3.6.1

[wangzhaode]

MNN 3.6.1 版本发布说明

发布日期: 2026年7月

📦 安装: pip install mnn==3.6.1 ｜ PyPI

📌 版本概述

MNN 3.6.1 版本聚焦于 Hexagon NPU 后端、Transformer C4 Fuse 全后端性能优化 与 稳定性修复 三大核心主题：

• Hexagon NPU 后端：新增高通 Hexagon NPU 直接编程后端，Qwen3-0.6B（W4 对称量化，block64）在骁龙 8 Elite（Hexagon v79）上 prefill 达 2667 tok/s，为 CPU 的 7.9×。
• Transformer C4 Fuse 全后端性能优化：新增 fused LLM 算子（RoPE / Attention / LayerNorm 等 C4 路径）并在 CPU / Metal / OpenCL / CUDA 全后端启用 Transformer C4 图；优化 C4 导出与 packed RoPE 路径；Metal 新增 LLM prefill/decode 优化 kernel 与 Fused Q4/Q8 dequant+GEMM（M_TILE=64）；OpenCL 修复 NC4HW4 raster L2 cache-set 抖动并优化 int4 GEMM。
• 稳定性修复与工具增强：新增 MNN-aware QLoRA 微调；修复 mmap 分配失败崩溃、推测解码 KVCache 索引、Arm82 LinearAttention fp16 精度等多项问题。

---

🚀 版本亮点

• Hexagon 后端（高通 NPU 直接编程）：Qwen3-0.6B 在骁龙 8 Elite 上 pp512 达 2667 tok/s（CPU 的 7.9×，OpenCL GPU 的 1.8×）
• Transformer C4 图全后端启用：fused LLM 算子 + C4 后端路径覆盖 CPU / Metal / OpenCL / CUDA，包含 packed RoPE fuse 优化
• Metal LLM Prefill 大幅优化：Fused Q4/Q8 dequant+GEMM kernel、M_TILE=64 变体（面向 M5）、prefill/decode kernel 与 shape/RoPE 基础设施重构，并新增 Metal Op Profile
• MNN-aware QLoRA 微调：新增面向 MNN 部署的 QLoRA 微调能力

---

✨ 新功能

后端：Hexagon（新增）

• Hexagon NPU 后端：新增高通 Hexagon NPU 直接编程后端（HexagonBackend / HexagonRuntime / DSP kernel），支持 Attention 等 LLM 关键算子；附带 skills/hexagon 开发文档

LLM / 推理加速

• Fused LLM 算子与 C4 后端路径：新增 CPURoPE、SharedGather 等 fused 算子及 CPU/Metal C4 计算路径
• C4 导出与 packed RoPE 优化：优化 C4 图导出；CPU / Metal / OpenCL 的 RoPE fuse（packed RoPE）路径优化
• Transformer C4 图全后端启用：CPU / Metal / OpenCL / CUDA（新增 RoPEExecution.cu、LayerNorm/LinearAttention C4 适配）统一启用 C4 图
• MNN-aware QLoRA 微调：新增 QLoRA 微调支持

GPU 后端

• Metal LLM prefill/decode kernel 优化：prefill/decode kernel、shape/RoPE 基础设施与 MetalConvolution1x1 SIMD-group 路径重构，新增 Metal Op Profile 工具
• Metal Fused Q4/Q8 dequant+GEMM：新增融合反量化+GEMM kernel 及 M_TILE=64 变体，优化 M5 上的 LLM prefill
• OpenCL raster 与 int4 GEMM 优化：转置遍历顺序修复 NC4HW4 L2 cache-set 抖动；优化 int4 GEMM 性能；拆分 binary_layernorm_c4 kernel
• Vulkan Range 算子：新增 Range 算子 image 实现
• QNN Interp/Resize 算子：QNN 后端新增 Interp/Resize 支持

CPU 后端

• RVV CommonOptFunction 适配：为 RISC-V RVV 架构适配 CommonOptFunction

工具

• llm_bench -fa 选项：新增 flash attention 开关（默认 1，启用 flash attention）

---

🐛 缺陷修复

LLM / VLM

• Qwen3.5 fused 推理正确性：修复 Qwen3.5 fused 推理结果错误
• C4 fuse 保护与 Metal raster：为 C4 fuse 增加 guard 条件并修复 Metal raster 问题
• RoPE schema 编译错误：修复 rope schema 引入的编译错误
• MPS 大 tensor Q4 量化静默全零：修复在 MPS 上量化大 tensor 时 Q4 权重静默全零的问题
• 通用 embedding scale 导出：修复导出时 embedding scale 缺失
• WikiText 校准数据集加载：修复量化校准数据集加载失败
• QNN 导出 chunk_size：修复 QNN prefill IO 生成未遵循 --chunk_size 的问题
• decoder embedding 导出 causal mask：decoder 类 embedding 模型导出改用 causal mask

