whn09 7f558a3775 neuron: workers bind memory to their own core's NUMA node; prompt chunks run the attention half as one graph; glibc trims freed heap early (TTFT 796 -> 720 ms on the bench prompt, jitter +-25 -> +-3 ms) ... - NUMA: each worker sees one core, so which_numa(local_rank) failed on all but local rank 0 and every worker inherited its MPOL_BIND to node 1: all host memory landed on one node and a run could be OOM-killed. - DEEPINFER_NEURON_WHOLE_ATTN_MIN_ROWS=512: no cuts inside the attention half of a >=512-row step, only before the MoE (its collectives are another group kind). - MALLOC_TRIM_THRESHOLD_/MALLOC_TOP_PAD_ defaults: compile-time host memory per rank stops growing ~15 GB. (MALLOC_ARENA_MAX=2 bounds it too but costs ~50 ms TTFT.) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>		2026-10-29 12:10:31 +08:00
..
mtp	neuron: drafter forward as one region, compiled dflash metadata fill; verification + drafts regions on by default	2026-10-27 00:14:32 +08:00
v4_modules	neuron: the grouped wv projection takes the bf16 context through the NKI qdq kernel instead of an in-graph fp8 encode (TTFT 1009 -> 942 ms, step 437 -> 414 ms, tokens identical)	2026-10-28 14:32:47 +08:00
vision_encoder	initial commit	2026-10-20 12:09:42 +08:00
__init__.py	initial commit	2026-10-20 12:09:42 +08:00
decoder_layer.py	neuron: workers bind memory to their own core's NUMA node; prompt chunks run the attention half as one graph; glibc trims freed heap early (TTFT 796 -> 720 ms on the bench prompt, jitter +-25 -> +-3 ms)	2026-10-29 12:10:31 +08:00
deepseek_vl.py	Neuron: full DeepSeek-V4.1-Flash load + decode region graphs at EP=64 (issues #21, #25)	2026-10-23 01:39:22 +08:00
deepseek.py	neuron: low-latency completion off by default; rank timeline (DEEPINFER_NEURON_RANK_TIMELINE) and Engram prefault	2026-10-27 22:24:34 +08:00