Skip to content

Pull requests: FlashML-org/FreeToken

New pull request New
75 Open 16 Closed
75 Open 16 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

feat(rocm): AMD ROCm support (gfx1100) + Qwen3.5-MoE GGUF loader
#217 opened Aug 26, 2026 by samuelishida Loading…
feat(models): add Qwen3-Next-80B-A3B support
#212 opened Aug 26, 2026 by akushonkamen Loading…
fix(gguf): chunk the moe_vec z grid past the 65535 cap
#211 opened Aug 26, 2026 by avlp12 Loading…
GGUF: serve DeepSeek-V4-Flash
#210 opened Aug 26, 2026 by vcruz305 Loading…
1
qwen3_5_moe: load mixed-precision compressed-tensors checkpoints (unsloth NVFP4+FP8)
#208 opened Aug 26, 2026 by chrisqianz Loading…
perf(moe): pool heterogeneous GGUF cache rows
#199 opened Aug 26, 2026 by pedro-moser • Draft
1
fix(engine): reserve explicit KV pages during MoE auto-sizing
#198 opened Aug 26, 2026 by pedro-moser Loading…
feat(server): add Poolside V1 parsers
#197 opened Aug 26, 2026 by pedro-moser • Draft
1
perf(gguf): sm_120 dispatch thresholds + upstream int8-MMA MMQ port (Ornith)
#196 opened Aug 26, 2026 by lucaspirola Loading…
demo: intentionally duplicate maintenance gate (radar fixture)
#192 opened Aug 25, 2026 by Mitaligrawal Loading…
1 task done
refactor(server): dedupe maintenance gate across API entrypoints
#191 opened Aug 25, 2026 by Mitaligrawal Loading…
2 tasks
fix(gemma4-gguf): accept scalar attention.head_count_kv
#190 opened Aug 25, 2026 by qxZap Loading…
fix(installer): isolate system curl from AppImage library paths
#189 opened Aug 25, 2026 by nekomario28 • Draft
1
fix(kernel): launch GGUF MoE GEMV token axis on grid.x, not grid.z
#185 opened Aug 25, 2026 by erichstuntebeck Loading…
1
2
fix(deps): use PyTorch 2.13 to fix SM89 hangs
#184 opened Aug 25, 2026 by endenis Loading…
1
fix(attention): map grouped decode heads to KV groups
#177 opened Aug 25, 2026 by taking-lying-flat Loading…
fix(engine): load prefill triton kernels at startup, not mid-request
#169 opened Aug 25, 2026 by jason-fxz Collaborator Loading…
cpu-device: OR-switch device resolver + RoPE multi-head fallback fix
#160 opened Aug 25, 2026 by spaz122772 Loading…
fix(kernel): GGUF JIT extension fails to build with gcc under C++17 - compile as C++20
#159 opened Aug 25, 2026 by avlp12 Loading…
GGUF: read multi-shard checkpoints
#154 opened Aug 24, 2026 by vcruz305 Loading…
fix(bench): skip an unsupported slot count instead of aborting the sweep
#153 opened Aug 24, 2026 by rakhimovv Loading…
1
3
test(e2e): let the cache-rebuild gate run on reasoning checkpoints
#145 opened Aug 24, 2026 by rakhimovv Loading…
fix(kernel): clear the latched CUDA error before raising in pinned_tensor
#144 opened Aug 24, 2026 by rakhimovv Loading…
GGUF kernels: raise on unsupported quant type instead of returning uninitialized memory
#138 opened Aug 24, 2026 by vcruz305 Loading…
1
feat(rocm): serve on AMD GPUs through the HIP toolchain
#137 opened Aug 24, 2026 by paralin Loading…
ProTip! Type g i on any issue or pull request to go back to the issue listing page.