From 5e89a12997886bd541323015c2a2835e5e263530 Mon Sep 17 00:00:00 2001 From: whn09 Date: Sat, 24 Oct 2026 10:15:43 +0800 Subject: [PATCH] Neuron: whole-layer region graphs by default (TPOT 447.8 -> 430.3 ms at TP=64) Co-Authored-By: Claude Opus 5.5 --- deepinfer/environ.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/deepinfer/environ.py b/deepinfer/environ.py index 9c6fc9e..b880f7c 100644 --- a/deepinfer/environ.py +++ b/deepinfer/environ.py @@ -115,7 +115,7 @@ class envs: # Neuron only: cut the region graphs further at the stage boundaries marked by neuron_ops.regions.split(). DEEPINFER_NEURON_REGION_SPLIT = EnvVar.bool(default=False) # Neuron only: one region graph per decoder layer (attention half + MoE) instead of two. - DEEPINFER_NEURON_REGION_LAYER = EnvVar.bool(default=False) + DEEPINFER_NEURON_REGION_LAYER = EnvVar.bool(default=True) # Neuron only: split the region graphs of prefill chunks (steps of more than one row) regardless. DEEPINFER_NEURON_REGION_SPLIT_PREFILL = EnvVar.bool(default=True) # Neuron only: synchronize around every region graph call and log per-kind times each decode step (profiling).