whn09 3c38fd5829 Neuron MoE: skip unrouted experts inside the fp4 kernels at decode sizes (DEEPINFER_NEURON_GATED_EXPERTS, default on) ... The fp4 expert GEMMs dequantize the whole weight to bf16 in HBM on every call (device profile: the MoE graph reads 756 MB a layer), and a rank computes all 6 of its experts while a 6-row step routes to about one. fp4_linear_2core_gated takes a per-expert flag (any row routed to it) and runs the dequant and GEMM in an nl.fori_loop bounded by it, so an unrouted expert costs a zero fill. Decode sizes only (<= 16 rows); prompt chunks route to every expert anyway. bench70 (vllm-sonnet 512, TP=64, nextn=5): TPOT 258.9 -> 248.4 ms, step 738 -> 711 ms, tokens per step identical. Tiny cp=2 nextn=5 on device: 9/9 golden tokens, drafts unchanged. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>		2026-10-26 06:36:48 +08:00
..
_C	initial commit	2026-10-20 12:09:42 +08:00
distributed	initial commit	2026-10-20 12:09:42 +08:00
engine	Neuron: sampling flags read once a step; runtime profiling on the driver only; one-lookup embedding	2026-10-25 03:52:47 +08:00
entrypoints	initial commit	2026-10-20 12:09:42 +08:00
hugging_face	Neuron: concatenate FP8 checkpoint tensors as bytes (fixes MoE on device)	2026-10-21 11:46:58 +08:00
model_executor	Neuron: eager embedding/gather without dtype casts or host scalar operands	2026-10-25 06:55:43 +08:00
neuron_ops	Neuron MoE: skip unrouted experts inside the fp4 kernels at decode sizes (DEEPINFER_NEURON_GATED_EXPERTS, default on)	2026-10-26 06:36:48 +08:00
npu_ops	initial commit	2026-10-20 12:09:42 +08:00
structure_output	initial commit	2026-10-20 12:09:42 +08:00
ve	initial commit	2026-10-20 12:09:42 +08:00
worker	Neuron table attention: decode only the blocks a decode step's requests reference	2026-10-25 17:58:11 +08:00
__init__.py	Neuron basic mode: try prefill on every step	2026-10-21 21:00:54 +08:00
attrdict.py	initial commit	2026-10-20 12:09:42 +08:00
config.py	initial commit	2026-10-20 12:09:42 +08:00
environ.py	Neuron DSpark: draft sampling as one region graph for greedy rows; token embedding on device by default	2026-10-25 06:06:50 +08:00
load_class.py	initial commit	2026-10-20 12:09:42 +08:00
logger.py	initial commit	2026-10-20 12:09:42 +08:00
model_config_loader.py	initial commit	2026-10-20 12:09:42 +08:00
multimodal_tokens.py	initial commit	2026-10-20 12:09:42 +08:00
platform.py	Neuron: static decode graphs, static-shape MoE, trn2 e4m3 fp8 (issues #9, #12, #24)	2026-10-22 11:59:37 +08:00
py.typed	initial commit	2026-10-20 12:09:42 +08:00
sampling_params.py	initial commit	2026-10-20 12:09:42 +08:00
sequence.py	initial commit	2026-10-20 12:09:42 +08:00
shared_memory.py	initial commit	2026-10-20 12:09:42 +08:00
testing.py	initial commit	2026-10-20 12:09:42 +08:00
timer.py	initial commit	2026-10-20 12:09:42 +08:00
utils.py	Neuron basic mode — part 3: end-to-end check vs the official reference	2026-10-20 20:20:02 +08:00