Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in
Toggle navigation
Menu
Open sidebar
OpenDAS
vllm_cscc
Commits
a2ef06e1
Unverified
Commit
a2ef06e1
authored
Jan 30, 2026
by
Russell Bryant
Committed by
GitHub
Jan 30, 2026
Browse files
[Misc] offest -> offset in comments and variable names (#33444)
Signed-off-by:
Russell Bryant
<
rbryant@redhat.com
>
parent
0a3c71e7
Changes
2
Show whitespace changes
Inline
Side-by-side
Showing
2 changed files
with
3 additions
and
3 deletions
+3
-3
vllm/v1/attention/backends/mamba2_attn.py
vllm/v1/attention/backends/mamba2_attn.py
+1
-1
vllm/v1/spec_decode/eagle.py
vllm/v1/spec_decode/eagle.py
+2
-2
No files found.
vllm/v1/attention/backends/mamba2_attn.py
View file @
a2ef06e1
...
@@ -105,7 +105,7 @@ class Mamba2AttentionMetadata(BaseMambaAttentionMetadata):
...
@@ -105,7 +105,7 @@ class Mamba2AttentionMetadata(BaseMambaAttentionMetadata):
# Chunk-related metadata (only for prefill)
# Chunk-related metadata (only for prefill)
seq_idx_p
:
torch
.
Tensor
|
None
=
None
seq_idx_p
:
torch
.
Tensor
|
None
=
None
# cu_chunk_seqlen_p is a tensor of shape (nchunks+1,) that contains, for
# cu_chunk_seqlen_p is a tensor of shape (nchunks+1,) that contains, for
# each chunk, its off
e
sts into the varlen sequence dimension. It is defined
# each chunk, its offs
e
ts into the varlen sequence dimension. It is defined
# such that the i-th chunk contains tokens from cu_chunk_seqlen_p[i] to
# such that the i-th chunk contains tokens from cu_chunk_seqlen_p[i] to
# cu_chunk_seqlen_p[i+1].
# cu_chunk_seqlen_p[i+1].
cu_chunk_seqlen_p
:
torch
.
Tensor
|
None
=
None
cu_chunk_seqlen_p
:
torch
.
Tensor
|
None
=
None
...
...
vllm/v1/spec_decode/eagle.py
View file @
a2ef06e1
...
@@ -1038,7 +1038,7 @@ class SpecDecodeBaseProposer:
...
@@ -1038,7 +1038,7 @@ class SpecDecodeBaseProposer:
# [0, 1, 2, 3, 4, 5, 6, 7, 8] ->
# [0, 1, 2, 3, 4, 5, 6, 7, 8] ->
# [0, 1, 0, 1, 2, 3, 0, 1, 2]
# [0, 1, 0, 1, 2, 3, 0, 1, 2]
# _r1_ ____r2____ ___r3__
# _r1_ ____r2____ ___r3__
token_off
e
sts
=
(
token_offs
e
ts
=
(
self
.
token_arange_np
[:
total_num_tokens
]
-
new_query_start_locs_expanded
self
.
token_arange_np
[:
total_num_tokens
]
-
new_query_start_locs_expanded
)
)
...
@@ -1053,7 +1053,7 @@ class SpecDecodeBaseProposer:
...
@@ -1053,7 +1053,7 @@ class SpecDecodeBaseProposer:
# [0, 1, // req 1
# [0, 1, // req 1
# q1 + 0, q1 + 1, q1 + 2, q1 + 3, // req 2
# q1 + 0, q1 + 1, q1 + 2, q1 + 3, // req 2
# q1 + q2 + 0, q1 + q2 + 1, q1 + q2 + 2] // req 3
# q1 + q2 + 0, q1 + q2 + 1, q1 + q2 + 2] // req 3
token_indices_np
=
token_off
e
sts
+
old_query_start_locs_expanded
token_indices_np
=
token_offs
e
ts
+
old_query_start_locs_expanded
token_indices
=
torch
.
from_numpy
(
token_indices_np
).
to
(
device
,
non_blocking
=
True
)
token_indices
=
torch
.
from_numpy
(
token_indices_np
).
to
(
device
,
non_blocking
=
True
)
spec_common_attn_metadata
=
CommonAttentionMetadata
(
spec_common_attn_metadata
=
CommonAttentionMetadata
(
...
...
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment