minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[Kernel] Relocate vendored fla and mamba kernel trees to sglang.kernels (RFC #29630, Phase 2.5, 7/7) (#30795)

Browse Source 
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
This commit is contained in:
Xiaoyu Zhang
2026-07-15 12:52:15 +08:00
committed by GitHub
co-authored by Claude Fable 5
parent 23f2b77d82
commit 4aadf94146
88 changed files with 226 additions and 183 deletions
Download Patch File Download Diff File Expand all files Collapse all files
benchmark/bench_linear_attention/bench_gdn_prefill_cutedsl.py
+4 -4
View File
@@ -26,14 +26,14 @@ sys.path.insert(0, os.path.join(os.path.dirname(__file__), "..", "..", "python")
import torch
from sglang.kernels.ops.attention.fla.chunk import (
chunk_gated_delta_rule as triton_chunk_gated_delta_rule,
)
from sglang.kernels.ops.attention.fla.l2norm import l2norm_fwd
from sglang.kernels.ops.attention.linear.gdn_blackwell import (
chunk_gated_delta_rule_cutedsl,
prepare_metadata_cutedsl,
)
from sglang.srt.layers.attention.fla.chunk import (
chunk_gated_delta_rule as triton_chunk_gated_delta_rule,
)
from sglang.srt.layers.attention.fla.l2norm import l2norm_fwd
# ---------------------------------------------------------------------------
# Helpers (shared shape: pool layout [N, H, K, V] with K-last stride)