monster/corridorkey-mrp-mlx
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

docs: retitle for the 4-machine matrix (M1/M3/M4/M5, Ultra->laptop)

Browse Source 
Also validated the tiled-compile patch against the repo's own standards:
ruff check + ruff format clean, pytest 64 passed / 0 failed.
This commit is contained in:
modelbeast 2026-07-16 23:56:38 +10:00
parent fd00c31b1d
commit 64be04aeac
1 changed files with 5 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
docs/2026-07-16-m-series-fleet-ablation-results.md
View File

@ -1,5 +1,5 @@
---
title: "M-series fleet ablation: M3 Ultra / M1 Ultra / M4 Pro"
title: "M-series fleet ablation: M1 / M3 / M4 / M5 (Ultra → laptop)"
date: 2026-07-16
device: "M3 Ultra 256GB · M1 Ultra 128GB · M4 Pro mini 24GB · M5 MacBook Pro 16GB (10-core)"
script: scripts/bench_optimizations.py (unmodified) + engine-level tiled×compile extension
@ -8,8 +8,10 @@ script: scripts/bench_optimizations.py (unmodified) + engine-level tiled×compil
# M-series Fleet Ablation Benchmarks
Same methodology as the 2026-03-09 wave2 doc (`--sweep ablation --resolution 512 1024 2048`,
checkpoint weights, 3 warmup + 10 bench runs), executed on three machines spanning the
Apple Silicon range. Contributed from a 3-Mac render-farm setup; happy to run follow-ups.
checkpoint weights, 3 warmup + 10 bench runs), executed across four machines spanning the
Apple Silicon range — **M1 Ultra (64c) · M3 Ultra (80c) · M4 Pro mini 24GB · M5 MacBook Pro
16GB (10c)** — i.e. oldest→newest GPU generation and Studio→laptop. Contributed from a
multi-Mac render-farm setup; happy to run follow-ups on any of them.
## Headline findings