GPU 后端

• Metal ATTENTION_C4 输出分支缺失：修复 prefill_qkv_tensor 缺少 ATTENTION_C4 输出分支
• Metal attention kv_cache flag：修复 Metal attention 未遵循 kv_cache flag
• OpenCL attention c4 kernel 编译：修复 attention c4 的 OpenCL kernel 编译错误
• OpenCL c4 长 prefill attention：修复 GEMM tune 时 V-buffer 越界与 NC4HW4 输出布局缺失
• OpenCL maskless attention setArg 溢出：修复导致 Adreno 上 SD UNet 输出错误的 setArg 溢出（issue 3.6.0 OpenCL regression on Adreno 830: SD UNet output wrong (mean-biased) from first inference, diverges to inf — CPU correct, 3.5.x correct #4653），同时修复 LinearAttention 越界写、共享 runtime gpuMode 被覆盖及 shared_gather 构建错误
• CUDA pooling pads 与 TopKV2 largest flag：修复显式 pooling pads 与 TopKV2 largest flag 处理，并重组相关测试
• QNN mmpose 模型：修复运行 mmpose 模型时的若干问题
• QNN context 生成栈溢出：提高 context 生成的栈上限

CPU 后端

• Arm82 LinearAttention fp16 精度：改用 fp32 累加器，修复 fp16 精度损失
• 推测解码 KVCache 索引：修复 CPU KVCache value 索引错误导致 Eagle3 输出质量下降（issue CPU KVCache索引问题降低 Eagle3输出质量 #4591）
• ConvInt8 SIGSEGV：修复 Conv 缺少量化权重时 ConvInt8 崩溃
• RVV 源文件缺失 Macro.h：补全 RISC-V RVV 源文件的头文件包含
• RVV Int8 函数修复：修复 MNNLineDepthWiseInt8AddBiasScaleUnit 等 RVV int8 函数问题（issue [CPU:Bugfix] supplement missing headfile, and fix a bug in MNNLineDep… #4454）

框架 / 其它

• mmap 分配失败崩溃：修复各后端 mmap 分配失败时的崩溃
• RuntimeManager::setCache 空指针：增加 null runtime 保护
• Android 构建缺失源文件：CMakeLists 补充 diffusion_session 与 sana_session 源文件

---

📚 其他改进

• 开发流程文档：新增 development workflow guidelines
• CI 脚本整合：整合 CI 测试脚本与文档
• Shader 构建脚本现代化：支持复杂路径
• NeuroPilot 代码移出公开仓库：相关实现改由内部维护
• iOS 发布包更名：移除包名中的 CoreML 后缀
• Copybara 同步脚本可配置化
• PyMNN 打包更新：更新 3.6.1 发布的 PyMNN 打包配置

---

📊 性能基准

相比 3.6.0（同设备、同量化配置，模型用 3.6.1 重新导出）：

• Android 性能大幅回升：Qwen3-4B OpenCL prefill (pp512) +101.7%，Qwen3-4B CPU decode (tg128) +72.2%，Qwen3-0.6B OpenCL decode +76.1%
• macOS Metal 全线提升：Qwen3-0.6B prefill (pp1024) +50.1%、decode +36.1%；Qwen3-4B prefill (pp512) +18.8%

测试环境: Mac mini M4 (16GB) / OnePlus PJZ110 (Snapdragon 8 Elite) | 模型: Qwen3-0.6B / Qwen3-4B | 量化: W4 (block=64) | 工具: llm_bench (rep=5, median)

Hexagon NPU 后端性能

Qwen3-0.6B（4-bit 对称量化，block64，C4）在 SM8750（骁龙 8 Elite，Hexagon v79）上三后端对比：

后端	pp512 (tok/s)	tg128 (tok/s)	vs CPU prefill
Hexagon NPU	2667	68.5	7.9×
OpenCL GPU	1491	71.4	4.4×
CPU	336	94.3	1.0×

