riotpiao-poimen/homelab
2
0
Fork 0
Code Issues 24 Pull Requests 3 Actions Packages Projects 1 Releases Wiki Activity

hold clef-candle at replicas 0 until the image carries the GDN fix

Browse Source 
candle-vllm PR #5 (the conv fix) could not be merged, so CI has not published
a :sm70 image containing it. Deploying clef against an older :sm70 fails
silently: the delta rule returns zero on all 24 linear-attention layers and
/v1/systemone still answers 200 with well-formed but wrong classifications.

Raise to 1 once a :sm70 build from the fix merge exists.
This commit is contained in:
rock
2026-11-05 09:41:28 +08:00
parent 64a3f131b5
commit 9617bbe814
1 changed files with 6 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
k8s/apps/llm-serving/clef-candle.yaml
+6 -1
View File
@@ -7,7 +7,12 @@ metadata:
app.kubernetes.io/part-of: llm-serving app.kubernetes.io/part-of: llm-serving
app: clef-candle app: clef-candle
spec: spec:
replicas: 1 # HOLD AT 0 until candle-vllm:sm70 contains the GDN conv fix (candle-vllm PR #5).
# An older :sm70 zeroes the key channel band, starves the delta rule and makes
# every linear-attention layer return zero. The response is still HTTP 200 and
# well formed, so the failure is silent. Raise to 1 only after CI publishes a
# :sm70 build from the merge that carries the fix.
replicas: 0
strategy: strategy:
# One replica holds one GPU, so Recreate keeps the rollout to a single # One replica holds one GPU, so Recreate keeps the rollout to a single
# weights load instead of needing a second card for a surge pod. # weights load instead of needing a second card for a surge pod.