Skip to content

Pull requests: ztxz16/fastllm

New pull request New
21 Open 177 Closed
21 Open 177 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

支持 compressed-tensors INT8 per-channel 权重加载与 CUDA 算子
#746 opened Sep 19, 2026 by wdyun Loading…
gguf: fix IQ2_S/IQ3_S mixed-quant loading, and let IQ3_S use the MMQ fast path
#744 opened Sep 19, 2026 by 113636xfh Loading…
新增 qwen3_5 多模态 E/PD 分离（Encode 独立部署）
#743 opened Sep 17, 2026 by zify9000 Loading…
均衡 DFlash2 草稿权重显存并修复图片能力上报与含图投机
#741 opened Sep 16, 2026 by tonykee Loading…
新增 NCCL 初始化通信自检，避免 TP 归约静默失败导致输出退化
#734 opened Sep 14, 2026 by zify9000 Loading…
单个请求的内部错误不再终止服务进程，并支持外部 MTP 与 mmproj 共存
#732 opened Sep 12, 2026 by 113636xfh Loading…
修复长上下文分块预填充卡顿与 MTP 草稿缓存失步
#731 opened Sep 12, 2026 by 113636xfh Loading…
1
改进 Qwen3.5 前缀快照：解码阶段记录、间隔与保留数量、共享缓冲
#730 opened Sep 12, 2026 by 113636xfh Loading…
fix: GGUF UD 混合量化类型导致 MTP 权重无法融合
#721 opened Sep 8, 2026 by jyf2100 Loading…
2 of 3 tasks
fix: qwen3.5 稠密模型的 cpu/cuda 分层设备字典被融合单卡路径忽略
#720 opened Sep 8, 2026 by jyf2100 Loading…
2 tasks done
Fix GGUF tokenization: BPE merges are applied left-to-right instead of by merge rank
#715 opened Aug 19, 2026 by cyjin-yl Loading…
Validate safetensors header fields before using them as sizes
#713 opened Aug 11, 2026 by professor-moody Loading…
1
支持 Qwen3.5/3.6 GGUF、V100 SM70 与原生 Turbo3 KV
#707 opened Jul 31, 2026 by cyjin-yl Loading…
5
ft支持w4a8, w4a8 moe, fp8 量化（仅针对SM90 卡子）
#706 opened Jul 30, 2026 by moguizhizi Loading…
7
Bound ModelLoader::ReadString/ReadBytes against the mapped file size (OOB read on crafted .flm)
#702 opened Jul 17, 2026 by professor-moody Loading…
修复MSVC 2017和GCC 8.x的CPU编译，clang编译
#694 opened Jun 7, 2026 by TylunasLi Contributor Loading…
增加对昇腾全系列NPU的支持 （on v0.1.5.1）
#677 opened May 19, 2026 by TylunasLi Contributor Loading…
添加了对Qwen3-coder-next的gguf模型的张量并行推理
#674 opened May 12, 2026 by a793181018 Loading…
修复 OpenAI 工具调用多轮消息的结构化 content 兼容问题
#663 opened Apr 21, 2026 by Li2718 Contributor Loading…
修复 --cache_dir 在子命令前使用时无法识别的问题
#662 opened Apr 21, 2026 by Li2718 Contributor Loading…
fix the numactl command control failure
#642 opened Jan 20, 2026 by pengjunjie2100 Loading…
ProTip! Exclude everything labeled bug with -label:bug.