Skip to content

Pull requests: FlashML-org/FreeToken

New pull request New
112 Open 26 Closed
112 Open 26 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

qwen4_exp: load modelopt MIXED_PRECISION (NVFP4 experts + block-FP8 dense) checkpoints
#320 opened Sep 1, 2026 by gberasmus87 Loading…
fix(ftw): preserve auxiliary expert banks
#317 opened Sep 1, 2026 by ErikBPF Loading…
fix(rocm): avoid indirect host pointers during HIP graph capture
#316 opened Sep 1, 2026 by ErikBPF Loading…
Fix GGUF user-defined token atomicity
#315 opened Aug 31, 2026 by tuxevil Loading…
feat(qwen4_exp): stream the PLE n-gram table from disk
#311 opened Aug 31, 2026 by jason-fxz Collaborator Loading…
2
feat(kvcache): add reliable quantized KV cache
#309 opened Aug 31, 2026 by plsgivemeachane Loading…
1
feat(server): add --api-key bearer authentication
#305 opened Aug 31, 2026 by benwilson Loading…
1
feat: Automatic KV/MoE Laddering for decode speed vs context-length trade off ( upto 33% faster decode )
#300 opened Aug 30, 2026 by aswinkumar1999 Loading…
Support poolside/Laguna-S-2.1-NVFP4, and fix serving from source on Windows
#298 opened Aug 30, 2026 by Eng-Ahmd Loading…
feat(qwen3_5_moe): support compressed-tensors NVFP4 experts with block-fp8 dense side
#296 opened Aug 30, 2026 by JIAQI13 Loading…
1
Add Docker container capability
#295 opened Aug 30, 2026 by recrudesce Loading…
4 tasks done
2
fix(qwen3_5_moe): detect shared-expert quant for modelopt mixed checkpoints
#294 opened Aug 30, 2026 by JIAQI13 Loading…
2
fix(qwen4_exp): fix nvfp4 ftw weight loading, qsa indexer, and ple tables
#293 opened Aug 30, 2026 by 3Simplex • Draft
3 tasks done
1
feat: support GLM-5.3-Flash NVFP4
#292 opened Aug 30, 2026 by earlvanze Loading…
2
fix(scheduler): hybrid radix hit corruption — GDN snapshot copy-on-donate + admission barriers
#287 opened Aug 30, 2026 by Cerynitius Loading…
1
Fix/windows serving
#285 opened Aug 30, 2026 by fangyuan-3149 Loading…
fix(models): load NVFP4 expert banks from checkpoints without a safetensors index
#283 opened Aug 30, 2026 by paulp83 Loading…
feat(qwen4-exp): mmap PLE tables from disc
#279 opened Aug 29, 2026 by ryseek Loading…
1
3
fix(moe): match the benchbw verdict to the served model's expert size
#278 opened Aug 29, 2026 by bernimccoy Contributor Loading…
fix: [BUG] qwen3_5 dense: mixed-precision NVFP4 crashes in ct_bf16_fuse (Float8 × BFloat16 promotion)
#275 opened Aug 29, 2026 by yuyi2439 Loading…
1
glm5_next: support GLM-5.3-Flash (hybrid KDA + DSA, mHC, NVFP4 expert offload)
#270 opened Aug 29, 2026 by Cerynitius Loading…
fix(server): report why a tokenizer/detokenizer worker died during load
#269 opened Aug 29, 2026 by sime2408 Loading…
feat(kvcache): sub-byte Q4_0 and Q6_0 KV cache quantization
#268 opened Aug 29, 2026 by fangyuan-3149 Loading…
4
Add loopback runtime process identity endpoint
#267 opened Aug 29, 2026 by HaileyStorm Loading…
fix(server): hide qwen transport markers from semantic responses
#266 opened Aug 29, 2026 by HaileyStorm Loading…
1
ProTip! Updated in the last three days: updated:>2026-08-29.