Skip to content

GitLab

Unverified Commit da6709c9 authored by Lucas Wilkinson's avatar Lucas Wilkinson Committed by GitHub
Browse files

[Misc] Delay deprecation of CommonAttentionMetadata properties (#32074) 


Signed-off-by: default avatarLucas Wilkinson <lwilkins@redhat.com>
parent d83becd5
Hide whitespace changes
Inline Side-by-side
Showing with 3 additions and 3 deletions
vllm/v1/attention/backends/utils.py
View file @ da6709c9
...@@ -96,7 +96,7 @@ class CommonAttentionMetadata: ...@@ -96,7 +96,7 @@ class CommonAttentionMetadata:
dcp_local_seq_lens_cpu: torch.Tensor | None = None dcp_local_seq_lens_cpu: torch.Tensor | None = None
"""Sequence lengths of the local rank in decode context parallelism world""" """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 _seq_lens_cpu: torch.Tensor | None = None
_num_computed_tokens_cpu: torch.Tensor | None = None _num_computed_tokens_cpu: torch.Tensor | None = None
...@@ -107,7 +107,7 @@ class CommonAttentionMetadata: ...@@ -107,7 +107,7 @@ class CommonAttentionMetadata:
""" """
Prefer using device seq_lens directly to avoid implicit H<>D sync. Prefer using device seq_lens directly to avoid implicit H<>D sync.
If a CPU copy is needed, use `seq_lens.cpu()` instead. 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: def seq_lens_cpu(self) -> torch.Tensor:
...@@ -121,7 +121,7 @@ class CommonAttentionMetadata: ...@@ -121,7 +121,7 @@ class CommonAttentionMetadata:
Prefer using device seq_lens directly to avoid implicit H<>D sync which breaks full 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 async scheduling. If a CPU copy is needed, it can be derived from
query_start_loc_cpu and seq_lens. 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: def num_computed_tokens_cpu(self) -> torch.Tensor:
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment