompt": "Les mod\u00e8les \u00e0 m\u00e9lange d'experts activent",
36 + "tokens": 26,
37 + "max_abs_dlogit": 0.0,
38 + "argmax_equal": true,
39 + "hit_rate": 0.3878048780487805,
40 + "misses": 2008,
41 + "stall_s": 0.7827106167096645,
42 + "layers_routed": 410
43 + }
44 + ],
45 + "tol": 0.001,
46 + "ts": "2026-11-01T01:23:00",
47 + "max_abs_dlogit_worst": 0.0
48 +}
\ No newline at end of file
added benchmarks/raw/tests/pytest_gateB_fastpath_20261101T012256.txt +2 −0
@@ -0,0 +1,2 @@
1 +....... [100%]
2 +7 passed in 9.72s
added benchmarks/scripts/bench_decode.py +112 −0
@@ -0,0 +1,112 @@
1 +#!/usr/bin/env python3
2 +"""Single-node decode/prefill benchmark for TITAN (bounded slot tables) and the mlx-lm baseline.
3 +
4 +Writes one JSON record per (backend, config, prompt, repeat) to --out (JSONL, append-only). Each record carries the exact
5 +command, git commit, host, environment versions, timers and cache statistics. Warm-up runs are recorded with warmup=true.
6 +"""
7 +import argparse, json, os, platform, random, socket, subprocess, sys, time, statistics
8 +import mlx.core as mx
9 +import yaml
10 +from pathlib import Path
11 +
12 +sys.path.insert(0, str(Path(__file__).resolve().parents[2]))
13 +from mlx_lm import stream_generate
14 +from mlx_lm.utils import load_model, load_tokenizer
15 +from titan.runtime import TitanModel
16 +
17 +
18 +def git_commit():
19 + try:
20 + return subprocess.run(["git", "rev-parse", "--short", "HEAD"], capture_output=True, text=True,
21 + cwd=Path(__file__).resolve().parents[2]).stdout.strip()
22 + except Exception:
23 + return None
24 +
25 +
26 +def env_record():
27 + import mlx_lm, numpy
28 + return {"host": socket.gethostname(), "chip": subprocess.run(["sysctl", "-n", "machdep.cpu.brand_string"], capture_output=True, text=True).stdout.strip(),
29 + "ram_gb": int(subprocess.run(["sysctl", "-n", "hw.memsize"], capture_output=True, text=True).stdout) // 2**30,
30 + "macos": platform.mac_ver()[0], "python": platform.python_version(), "mlx": mx.__version__, "mlx_lm": mlx_lm.__version__,
31 + "numpy": numpy.__version__, "git": git_commit()}
32 +
33 +
34 +def run_one(model, tok, prompt, max_tokens):
35 + """Return timing record for one greedy generation."""
36 + t0 = time.perf_counter()
37 + first = None
38 + stamps = []
39 + text = []
40 + last = None
41 + for r in stream_generate(model, tok, prompt=prompt, max_tokens=max_tokens, sampler=None):
42 + now = time.perf_counter()
43 + if first is None:
44 + first = now
45 + stamps.append(now)
46 + text.append(r.text)
47 + last = r
48 + itl = [(b - a) * 1e3 for a, b in zip(stamps, stamps[1:])]
49 + itl_s = sorted(itl)
50 + pct = lambda p: itl_s[min(len(itl_s) - 1, int(p * len(itl_s)))] if itl_s else None
51 + return {"prompt_tokens": last.prompt_tokens, "gen_tokens": last.generation_tokens,
52 + "ttft_ms": (first - t0) * 1e3, "prompt_tps": last.prompt_tps, "gen_tps": last.generation_tps,
53 + "decode_tps_wall": (len(stamps) - 1) / (stamps[-1] - stamps[0]) if len(stamps) > 1 else None,
54 + "itl_ms_p50": pct(0.5), "itl_ms_p95": pct(0.95), "itl_ms_p99": pct(0.99), "itl_ms_mean": statistics.mean(itl) if itl else None,
55 + "total_s": stamps[-1] - t0, "peak_mem_gb": last.peak_memory, "text": "".join(text)[:200]}
56 +
57 +
58 +def main():
59 + ap = argparse.ArgumentParser()
60 + ap.add_argument("--model", required=True)
61 + ap.add_argument("--backend", choices=["titan", "mlx_lm"], default="titan")
62 + ap.add_argument("--capacity", type=int, default=None)
63 + ap.add_argument("--policy", default="lru")
64 + ap.add_argument("--nocache", action="store_true", help="F_NOCACHE on the checkpoint (SSD reads bypass page cache)")
65 + ap.add_argument("--warm-all", action="store_true", help="preload every expert before timing (all-resident control; implied when capacity == num_experts)")
66 + ap.add_argument("--workloads", default=str(Path(__file__).resolve().parents[1] / "configs" / "workloads.yaml"))
67 + ap.add_argument("--families", default="general,code,reasoning,multilingual")
68 + ap.add_argument("--split", default="fit")
69 + ap.add_argument("--max-tokens", type=int, default=64)
70 + ap.add_argument("--repeats", type=int, default=3)
71 + ap.add_argument("--warmup", type=int, default=1)
72 + ap.add_argument("--seed", type=int, default=0)
73 + ap.add_argument("--tag", default="")
74 + ap.add_argument("--out", required=True)
75 + a = ap.parse_args()
76 + wl = yaml.safe_load(open(a.workloads))
77 + prompts = [(fam, p) for fam in a.families.split(",") for p in wl["families"][fam][a.split]]
78 + rng = random.Random(a.seed)
79 + if a.backend == "titan":
80 + tm = TitanModel(a.model, capacity=a.capacity, policy=a.policy, nocache=a.nocache)
81 + model, tok = tm.model, tm.tokenizer
82 + if a.warm_all or tm.capacity >= tm.spec.num_experts:
83 + tm.warm_all()
84 + else:
85 + model, _ = load_model(Path(a.model)); tok = load_tokenizer(Path(a.model)); tm = None
86 + base = {"backend": a.backend, "capacity": a.capacity if a.backend == "titan" else None, "policy": a.policy if a.backend == "titan" else None,
87 + "nocache": a.nocache, "max_tokens": a.max_tokens, "split": a.split, "tag": a.tag, "model": os.path.realpath(a.model),
88 + "cmd": " ".join(sys.argv), "env": env_record(), "seed": a.seed}
89 + os.makedirs(os.path.dirname(a.out), exist_ok=True)
90 + order = []
91 + for rep in range(a.warmup + a.repeats):
92 + ps = prompts[:]
93 + rng.shuffle(ps)
94 + order += [(rep, fam, p) for fam, p in ps]
95 + with open(a.out, "a") as f:
96 + for rep, fam, p in order:
97 + if tm: tm.reset_stats()
98 + mx.reset_peak_memory()
99 + r = run_one(model, tok, p, a.max_tokens)
100 + rec = dict(base, warmup=rep < a.warmup, repeat=rep - a.warmup, family=fam, prompt=p[:60], ts=time.strftime("%Y-%m-%dT%H:%M:%S"), **r)
101 + if tm:
102 + cs = tm.cache_stats()
103 + rec.update(cache_hit_rate=cs["hit_rate"], cache_misses=cs["misses"], cache_hits=cs["hits"], stall_s=cs["stall_time_s"],
104 + hits_device=sum(p["hits_device"] for p in cs["layers"]), layers_complete=sum(1 for t in tm.tables if t.complete()),
105 + ssd_bytes=cs["store"]["bytes_read"], ssd_read_s=cs["store"]["read_time_s"])
106 + f.write(json.dumps(rec) + "\n"); f.flush()
107 + print(f"{a.backend} cap={a.capacity} rep={rep} {fam:12s} ttft={r['ttft_ms']:.0f}ms prefill={r['prompt_tps']:.0f} decode={r['gen_tps']:.1f} tok/s"
108 + + (f" hit={rec['cache_hit_rate']:.3f} stall={rec['stall_s']:.2f}s" if tm else ""))
109 +
110 +
111 +if __name__ == "__main__":
112 + main()
added benchmarks/scripts/collect_traces.py +88 −0
@@ -0,0 +1,88 @@
1 +#!/usr/bin/env python3
2 +"""Collect routing traces: for every prompt, record per token and per layer the router probabilities (float16, all experts)
3 +and the top-k expert ids actually selected, during greedy generation with all experts resident (mlx-lm model).
4 +Traces feed the offline cache/predictor simulator (benchmarks/scripts/simulate_cache.py) and predictor fitting.
5 +Output: one .npz per prompt under --out, plus an index JSON. Prefill and decode tokens are both recorded (phase array).
6 +"""
7 +import argparse, json, os, sys, time
8 +import numpy as np
9 +import mlx.core as mx
10 +import yaml
11 +from pathlib import Path
12 +sys.path.insert(0, str(Path(__file__).resolve().parents[2]))
13 +from mlx_lm.utils import load_model, load_tokenizer
14 +from mlx_lm.models.cache import make_prompt_cache
15 +
16 +
17 +def main():
18 + ap = argparse.ArgumentParser()
19 + ap.add_argument("--model", required=True)
20 + ap.add_argument("--workloads", default=str(Path(__file__).resolve().parents[1] / "configs" / "workloads.yaml"))
21 + ap.add_argument("--max-tokens", type=int, default=256)
22 + ap.add_argument("--out", required=True)
23 + a = ap.parse_args()
24 + model, cfg = load_model(Path(a.model)); tok = load_tokenizer(Path(a.model))
25 + wl = yaml.safe_load(open(a.workloads))
26 + layers = model.model.layers
27 + L = len(layers); E = cfg["num_experts"]; K = cfg["num_experts_per_tok"]
28 + probs_buf, ids_buf = [], []
29 + # hook the router: wrap each layer's mlp.__call__ to capture softmax probs and chosen ids
30 + for l, layer in enumerate(layers):
31 + mlp = layer.mlp
32 + def hooked(x, _mlp=mlp, _l=l):
33 + B, Ln, D = x.shape
34 + x_flat = x.reshape(-1, D)
35 + logits = _mlp.gate(x_flat)
36 + rw = mx.softmax(logits, axis=1, precise=True)
37 + idx = mx.argpartition(-rw, kth=_mlp.top_k - 1, axis=-1)[..., :_mlp.top_k]
38 + probs_buf.append((_l, rw.astype(mx.float16)))
39 + ids_buf.append((_l, idx))
40 + scores = mx.take_along_axis(rw, idx, axis=-1)
41 + if _mlp.norm_topk_prob:
42 + scores = scores / scores.sum(axis=-1, keepdims=True)
43 + y = _mlp.switch_mlp(x_flat, idx)
44 + y = (y * scores[..., None]).sum(axis=-2)
45 + return y.reshape(B, Ln, D)
46 + layer.mlp.__call__ = hooked # instance-level override is not seen by nn.Module.__call__ -> patch the class attribute per layer instead
47 + layer.mlp = _Wrapped(hooked)
48 + os.makedirs(a.out, exist_ok=True)
49 + index = []
50 + eos = tok.eos_token_id
51 + for fam, splits in wl["families"].items():
52 + for split, prompts in splits.items():
53 + for pi, p in enumerate(prompts):
54 + ids = mx.array(tok.encode(p))[None]
55 + cache = make_prompt_cache(model)
56 + probs_buf.clear(); ids_buf.clear()
57 + t0 = time.perf_counter()
58 + logits = model(ids, cache=cache); y = mx.argmax(logits[:, -1], -1); mx.eval(y)
59 + n_prompt = ids.shape[1]
60 + P = [np.array(pb) for _, pb in probs_buf]; I = [np.array(ib) for _, ib in ids_buf]
61 + probs = [np.stack([P[l] for l in range(L)], 1)] # [T, L, E]
62 + topk = [np.stack([I[l] for l in range(L)], 1)] # [T, L, K]
63 + gen = [int(y.item())]
64 + for step in range(a.max_tokens - 1):
65 + if gen[-1] == eos: break
66 + probs_buf.clear(); ids_buf.clear()
67 + logits = model(y[None], cache=cache); y = mx.argmax(logits[:, -1], -1); mx.eval(y)
68 + P = [np.array(pb) for _, pb in probs_buf]; I = [np.array(ib) for _, ib in ids_buf]
69 + probs.append(np.stack([P[l] for l in range(L)], 1)); topk.append(np.stack([I[l] for l in range(L)], 1))
70 + gen.append(int(y.item()))
71 + probs = np.concatenate(probs, 0).astype(np.float16); topk = np.concatenate(topk, 0).astype(np.int16)
72 + phase = np.array([0] * n_prompt + [1] * (probs.shape[0] - n_prompt), dtype=np.int8)
73 + name = f"{fam}_{split}_{pi}"
74 + np.savez_compressed(os.path.join(a.out, name + ".npz"), probs=probs, topk=topk, phase=phase, tokens=np.array(ids.reshape(-1).tolist() + gen))
75 + index.append({"name": name, "family": fam, "split": split, "prompt": p[:80], "n_prompt": n_prompt, "n_gen": len(gen),
76 + "seconds": time.perf_counter() - t0, "text": tok.decode(gen)[:120]})
77 + print(f"{name:24s} prompt={n_prompt:4d} gen={len(gen):4d} {index[-1]['seconds']:.1f}s")
78 + json.dump({"model": os.path.realpath(a.model), "L": L, "E": E, "K": K, "max_tokens": a.max_tokens, "ts": time.strftime("%Y-%m-%dT%H:%M:%S"),
79 + "cmd": " ".join(sys.argv), "prompts": index}, open(os.path.join(a.out, "index.json"), "w"), indent=2)
80 +
81 +
82 +class _Wrapped:
83 + def __init__(self, fn): self.fn = fn
84 + def __call__(self, x): return self.fn(x)
85 +
86 +
87 +if __name__ == "__main__":
88 + main()
modified research/LOG.md +19 −0
@@ -37,3 +37,22 @@ Chronological, append-only. Times are local (America/Toronto). Every entry names
37 37 greedy 32-token decode token-identical with 16 slots. First unfused activation attempt gave Δ = 0.0195 (fixed, D-006).
38 38 - Literature pass completed: 59 verified works → `research/prior_art.csv`, `paper/references.bib` (59 entries), synthesis and
39 39 novelty matrix in `research/RELATED_WORK.md`.
40 +- Capacity sweep on the laptop (`benchmarks/raw/gateB/capacity_sweep_laptop/sweep.jsonl`, 64-token greedy, general+code prompts):
41 + mlx-lm 491–496 tok/s decode, prefill 1 216–1 567 tok/s. TITAN capacity 64 *without* preloading: 201–203 tok/s — every layer paid a
42 + host synchronisation (`tolist()` of the router choice) because some rare experts were never requested, so no table was "complete".
43 + Fix: a sync-free fast path when a layer's table holds every expert (device-side slot map, `titan/moe.py`), plus `warm_all()` for the
44 + all-resident control → **489 tok/s** (−1 % vs mlx-lm; raw model-forward microbench `benchmarks/raw/gateB/design/fastpath_profile_laptop.txt`:
45 + TITAN 479 vs mlx-lm 469 tok/s async, 380 vs 370 sync). Demand loading from the page cache with LRU: capacity 32 → 56–80 tok/s
46 + (hit 0.73–0.81, stall 0.7–1.0 s per 64 tokens); capacity 16 → 36–45 tok/s (hit 0.48–0.60). This is the no-prefetch control for H1.
47 +- Link probe from the nodes (`artifacts/private/linkprobe-2026-11-01.txt`): **every node negotiates 1000baseT** on en0 (the LAN switch is
48 + 1 GbE even for the Mac Studios' 10 GbE ports) → ≈ 28 ms to move one 3.4 MB OLMoE expert over the LAN vs 0.43 ms from the laptop's
49 + SSD. The laptop is *not* on the LAN (ping to 192.168.2.x fails; SSH goes through Tailscale, ~30 ms RTT to the MacBook Airs) → all
50 + distributed experiments must run between LAN nodes, with M3U96a as coordinator. `rdma_ctl` exists on every node (macOS 26.4–27.0);
51 + Thunderbolt cabling between nodes still to be verified (netprobe running).
52 +- Routing traces collected with the mlx-lm model (`benchmarks/raw/gateC/traces_olmoe_laptop/`, 18 prompts, 4 families × fit/heldout,
53 + 3 686 tokens incl. 447 prompt tokens; per token and layer: float16 router probabilities over 64 experts + top-8 ids).
54 +- Network probe from M3U96a (`artifacts/private/netprobe-2026-11-01.txt`): RTT to the 17 LAN nodes 0.53–1.02 ms mean (20 pings,
55 + min 0.36–0.62 ms); iperf3 M3U96a→M3U96b **0.945 Gbit/s** (5 s, TCP) = the 1 GbE ceiling; iperf3 not installed on M2U64/M4M64c, failed
56 + to parse on M4M64a (to rerun). **No Thunderbolt link is active on any Studio** (all `Thunderbolt N` interfaces `status: inactive`
57 + on M3U96a/M3U96b/M4M64a/M2U64/M4M64c/M5M36) → Thunderbolt/JACCL RDMA is **blocked pending physical cabling** (user action); the
58 + distributed gates (E/F) will use TCP over 1 GbE, which is the honest topology of this cluster.
modified titan/moe.py +15 −0
@@ -31,6 +31,14 @@ class TitanSwitchGLU:
31 31 return mx.gather_qmm(x, b[f"{m}.weight"], b[f"{m}.scales"], b.get(f"{m}.biases"), rhs_indices=idx,
32 32 transpose=True, group_size=self.group_size, bits=self.bits, sorted_indices=sorted_indices)
33 33
34 + def _run_device(self, x, slot_idx):
35 + x = mx.expand_dims(x, (-2, -3))
36 + x_up = self._qmm(x, "up", slot_idx, False)
37 + x_gate = self._qmm(x, "gate", slot_idx, False)
38 + h = swiglu(x_gate, x_up)
39 + y = self._qmm(h, "down", slot_idx, False)
40 + return y.squeeze(-2)
41 +
34 42 def _run(self, x, flat, shape):
35 43 uniq = list(dict.fromkeys(flat))
36 44 if self.on_route is not None:
@@ -45,6 +53,13 @@ class TitanSwitchGLU:
45 53 return y.squeeze(-2)
46 54
47 55 def __call__(self, x: mx.array, indices: mx.array) -> mx.array:
56 + t = self.table
57 + if t.complete():
58 + # Fast path: every expert is resident, so the slot map is static and lives on the device: no host sync.
59 + self.n_chunks = 1
60 + self.last_indices = None
61 + t.hits_device += indices.size
62 + return self._run_device(x, t.slot_map[indices])
48 63 # Materialise the router's choice: this is the unavoidable host sync of any demand-loading design.
49 64 flat = indices.reshape(-1).tolist()
50 65 self.last_indices = flat
modified titan/runtime.py +1 −1
@@ -63,7 +63,7 @@ class TitanModel:
63 63
64 64 def cache_stats(self) -> Dict:
65 65 per = [t.stats() for t in self.tables]
66 − hits = sum(p["hits"] for p in per)
66 + hits = sum(p["hits"] + p["hits_device"] for p in per)
67 67 misses = sum(p["misses"] for p in per)
68 68 return {"layers": per, "hits": hits, "misses": misses,
69 69 "hit_rate": hits / (hits + misses) if hits + misses else None,
modified titan/slots.py +14 −2
@@ -88,6 +88,9 @@ class SlotTable:
88 88 self.stall_time = 0.0
89 89 self.loads = 0
90 90 self.prefetched_used = 0
91 + self.hits_device = 0 # accesses served on the fast path (table complete, no host sync)
92 + self._complete = False
93 + self.slot_map = None # device array expert -> slot, rebuilt when the table becomes complete
91 94
92 95 # ---- residency management -------------------------------------------------
93 96 def _install(self, expert: int, tensors: Dict[str, mx.array], protected: Iterable[int]) -> int:
@@ -103,8 +106,17 @@ class SlotTable:
103 106 self.slot_of[expert] = slot
104 107 self.expert_in[slot] = expert
105 108 self.loads += 1
109 + self._complete = False
106 110 return slot
107 111
112 + def complete(self) -> bool:
113 + """True iff every expert of the layer is resident (then no host synchronisation is needed)."""
114 + if not self._complete and len(self.slot_of) == self.spec.num_experts:
115 + self.slot_map = mx.array([self.slot_of[e] for e in range(self.spec.num_experts)], dtype=mx.uint32)
116 + mx.eval(self.slot_map)
117 + self._complete = True
118 + return self._complete
119 +
108 120 def load(self, expert: int, protected: Iterable[int] = ()) -> int:
109 121 """Synchronously load an expert from the store into a slot (demand miss)."""
110 122 t = time.perf_counter()
@@ -141,10 +153,10 @@ class SlotTable:
141 153
142 154 def stats(self) -> dict:
143 155 n = self.hits + self.misses
144 − return {"layer": self.layer, "capacity": self.capacity, "hits": self.hits, "misses": self.misses,
156 + return {"layer": self.layer, "capacity": self.capacity, "hits": self.hits, "misses": self.misses, "hits_device": self.hits_device,
145 157 "hit_rate": (self.hits / n) if n else None, "evictions": self.evictions, "loads": self.loads,
146 158 "stall_time_s": self.stall_time, "policy": self.policy.name}
147 159
148 160 def reset_stats(self):
149 − self.hits = self.misses = self.evictions = self.loads = 0
161 + self.hits = self.misses = self.evictions = self.loads = self.hits_device = 0
150 162 self.stall_time = 0.0
151 163
Menu

Theme
↑↓ navigate ↵ open esc close / or ⌘K anywhere