Skip to content

refactor(attn): let the bucketer define the batched decode kernel's buckets #792

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
jvlunteren wants to merge 2 commits into from
Closed

refactor(attn): let the bucketer define the batched decode kernel's buckets #792

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d8dd7ef
rename the bucketed decode kernel to batched decode
jvlunteren Sep 7, 2026
5084ae2
let the bucketer define the batched decode kernel's buckets
jvlunteren Sep 7, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 3 additions & 3 deletions spyre_inference/envs.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,7 +33,7 @@
SPYRE_ATTN_RECORD: bool = True
SPYRE_ATTN_KV_BUCKETS: str | None = None
SPYRE_ATTN_QUERY_BUCKETS: str | None = None
SPYRE_BUCKETED_DECODE: bool = False
SPYRE_BATCHED_DECODE: bool = False
SPYRE_NUM_CPUS: int = 0
SPYRE_UPDATE_THREAD_CONFIG: bool = True

Expand Down Expand Up @@ -63,10 +63,10 @@
# Comma-separated query_len buckets to record, unset uses the default buckets
# [1] + multiples of min(512, max_num_batched_tokens) up to max_num_batched_tokens.
"SPYRE_ATTN_QUERY_BUCKETS": lambda: os.getenv("SPYRE_ATTN_QUERY_BUCKETS"),
# When "1", enables the bucketed multi-sequence decode kernel. Off by default
# When "1", enables the batched multi-sequence decode kernel. Off by default
# pending performance characterisation at small batch sizes (num_seqs <= 4).
# Re-enable to measure the path or to restore it after calibration.
"SPYRE_BUCKETED_DECODE": lambda: bool(int(os.getenv("SPYRE_BUCKETED_DECODE", "0"))),
"SPYRE_BATCHED_DECODE": lambda: bool(int(os.getenv("SPYRE_BATCHED_DECODE", "0"))),
# CPU budget used to size thread pools. "0" (default) auto-detects the budget
# (cgroup CPU quota, then physical core count).
"SPYRE_NUM_CPUS": lambda: int(os.getenv("SPYRE_NUM_CPUS", "0")),
Expand Down
Loading
Loading