# glm53-jovian-r12 build inputs — rtx6kpro models/glm-5.3-flash.md pins.
# digest-pinned day-0 arm64 cu130 base (same as glm53-flash-nvfp4-lab).
# NOTE: the doc's "composition commits" (7d8a09c… / d064ca4c…) are not
# public refs; we reproduce them as base + sequential PR-head merges —
# so here the REFs are the upstream BASES (branch tip / master).
BASE_IMAGE="vllm/vllm-openai:glm53-flash-arm64-cu130@sha256:905c02933be6021301db2dc284e24e3727467aa3a0f63b41d609885778a07bce"
VLLM_REF="54f6e9826c20ef06ed65d838c0ad497ad0abdecf"
B12X_REF="139e04048bc3bb4f7210c99e7184d8d2f0e345e7"