whn09 380e160b4b Neuron flash_mla: round attention P to bf16 like the CUDA kernels ... The pure-torch sparse attention followed FlashMLA's tests/ref.py (fp32 P). The CUDA kernels (sm100 prefill phase1.cuh:238-253) and the official reference round P = exp(s - running_max) to bf16 before the P@V GEMM in an online softmax; do the same with 64-key blocks. Localized with examples/neuron/probe_layer0.py (per-op reference vs DeepInfer capture): layer 0 is now bit-exact up to two rounding points that are DeepInfer's own GPU/Ascend behaviour (SwiGLU -> FP8 from fp32, bf16 routed partial in the MoE combine). CPU end-to-end on the tiny checkpoint: 26/27 greedy tokens match the reference, the miss being a 0.0096-margin step the reference flips against itself. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>		2026-10-20 20:48:36 +08:00
..
vendor	Neuron flash_mla: round attention P to bf16 like the CUDA kernels	2026-10-20 20:48:36 +08:00
__init__.py	Add AWS Trainium (Neuron) platform, basic mode — part 1	2026-10-20 18:34:51 +08:00
NEURON_PORT.md	Add AWS Trainium (Neuron) platform, basic mode — part 1	2026-10-20 18:34:51 +08:00
transfer_to_neuron.py	Neuron basic mode — part 2: deep_ep, sampling kernels, fixes	2026-10-20 18:57:27 +08:00