[PCP][DCP] Enable PCP+DCP on sparse-MLA models (#56157)
Signed-off-by: patryk.saffer <patryk.saffer@blc-login-1.blc1.mistralcompute.ai> Signed-off-by: patryk.saffer <patryk.saffer@blc-login-0.blc1.mistralcompute.ai> Signed-off-by: PatrykSaffer <patryk.saffer@mistral.ai> Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com> Co-authored-by: patryk.saffer <patryk.saffer@blc-login-1.blc1.mistralcompute.ai> Co-authored-by: patryk.saffer <patryk.saffer@blc-login-0.blc1.mistralcompute.ai> Co-authored-by: Lucas Wilkinson <lwilkins@redhat.com> Co-authored-by: Patryk Saffer <PatrykSaffer@users.noreply.github.com> Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
This commit is contained in:
co-authored by
patryk.saffer
patryk.saffer
Lucas Wilkinson
Patryk Saffer
Claude Fable 5.1
parent
72d4d83157
commit
ebe1dec2da
@@ -114,6 +114,9 @@ class InputBatch:
|
||||
# (cudagraph capture) batches, which run the KV-sharing layers in full.
|
||||
fast_prefill: "FastPrefillBatchMetadata | None" = None
|
||||
|
||||
# [num_reqs] set only under PCP+DCP (see CommonAttentionMetadata).
|
||||
dcp_local_seq_lens_cpu_upper_bound: torch.Tensor | None = None
|
||||
|
||||
@classmethod
|
||||
def make_dummy(
|
||||
cls,
|
||||
|
||||
Reference in New Issue
Block a user