From da6709c9fe6965b7348692576ffadeee8439388e Mon Sep 17 00:00:00 2001 From: Lucas Wilkinson Date: Sat, 10 Jan 2026 00:06:44 -0500 Subject: [PATCH] [Misc] Delay deprecation of CommonAttentionMetadata properties (#32074) Signed-off-by: Lucas Wilkinson --- vllm/v1/attention/backends/utils.py | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/vllm/v1/attention/backends/utils.py b/vllm/v1/attention/backends/utils.py index eecd81a696..8e948e22a3 100644 --- a/vllm/v1/attention/backends/utils.py +++ b/vllm/v1/attention/backends/utils.py @@ -96,7 +96,7 @@ class CommonAttentionMetadata: dcp_local_seq_lens_cpu: torch.Tensor | None = None """Sequence lengths of the local rank in decode context parallelism world""" - # WARNING: Deprecated fields. Will be removed in a future release (v0.14.0) + # WARNING: Deprecated fields. Will be removed in a future release (v0.15.0) _seq_lens_cpu: torch.Tensor | None = None _num_computed_tokens_cpu: torch.Tensor | None = None @@ -107,7 +107,7 @@ class CommonAttentionMetadata: """ Prefer using device seq_lens directly to avoid implicit H<>D sync. If a CPU copy is needed, use `seq_lens.cpu()` instead. - Will be removed in a future release (v0.14.0) + Will be removed in a future release (v0.15.0) """ ) def seq_lens_cpu(self) -> torch.Tensor: @@ -121,7 +121,7 @@ class CommonAttentionMetadata: Prefer using device seq_lens directly to avoid implicit H<>D sync which breaks full async scheduling. If a CPU copy is needed, it can be derived from query_start_loc_cpu and seq_lens. - Will be removed in a future release (v0.14.0) + Will be removed in a future release (v0.15.0) """ ) def num_computed_tokens_cpu(self) -> torch.Tensor: