[Mamba] Add FlashInfer ReplaySSM support for MTP (#52928)

Signed-off-by: Andrii Skliar <askliar@nvidia.com>
Co-authored-by: Andrii Skliar <askliar@nvidia.com>
This commit is contained in:
Andrii Skliar
2026-09-30 21:41:17 +00:00
committed by GitHub
co-authored by Andrii Skliar
parent 7e583e615c
commit deb148e730
18 changed files with 1029 additions and 113 deletions
+1
View File
@@ -815,6 +815,7 @@ class NemotronHForCausalLM(
tp_world_size=parallel_config.tensor_parallel_size,
logical_window=cache_config.replayssm_buffer_len,
backend=vllm_config.mamba_config.backend,
num_speculative_tokens=vllm_config.num_speculative_tokens,
)
return base_shape