[Model] Support DeepSeek-V4.1-Flash (#56214)

Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Signed-off-by: Jee Jee Li <jeejeelee@inferact.ai>
Signed-off-by: andyluo7 <andy.luo@amd.com>
Signed-off-by: zjy0516 <riverclouds.zhu@qq.com>
Co-authored-by: Jee Jee Li <jeejeelee@inferact.ai>
Co-authored-by: andyluo7 <andy.luo@amd.com>
Co-authored-by: OpenAI Codex <codex@openai.com>
Co-authored-by: zjy0516 <riverclouds.zhu@qq.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This commit is contained in:
Yongye Zhu
2026-09-11 02:11:19 -07:00
committed by GitHub
co-authored by Jee Jee Li andyluo7 OpenAI Codex zjy0516 Claude Opus 5
parent e6821ceac9
commit e77daef89e
49 changed files with 3996 additions and 138 deletions
@@ -91,6 +91,7 @@ class MooncakeStoreCoordinator:
assert all(
scheduler_block_size % g.kv_cache_spec.block_size == 0
for g in kv_cache_groups
if g.kv_cache_spec.prefix_cacheable
), "scheduler_block_size must be a multiple of each group's block_size"
self.kv_cache_groups = kv_cache_groups
self.mamba_group_ids = {
@@ -460,11 +460,13 @@ class NixlBaseConnectorWorker:
for spec in iter_layer_specs(group.kv_cache_spec)
)
]
if len(ple_groups) != 1 or len(ple_groups[0][1].layer_names) != 1:
if len(ple_groups) > 1 or any(
len(group.layer_names) != 1 for _, group in ple_groups
):
raise ValueError(
"CSA-linear NIXL requires exactly one PLE cache owner."
"CSA-linear NIXL requires at most one PLE cache owner."
)
self._ple_group_index = ple_groups[0][0]
self._ple_group_index = ple_groups[0][0] if ple_groups else None
if self._has_mamba:
assert self._is_hma_required