2026-11-05 02:20:01 +00:00
1 changed files with 13 additions and 1 deletions
@@ -37,7 +37,19 @@ spec:
          # and makes every linear-attention layer return zero. The symptom is
          # a well-formed /v1/systemone response with wrong answers, not an error.
          image: forgejo.riotpiao.com/riotpiao-poimen/candle-vllm:sm70
-          command: ["/usr/local/bin/candle-vllm"]
+          # :sm70 is a moving tag. Default IfNotPresent keeps a cached copy and a
          # rebuilt image never reaches the pod, so a merged fix looks like it did
          # not land. Always forces the pull on every pod start.
          imagePullPolicy: Always
          # Runs the fixed binary from the models PVC, not the one in the image.
          # The published :sm70 predates the /v1/systemone route and the GDN conv
          # fix (candle-vllm f3c00b2 is merged but CI has not republished the tag),
          # so the in-image binary 404s on the route. The PVC copy at
          # /mnt/models/bin/candle-vllm contains both fixes.
          #
          # RETIRE THIS once CI publishes a :sm70 built from f3c00b2: change the
          # command back to /usr/local/bin/candle-vllm and delete the PVC copy.
          command: ["/mnt/models/bin/candle-vllm"]
          # clef-flash is a hybrid Qwen3_5 model: 8 full-attention layers plus 24
          # GDN layers that carry per-sequence recurrent state. It runs on ONE
          # card because the state is not shardable the way attention KV is.