fix: restore YaRN rope-scaling for reasoning-predictor (GPTQ requant dropped it, checkpoint's own ceiling was 40960 not 131072)

k8s/apps/llm-serving/reasoning.yaml

@@ -38,11 +38,19 @@ spec:
       # TP>1 restriction unlike bnb, so tensor-parallel-size=2 is worth
       # trying later, but not risking a parallelism-strategy change in the
       # same rollout as the quant+context-length change.
+      # This GPTQ requant's own config.json ships max_position_embeddings=
+      # 40960 and rope_scaling=None -- confirmed directly (curl'd the raw
+      # config.json), the base Qwen3-32B repo's YaRN block did NOT carry
+      # over during quantization. Re-applying it explicitly here restores
+      # the same math the base model documents (32768 * 4.0 = 131072);
+      # without this, --max-model-len=131072 fails ModelConfig validation
+      # against the checkpoint's own (unscaled) 40960 ceiling.
       - --model=JunHowie/Qwen3-32B-GPTQ-Int4
       - --served-model-name=reasoning
       - --quantization=gptq
       - --dtype=float16
       - --kv-cache-dtype=auto
+      - --rope-scaling={"rope_type":"yarn","factor":4.0,"original_max_position_embeddings":32768}
       - --tensor-parallel-size=1
       - --pipeline-parallel-size=2
       - --max-model-len=131072