Skip to content
GitLab
Explore
Sign in
铁森林
2026abc
Repository
Branches
Overview
Active
Stale
All
exp/dense-w8a8-a47-20260720
e8ccfa3b
·
Explore dense MLP W8A8 runtime
·
Jul 20, 2026
exp/lm-head-w8-r4-a44-20260720
1eb112fb
·
docs: record A44 wheel verification
·
Jul 20, 2026
exp/int4-load-quant-a42-20260719
e85381df
·
[PRA26] explore persistent INT4 decode weights
·
Jul 20, 2026
exp/gdn-stateio-a37-20260716
51cdd892
·
Document A37 and H8 validation
·
Jul 16, 2026
opt/mrope-h2d-next-20260716
44384816
·
Add H1 scheduler trace evidence
·
Jul 16, 2026
experiment-score90-packed-qk
512fdddd
·
docs: record packed qk prefill candidate
·
Jul 15, 2026
experiment-score90-gfx936
c14ff064
·
docs: record profiled prefill gemm results
·
Jul 15, 2026
opt/final-gfx936-20260715
3f298197
·
Align PRA26 submission documents with true baseline
·
Jul 15, 2026
final-gfx936-compliant
1ce9b954
·
Submit compliant gfx936 single-request optimization
·
Jul 14, 2026
hwj-714-tunableop-submit108
48deb293
·
submit 030 force full streaming throughput decode
·
Jul 14, 2026
submit-121-packed-qk
57fb9e94
·
Enable packed QK path for submission
·
Jul 14, 2026
merge-108-safe
f45ecc31
·
docs: mark C3 excluded, fix serve_args=CUSTOM description
·
Jul 14, 2026
hwj-714
c7b146a3
·
revert serve_args.txt to empty (drop --attention-backend CUSTOM)
·
Jul 14, 2026
hwj713
c7b146a3
·
revert serve_args.txt to empty (drop --attention-backend CUSTOM)
·
Jul 14, 2026
submit-108-a1
168402b8
·
Add guarded gfx936 flash SDPA prefill backend
·
Jul 14, 2026
hwj20260713
f0e6c72d
·
088 legal chain: compliant optimizations on vllm_cscc baseline
·
Jul 13, 2026
submit-093-9a
9a343af5
·
Tune fixed decode GEMV launch configs
·
Jul 13, 2026
hwj0629
8b4d152a
·
082 clean: 16384x5120 BN=16 microbench -3.1% (基于submit-030)
·
Jul 12, 2026
submit-030-stream-ignore-eos
ac155d6c
·
remove serve_args.txt restore 88.31
·
Jul 12, 2026
submit-024-gateup-triton
351f0bfe
·
submit 024 gateup triton optimization
·
Jul 02, 2026
Prev
1
2
3
Next