[Bugfix][MLA] Scope DSpark non-causal capability to builder layers (#57992)

Signed-off-by: Xavier Aguilar <xavier.aguilarfruto@amd.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
xaguilar-amd
2026-10-07 10:22:32 -05:00
committed by GitHub
co-authored by mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent 71f9a74920
commit 9367d8b969
2 changed files with 127 additions and 2 deletions
+121
View File
@@ -19,6 +19,21 @@ class _NonCausalMLAMetadataBuilder(MLACommonMetadataBuilder[MLACommonMetadata]):
supports_non_causal_multi_token_decode = True
class _CausalOnlyMLAMetadataBuilder(MLACommonMetadataBuilder[MLACommonMetadata]):
pass
class _DraftMLAMetadataBuilder(MLACommonMetadataBuilder[MLACommonMetadata]):
"""Triton-like: non-causal decode is supported, DCP draft decode is not."""
supports_non_causal_multi_token_decode = True
class _DcpCapableMLAMetadataBuilder(MLACommonMetadataBuilder[MLACommonMetadata]):
supports_non_causal_multi_token_decode = True
supports_non_causal_multi_token_dcp = True
def _metadata(
query_start_loc: list[int],
num_tokens: int | None = None,
@@ -60,6 +75,54 @@ def _builder(marked: bool = True) -> _NonCausalMLAMetadataBuilder:
return builder
def _mla_layer(*, non_causal: bool = False) -> SimpleNamespace:
return SimpleNamespace(
non_causal_multi_token_decode=non_causal,
q_lora_rank=None,
kv_lora_rank=512,
qk_nope_head_dim=128,
qk_rope_head_dim=64,
v_head_dim=128,
prefill_backend=SimpleNamespace(clone=lambda: None),
)
def _merged_mla_spec() -> MLAAttentionSpec:
return MLAAttentionSpec(
block_size=64,
num_kv_heads=1,
head_size=576,
dtype=torch.bfloat16,
non_causal_multi_token_decode=True,
)
def _dspark_dcp_vllm_config(
static_forward_context: dict[str, SimpleNamespace],
) -> SimpleNamespace:
return SimpleNamespace(
speculative_config=SimpleNamespace(
method="dspark", num_speculative_tokens=None
),
parallel_config=SimpleNamespace(
decode_context_parallel_size=2,
prefill_context_parallel_size=1,
cp_kv_cache_interleave_size=1,
),
compilation_config=SimpleNamespace(
static_forward_context=static_forward_context
),
model_config=SimpleNamespace(
dtype=torch.bfloat16,
max_model_len=128,
get_num_attention_heads=lambda _parallel_config: 16,
),
cache_config=SimpleNamespace(cache_dtype="auto", block_size=64),
scheduler_config=SimpleNamespace(max_num_seqs=1),
attention_config=SimpleNamespace(use_prefill_query_quantization=False),
)
def test_group_capability_keeps_runtime_causality_per_step():
builder = _builder()
@@ -137,3 +200,61 @@ def test_mla_cache_marker_is_promoted_to_group_capability():
[unmarked, unmarked]
).non_causal_multi_token_decode
assert MLAAttentionSpec.merge([unmarked, marked]).non_causal_multi_token_decode
def test_builder_scopes_noncausal_capability_to_its_layers():
merged_spec = _merged_mla_spec()
static_forward_context = {
"target": _mla_layer(non_causal=False),
"draft": _mla_layer(non_causal=True),
# Indexer and compressor caches share an MLA group but predate the flag.
"indexer": SimpleNamespace(),
}
vllm_config = _dspark_dcp_vllm_config(static_forward_context)
device = torch.device("cpu")
def _flag(layer_names: list[str]) -> bool:
return _DcpCapableMLAMetadataBuilder(
merged_spec,
layer_names,
vllm_config,
device,
supports_dcp_with_varlen=True,
).non_causal_multi_token_decode
assert not _flag(["target"])
assert _flag(["draft"])
assert _flag(["target", "draft"])
assert not _flag(["target", "indexer"])
def test_merged_group_spec_does_not_mark_a_target_only_builder():
# Both builders receive the same merged spec; only the draft layer is
# non-causal. Constructing through __init__ would still fail DCP
# validation if the builder read the group flag.
merged_spec = _merged_mla_spec()
static_forward_context = {
"target": _mla_layer(non_causal=False),
"draft": _mla_layer(non_causal=True),
}
vllm_config = _dspark_dcp_vllm_config(static_forward_context)
device = torch.device("cpu")
assert merged_spec.non_causal_multi_token_decode
target = _CausalOnlyMLAMetadataBuilder(
merged_spec,
["target"],
vllm_config,
device,
supports_dcp_with_varlen=True,
)
assert not target.non_causal_multi_token_decode
with pytest.raises(ValueError, match="non-causal draft"):
_DraftMLAMetadataBuilder(
merged_spec,
["draft"],
vllm_config,
device,
supports_dcp_with_varlen=True,
)
@@ -2384,8 +2384,12 @@ class MLACommonMetadataBuilder(AttentionMetadataBuilder[M]):
self.vllm_config = vllm_config
self.device = device
self.use_pcp = parallel_config.prefill_context_parallel_size > 1
self.non_causal_multi_token_decode = getattr(
kv_cache_spec, "non_causal_multi_token_decode", False
# merge() unions this flag across the KV group, so a shared target/draft
# group would mark the target builder as non-causal.
ctx = self.compilation_config.static_forward_context
self.non_causal_multi_token_decode = any(
getattr(ctx[name], "non_causal_multi_token_decode", False)
for name in layer_names
)
self._validate_dspark_dcp_support(supports_dcp_with_varlen)