[Bugfix] Fix full-payload mm splitting for dual hidden/scheduled batch axes (#4870 follow-up) (#4910)

Signed-off-by: amy-why-3459 <wuhaiyan17@huawei.com>
This commit is contained in:
amy-why-3459 authored and GitHub committed 2026-07-06 15:08:05 +08:00
1 parent ba7e9e6496
commit 9ceca4ec85
4 files changed
+96 -18

No files matched your search

+12 -6
View File
@@ -839,7 +839,8 @@ class GPUARModelRunner(OmniGPUModelRunner, OmniConnectorModelRunnerMixin):
end: int,
audio_sparse_output: bool,
sparse_mm_index: dict[str, int],
seq_len: int,
hidden_seq_len: int,
scheduled_seq_len: int,
) -> dict[str, object]:
if combined_multimodal_outputs:
return self._build_combined_prefix_cache_mm_payload(
@@ -876,7 +877,8 @@ class GPUARModelRunner(OmniGPUModelRunner, OmniConnectorModelRunnerMixin):
start=start,
end=end,
pass_lists_through=False,
seq_len=seq_len,
seq_len=hidden_seq_len,
scheduled_seq_len=scheduled_seq_len,
)
return mm_payload
@@ -894,7 +896,8 @@ class GPUARModelRunner(OmniGPUModelRunner, OmniConnectorModelRunnerMixin):
mm_cpu: dict[str, object] | None,
audio_sparse_output: bool,
sparse_mm_index: dict[str, int],
seq_len: int,
hidden_seq_len: int,
scheduled_seq_len: int,
) -> dict[str, object]:
payload: dict[str, object] = {}
if not audio_sparse_output:
@@ -920,7 +923,8 @@ class GPUARModelRunner(OmniGPUModelRunner, OmniConnectorModelRunnerMixin):
end=end,
audio_sparse_output=audio_sparse_output,
sparse_mm_index=sparse_mm_index,
seq_len=seq_len,
hidden_seq_len=hidden_seq_len,
scheduled_seq_len=scheduled_seq_len,
)
payload.update(mm_payload)
return payload
@@ -1731,7 +1735,8 @@ class GPUARModelRunner(OmniGPUModelRunner, OmniConnectorModelRunnerMixin):
# The actual multimodal wire transport uses multimodal_outputs instead.
pooler_output: list[dict[str, object]] | None = None
if needs_pooler_payload:
mm_seq_len = int(scheduler_output.total_num_scheduled_tokens)
hidden_seq_len = int(hidden_states.shape[0])
scheduled_seq_len = int(scheduler_output.total_num_scheduled_tokens)
mm_cpu = None
if self.omni_prefix_cache is not None:
(
@@ -1787,7 +1792,8 @@ class GPUARModelRunner(OmniGPUModelRunner, OmniConnectorModelRunnerMixin):
mm_cpu=mm_cpu,
audio_sparse_output=audio_sparse_output,
sparse_mm_index=sparse_mm_index,
seq_len=mm_seq_len,
hidden_seq_len=hidden_seq_len,
scheduled_seq_len=scheduled_seq_len,
)
pooler_output.append(flatten_payload(payload))