frostbyte_neo
  • 加入于 2026-07-22
a1da37bcc7 feat(gallery): add Qwopus Flash Q8 variant
2026-09-08 09:17:01 +00:00
74a9716e7b feat(gallery): add Huihui Qwen3.8 quantizations
2026-09-08 09:17:01 +00:00
2026-09-08 09:17:01 +00:00
2026-09-08 09:17:01 +00:00
e8546965c7 fix(gallery): default audio-cpp models to backend:best (#11892)
ba88fb13ce fix(diffusers): auto-detect CUDA instead of defaulting to CPU (#11891)
e0179a1d3b chore(website): refresh the counters (#11902)
caa92a7af4 feat(swagger): update swagger (#11914)
c718937900 chore: ⬆️ Update 0xShug0/audio.cpp to 9c6a282337cc83f227cc10428867a478947706ad (#11915)
比较 18 提交 »
2026-09-08 09:17:00 +00:00
490e4ba778 test(worker): canonicalize missing upload fixture
4e4597dfc2 fix(worker): resolve staging directory symlinks
2756eaeab0 Merge master into the distributed test branch
561d623976 feat(gallery): add EXL3 vllm.cpp models (#11912)
aff9db9758 fix(distributed): resolve paths for virtual models (#11911)
比较 17 提交 »
2026-09-08 09:17:00 +00:00
946a7d5c7a fix(gallery): avoid duplicate Ornith 35B variants
561d623976 feat(gallery): add EXL3 vllm.cpp models (#11912)
aff9db9758 fix(distributed): resolve paths for virtual models (#11911)
e494033607 fix(distributed): finalize stalled model uploads (#11910)
88d19567b8 feat(faces): replay saved face enrollments (#11908)
比较 158 提交 »
2026-09-08 09:16:59 +00:00
7f9cf9c63d feat(diffusers): add AudioLDM2 generation
e8546965c7 fix(gallery): default audio-cpp models to backend:best (#11892)
ba88fb13ce fix(diffusers): auto-detect CUDA instead of defaulting to CPU (#11891)
e0179a1d3b chore(website): refresh the counters (#11902)
caa92a7af4 feat(swagger): update swagger (#11914)
比较 21 提交 »
2026-09-08 09:16:58 +00:00
28c4ad994e fix(gallery): persist inference defaults under parameters
57f802aa7a chore: ⬆️ Update leejet/stable-diffusion.cpp to d8fb10c02977c8ca999f3fb4e02df9ecf10f7ba6 (#11898)
f94ce520ce chore: ⬆️ Update antirez/ds4 to 9ab705347c1775e7599ede7eb81a6255ec7dccb5 (#11808)
5728f78f2d chore: ⬆️ Update ggml-org/llama.cpp to 465e49b9cea78a68b9c244ffb48d0ee24a82873d (#11897)
763bf6fbab chore: ⬆️ Update 0xShug0/audio.cpp to f6277c1695a83cf388a8282c1c1a8757cf626f18 (#11899)
比较 479 提交 »
2026-09-08 09:16:58 +00:00
5aa22ab2f8 feat(window-layouts): add incremental width and height resize actions (#382) (#385)
c9af3318a6 Merge pull request #391 from xcv58/codex/issue-309-cli-nightly-distribution
e03d5fb7b4 fix(nightly): preserve build artifacts across partial retries
2b889c028d Preserve multi-finger taps and prevent duplicate clicks around TipTap (#392)
6c6ebaf99f fix(settings): match sidebar sort icon to section title
比较 22 提交 »
2026-09-08 09:15:26 +00:00
b0905a8880 Merge branch 'main' into yweng/dev/opt_dflash2
606a92ffc7 fix ci
a1666e3856 fix ci
fe70b6059a drop the capture-sink source scan the runtime already fails loudly without
c8cb24f1a8 share one cold-L2 harness and shape table between the two GEMM tuning scripts
比较 39 提交 »
2026-09-08 09:13:48 +00:00
902b437f54 fix(v4): make benchmark invocation explicit
cd481b56ae perf(v4): add agentic benchmark
比较 2 提交 »
2026-09-08 09:13:48 +00:00
495dcd04d6 fix(pd): stop failing prefill rooms that are only queued on the decode
2026-09-08 09:13:48 +00:00
2026-09-08 09:13:48 +00:00
4df638af12 fix(attention): keep window_left 0 as a one-token window
9cb255115d test(cache): hold the remaining cache-group fixtures to their family shape
5ff98774ef refactor(scheduler): a State cache group is the checkpoint shape
3e02c6ae3a refactor(cache): hold each cache-group family to its declaration shape
8bf5b806cb refactor(scheduler): carry one block_granularity across the cache-group bridge
比较 15 提交 »
2026-09-08 09:13:47 +00:00
7ac3dd3f72 fix(cache): combine every rank's local checkpoint digest
cd6be4e009 test(cache): require two-page lookback in L3 shortage regression
53e024db49 fix(cache): rematch shortened Host L3 prefixes and converge backup errors
2ef6017c3b fix(cache): hash imported local checkpoint package code
aae6fd29ef fix(cache): require HF hub cache layout for snapshot commit IDs
比较 17 提交 »
2026-09-08 09:13:47 +00:00
acd12a30ca perf(gfx950): load compact e8m0 scales in the SiTU decode GEMVs (#1426)
11b47c13b5 perf(l2): static geometry, block-id transfers, and one-kernel H2D flags (#1384)
edcd0ab3bb fix(qwen): prevent eager PLE index corruption after graph replay (#1425)
4a60ccf97c perf(k3): shard the latent MoE down projection by column (#1383)
5dac46e2db perf(qsa): reduce CuTe sparse decode latency (#1420)
比较 7 提交 »
2026-09-08 09:13:47 +00:00
2f7b1957d2 refactor(attention): commit QSA verify state from forward context
227fdc3a50 Merge branch 'main' into codex/qsa-backend-refactor
0918752ed7 Merge main into codex/qsa-backend-refactor
acd12a30ca perf(gfx950): load compact e8m0 scales in the SiTU decode GEMVs (#1426)
11b47c13b5 perf(l2): static geometry, block-id transfers, and one-kernel H2D flags (#1384)
比较 24 提交 »
2026-09-08 09:13:46 +00:00
2026-09-08 09:13:16 +00:00