[WIP][Spec Decode] DSpark confidence-scheduled verification

tests/evals/gsm8k/configs/DeepSeek-V4-Flash-DSpark-confidence-TP4.yaml

@@ -0,0 +1,22 @@
+model_name: "deepseek-ai/DeepSeek-V4-Flash-DSpark"
+accuracy_threshold: 0.92
+num_questions: 1319
+num_fewshot: 5
+startup_max_wait_seconds: 1800
+server_args: >-
+  --tokenizer-mode deepseek_v4
+  --trust-remote-code
+  --dtype bfloat16
+  --max-model-len 8192
+  --tensor-parallel-size 4
+  --enable-expert-parallel
+  --block-size 256
+  --gpu-memory-utilization 0.5
+  --kv-cache-dtype fp8
+  --max-num-batched-tokens 16384
+  --max-num-seqs 32
+  --speculative-config '{"method":"dspark",
+  "model":"deepseek-ai/DeepSeek-V4-Flash-DSpark",
+  "attention_backend":"FLASH_ATTN","num_speculative_tokens":7,
+  "draft_sample_method":"probabilistic","dspark_sps_curve":"auto",
+  "confidence_based_verification":"auto"}'

tests/test_config.py

@@ -1572,6 +1572,33 @@ def test_draft_sample_method_gumbel_is_rejected():
         )
 
 
+def test_dspark_confidence_config_validation():
+    config = SpeculativeConfig(
+        method="ngram",
+        num_speculative_tokens=1,
+        dspark_budget_frac=0.5,
+        confidence_based_verification="mask",
+    )
+    assert config.dspark_budget_frac == 0.5
+    assert config.confidence_based_verification == "mask"
+    with pytest.raises(ValidationError, match="confidence_based_verification"):
+        SpeculativeConfig(
+            method="ngram",
+            num_speculative_tokens=1,
+            confidence_based_verification="compact",
+        )
+
+
+@pytest.mark.parametrize("value", [0.0, 1.1])
+def test_dspark_confidence_config_rejects_invalid_ranges(value):
+    with pytest.raises(ValueError, match="dspark_budget_frac"):
+        SpeculativeConfig(
+            method="ngram",
+            num_speculative_tokens=1,
+            dspark_budget_frac=value,
+        )
+
+
 def test_ir_op_priority_default():
     """Test that IR op priority defaults are set correctly."""
     from vllm.config.kernel import IrOpPriorityConfig