serverAppConfigs

Files

T

Frank caee26e3c3 llm: patch 04b v2 - plan-time signal is the spec-config METHOD

use_aux_hidden_state_outputs lives on the gpu_model_runner instance
(gpu_model_runner.py:623/678), not on the spec config, so the first
registry check never fired and the reserve stayed at 13,600 B/token.
Switch to method in {dflash, dspark, extract_hidden_states}.

2026-09-30 19:29:18 -07:00

glm53-compile-01-model-decorator.patch

frank-primary: compile changelist — real diffs, break diagnostics, patch 04, corrected overclaim

2026-09-30 13:59:09 -07:00

glm53-compile-02-trace-safe-probes.patch

frank-primary: compile changelist — real diffs, break diagnostics, patch 04, corrected overclaim

2026-09-30 13:59:09 -07:00

glm53-compile-03-fullgraph-optout.patch

frank-primary: pin down the compile blocker — a name-resolution KeyError in an inlined callee

2026-09-30 15:39:10 -07:00

glm53-compile-04-fused-qkv-rmsnorm-optout.patch

frank-primary: patch 05, five verified artifacts, and correct the wrong confound theory