[Misc] Name each backend and its kernel block sizes in block-size errors (#58557)

Signed-off-by: Eugenio "Jay" Zuccarelli <11176606+jayzuccarelli@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>

tests/v1/worker/test_attn_utils.py

@@ -77,6 +77,7 @@ def test_get_kv_cache_spec_resolves_hisparse_block_size(
     layers = {}
     for name, sizes in zip(specs, [main_sizes, indexer_sizes, [block_size]]):
         backend = SimpleNamespace(
+            get_name=lambda name=name: name,
             customize_spec=AttentionBackend.customize_spec,
             get_supported_kernel_block_sizes=lambda sizes=sizes: sizes,
         )

tests/v1/worker/test_gpu_model_runner.py

@@ -298,6 +298,10 @@ def _make_mock_backend_for_kernel_block_size(
     supported_sizes: list[int | MultipleOf],
 ):
     class _MockBackend:
+        @staticmethod
+        def get_name():
+            return "MOCK"
+
         @staticmethod
         def get_supported_kernel_block_sizes():
             return supported_sizes

vllm/v1/attention/backend.py

@@ -54,6 +54,9 @@ class MultipleOf:
     def __init__(self, base: int):
         self.base = base
 
+    def __repr__(self) -> str:
+        return f"MultipleOf({self.base})"
+
 
 class AttentionBackend(ABC):
     """Abstract class for attention backends."""

vllm/v1/worker/utils.py

@@ -383,7 +383,13 @@ def select_common_block_size(
     for size in sorted(candidates, reverse=True):
         if block_size_is_supported(backends, size):
             return size
-    raise ValueError(f"No common block size for {kv_manager_block_size}.")
+    raise ValueError(
+        f"No common block size for {kv_manager_block_size} ("
+        + "; ".join(
+            f"{b.get_name()}: {b.get_supported_kernel_block_sizes()}" for b in backends
+        )
+        + ")."
+    )
 
 
 def allocate_kv_cache(