leavelet d6aac79eb1 Lean tensor dump: residual trajectory only (attn_in/attn_out + q_all), drop kv/MoE ...

The full-stage+kv dump (1.7GB/file, synchronous in-forward) hung/crashed a scheduler
child. Trim to the residual-stream trajectory (attn_in/attn_out per layer) + q_all +
final_hidden (~tens of MB/forward) — enough to localize the first divergent layer (Stage 1).
Drop dense-space kv_cache/topk (not comparable anyway) and the MoE substages. Dumps go to
local NVMe (/ssd) to avoid beegfs stall.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

2026-06-17 15:09:39 +00:00

..

sglang

Lean tensor dump: residual trajectory only (attn_in/attn_out + q_all), drop kv/MoE

2026-06-17 15:09:39 +00:00

pyproject_cpu.toml

Upgrade transformers==5.3.0 (#17784)

2026-03-18 13:50:43 -07:00

pyproject_npu.toml

Upgrade transformers==5.3.0 (#17784)

2026-03-18 13:50:43 -07:00

pyproject_other.toml

chore: extract diffusion_common in python/pyproject_other.toml (#20803)

2026-03-19 10:39:16 +08:00

pyproject_xpu.toml

Upgrade transformers==5.3.0 (#17784)

2026-03-18 13:50:43 -07:00

pyproject.toml

new pyprojectoml

2026-06-10 05:54:43 +08:00