The first local-NVMe recovery root is closed as an infrastructure-only abort:
/mnt/fast-ai/llm-optimization-artifacts/laguna-s-2.1/runs/w1-n128-nvme-postreboot-recovery-20260723T131632Z
It ran on clean boot 0b7f98a5-e50a-46a5-81ea-15938b55317a, kernel
7.0.0-28-generic, with taint 0. The gate verified all 118 local model
files, xpu-smi version and four-card discovery, exact PCI/DRM mapping, and
strict four-card idleness.
The next bounded command invoked oneAPI 2026 sycl-ls with
UR_LOG_LOADER=level_info. That value uses an obsolete logger syntax. The
tool printed:
Wrong format of the UR_LOG_LOADER environment variable value: 'level_info'
No platforms found
Platforms: 0
sycl-ls nevertheless exited zero. The gate then required all four
level_zero GPU rows, found none, and failed closed with status 1. It did not
run the peer binary, either XCCL pass, either N64 gate, N128, a model service,
or model generation. Kernel delta contained only -- No entries --; the Xe
reject file was empty and kernel taint remained zero.
The evidence manifest verifies completely. Stable hashes are:
evidence.sha256 06b27c3e2ac0e254c8294aa679face6329323dad2ce06579281bc2d10c343ca3
final-status.txt 73f83e71c30324b36f15b2fe2afefc97112fa2e66a0a23eb25019d784a5a42df
sycl-ls-verbose.log 26baddc04f0c4f14da045b85b32d38d192e832b4119fc5a15971ea353b47ef0b
kernel-delta.txt 19c4e28b2f54ea8f217db4e622b0a5c758efa7b77ce2dc0394b6d159568c0b2e
The only authorized correction is to explicitly unset UR_LOG_LOADER for
the already verbose sycl-ls command and use the fresh local root:
/mnt/fast-ai/llm-optimization-artifacts/laguna-s-2.1/runs/w1-n128-nvme-postreboot-recovery-20260723T134648Z
This retry is not conditioned on performance or candidate output. A1 remains the first model generation after a complete recovery pass.