Skip to content

Actions: FlashML-org/FreeToken

Actions

All workflows

Actions

Loading...
Loading

Showing runs from all workflows
312 workflow runs
312 workflow runs

Filter by Workflow

⌥ + click/return to exclude

Filter by Event

⌥ + click/return to exclude

Filter by Status

⌥ + click/return to exclude

Filter by Branch

⌥ + click/return to exclude

Filter by Actor

⌥ + click/return to exclude
bench bw verdict misses in the other direction: auto stays on offload (ratio ~1.9), while hybrid decodes 21% / 55% faster (Qwen3.8-Flash-Next NVFP4, RTX 4090 + dual Xeon) Label issues and PRs #228: Issue #664 opened by oscarlius
12s
12s
fix(rocm): make CPU/Hybrid MoE graph replay safe Label issues and PRs #227: Pull request #378 edited by zihaomu
7s
7s
Nightly wheels Nightly wheels #67: Scheduled
4m 23s main
main
4m 23s
AMD Roadmap 2026 Q4-2027 Q1 Label issues and PRs #226: Issue #541 edited by zihaomu
9s
9s
AMD Roadmap 2026 Q4-2027 Q1 Label issues and PRs #225: Issue #541 edited by zihaomu
9s
9s
sm75 (Turing) fork: 35B MoE on an 8 GB RTX 2080 — 350/25 t/s, 64K context Label issues and PRs #224: Issue #663 opened by ljmbot
9s
9s
feat(cpu-moe): serve block-fp8 experts on the cpu and hybrid backends Label issues and PRs #223: Pull request #662 edited by jason-fxz
7s
7s
feat(cpu-moe): serve block-fp8 experts on the cpu and hybrid backends Label issues and PRs #222: Pull request #662 edited by jason-fxz
8s
8s
feat(cpu-moe): serve block-fp8 experts on the cpu and hybrid backends Label issues and PRs #221: Pull request #662 opened by jason-fxz
7s
7s
Requests stall and time out above --max-running-requests 8; GPU stays at ~28% of TDP and only one CPU thread is busy Label issues and PRs #220: Issue #661 opened by yakold
9s
9s
Nightly wheels Nightly wheels #66: Scheduled
4m 19s main
main
4m 19s
feat(moe): support gsq q2_0 routed experts Label issues and PRs #219: Pull request #660 opened by wenlive
7s
7s
fix(kernel): enable disk row stores on windows Label issues and PRs #218: Pull request #659 opened by wenlive
7s
7s
Upgrade to torch 2.14.1, Triton 3.8 and sglang-kernel 0.4.9 Label issues and PRs #217: Issue #629 edited by jason-fxz
7s
7s
build(deps): move to torch 2.14.1, triton 3.8 and sglang-kernel 0.4.9 Label issues and PRs #216: Pull request #630 edited by jason-fxz
5s
5s
Nightly wheels Nightly wheels #65: Scheduled
3m 30s main
main
3m 30s
Support Qwen3.8-Flash-Next-GSQ-RCO GGUF and integrate Strata + KVMem Label issues and PRs #215: Issue #658 opened by yutils
11s
11s
feat(server): add --api-key authentication Label issues and PRs #214: Pull request #639 edited by jason-fxz
8s
8s
Upgrade to torch 2.14.1, Triton 3.8 and sglang-kernel 0.4.9 Label issues and PRs #213: Issue #629 edited by jason-fxz
9s
9s
feat(models): add bounded encoder swa replay behind --swa-encoder-replay Label issues and PRs #212: Pull request #657 opened by void-mckenzie
7s
7s
fix(moe): refuse e8m0 scale code 0xFF when MXFP4 expert pieces are read Label issues and PRs #211: Pull request #656 opened by Avicennasis
7s
7s
MXFP4 expert loaders accept the e8m0 NaN scale code (0xFF) and serve NaN blocks Label issues and PRs #210: Issue #655 opened by Avicennasis
7s
7s
fix(kernel): avoid nan in triton sampling softmax when a chunk starts with -inf Label issues and PRs #209: Pull request #654 opened by ischencheng
9s
9s
Nightly wheels Nightly wheels #64: Scheduled
4m 1s main
main
4m 1s
test(kernel): cover Triton activation and norm fallbacks Label issues and PRs #208: Pull request #652 edited by zihaomu
7s
7s