Fix topology-aware incumbents for qwen27b tuning

configs/examples/dash0_qwen27b_tight_slo_run4_0_8k.json

@@ -1,7 +1,7 @@
 {
   "study_id": "dash0-qwen27b-tight-slo-10min-run4-chat-0-8k",
   "hardware": {
-    "gpu_count": 4,
+    "gpu_count": 8,
     "gpu_model": "H20",
     "host_candidates": [
       "dash0"
@@ -26,7 +26,6 @@
       "/home/admin/resource/model/464482ce/qwen3.5-27b/256k-0223-internal"
     ],
     "base_envs": {
-      "CUDA_VISIBLE_DEVICES": "4,5,6,7",
       "VLLM_DISABLE_COMPILE_CACHE": "1",
       "DS_LLM_IGNORE_WARMUP": "1",
       "DS_LLM_IGNORE_CHECK_WARMUP": "1",
@@ -73,10 +72,7 @@
       "mamba-cache-dtype": "float32",
       "skip-mm-profiling": true,
       "quantization": "fp8",
-      "tensor-parallel-size": 4,
-      "data-parallel-size": 1,
-      "expert-parallel-size": 1,
-      "max-num-seqs": 16,
+      "tensor-parallel-size": 1,
       "disable-log-requests": true
     },
     "tunable_envs": [
@@ -100,7 +96,7 @@
       "require_enable_expert_parallel_when_ep_gt_one": true,
       "validate_cuda_graph_sizes_divisible_by_tp_when_tp_ep_reduce_scatter": true,
       "allowed_tp_dp_products": [1, 2, 4, 8],
-      "allowed_tensor_parallel_sizes": [1, 2, 4],
+      "allowed_tensor_parallel_sizes": [1, 2, 4, 8],
       "allowed_data_parallel_sizes": [1, 2, 4, 8],
       "allowed_expert_parallel_sizes": [1]
     },