[Spec decode] Support variable-length decode for Kimi-K3 adaptive ver (#52988)

Signed-off-by: Albert Cheng <albecheng@nvidia.com>
Signed-off-by: Albert Cheng (Engrg-Hardware 1) <albecheng@login-bia01.bia.clusters.nvidia.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
Co-authored-by: Benjamin Chislett <bchislett@nvidia.com>
This commit is contained in:
Albert Cheng
2026-09-22 20:43:32 -07:00
committed by GitHub
co-authored by mergify[bot] <37929162+mergify[bot]@users.noreply.github.com> Benjamin Chislett
parent 355d789f0d
commit 88aa0d287d
13 changed files with 203 additions and 22 deletions
+2 -1
View File
@@ -353,9 +353,10 @@ class Dots3NoteMLAMetadataBuilder(TritonMLAMetadataBuilder):
query_start_loc_cpu: torch.Tensor,
query_start_loc_device: torch.Tensor,
num_decode_tokens: int,
max_query_len: int,
dcp_tot_seq_lens_device: torch.Tensor | None,
) -> Dots3NoteDecodeMetadata:
del max_seq_len, query_start_loc_device, num_decode_tokens
del max_seq_len, query_start_loc_device, num_decode_tokens, max_query_len
query_len = int(query_start_loc_cpu[1] - query_start_loc_cpu[0])
return Dots3NoteDecodeMetadata(
block_table=block_table_tensor,