Hexagon NPU 在 prefill 上优势显著（7.9×），decode 目前仍以 CPU 最优。Hexagon SDK 5.x 与 6.x 编译的 DSP skeleton 性能一致（差异 <2%，误差范围内）。

---

🙏 致谢

衷心感谢所有贡献者对本版本的宝贵贡献（按 GitHub login 列出）：

• @jxt1234 — Hexagon 后端 (Hexagon 后端【高通NPU直接编程支持】 #4633)、fused LLM 算子与 C4 后端路径 ([LLM:Feature] Add segment operator runtime support #4554)
• @wangzhaode — Transformer C4 图全后端启用 ([LLM:Perf] Enable Transformer C4 graphs across backends #4629)、C4 导出与 packed RoPE 优化 ([LLM:Perf] Optimize C4 export and packed RoPE paths #4598)、QLoRA 微调 ([LLM:Feature] Add MNN-aware QLoRA finetuning #4560)、Qwen3.5 fused 推理修复 ([LLM:Bugfix] Fix Qwen3.5 fused inference correctness #4631)、C4 fuse guard 与 Metal raster 修复 ([LLM:Bugfix] Guard C4 fuse and fix Metal raster #4607)、Vulkan Range 算子 ([Vulkan:Feature] Add Range op image implementation #4563)、Metal kv_cache flag 修复 ([Metal:Bugfix] Respect attention kv_cache flag #4556)、embedding scale 导出修复 ([LLM:Bugfix] Export generic embedding scale #4615)
• @v0jiuqi — Metal LLM prefill/decode kernel 与 RoPE 基础设施 (* [Metal:Speed] LLM prefill/decode kernel + shape/RoPE infrastructure… #4650)、Fused Q4/Q8 dequant+GEMM ([Metal:Speed] Fused Q4/Q8 dequant+GEMM kernel + M_TILE=64 variant for… #4651)、llm_bench -fa 选项 ([Tools:Feature] llm_bench: add '-fa' meaning whether to use flash att… #4658)、Metal ATTENTION_C4 分支修复 ([Metal:Bugfix] prefill_qkv_tensor missing ATTENTION_C4 output branch #4654)、MPS Q4 全零权重修复 ([LLM:Bugfix] Fix silent all-zero Q4 weights when quantizing large ten… #4649)、Arm82 LinearAttention fp16 修复 ([CPU:Bugfix] Arm82 LinearAttention fp16: use fp32 accumulators #4634)、mmap 失败崩溃修复 ([Core:Bugfix] Fix crashes when mmap allocation fails across all backends #4630)
• @Qxinyu — OpenCL NC4HW4 raster / int4 GEMM 优化 ([OpenCL:Opt]Fix NC4HW4 L2 cache-set thrashing in raster by transposin… #4606)、c4 长 prefill attention 修复 ([OpenCL:Bugfix] Fix c4 long-prefill attention: V-buffer OOB in gemm t… #4609)、maskless attention setArg 溢出等修复 ([OpenCL:Bugfix] Fix maskless attention setArg overflow causing wrong SD UNet output on Adreno (#4653) #4659)、attention c4 kernel 编译修复 ([OpenCL:Bugfix] Fix opencl kernel compile bug for attention c4 #4644)、QNN mmpose 修复 ([QNN:Bugfix] Fix some bugs when run mmpose model #4626)、setCache 空指针保护 ([Express:Bugfix] Guard null runtime in RuntimeManager::setCache #4617)
• @Fatemanx (xly) — CUDA pooling pads 与 TopKV2 largest flag 修复 ([Backend] Fix explicit pooling pads and CUDA TopKV2 largest flag #4627)
• @LudovicoYIN — QNN 导出 chunk_size 修复 ([LLM:Bugfix] Fix QNN export chunk-size and per-task workdir handling #4611)、WikiText 校准数据集修复 ([LLM:Bugfix] Fix WikiText calibration dataset loading #4566)、QNN context 生成栈上限
• @huluhuluu — 推测解码 KVCache 索引修复 ([CPU:Bugfix] CPU KVCache value index for speculative decoding(issue#4591) #4600)
• @lanling-47 — QNN Interp/Resize 算子 ([QNN:Feature] Add Interp/Resize op support for QNN backend #4582)
• @jxgxxx — RVV CommonOptFunction 适配 ([CPU:Perf] Adapt CommonOptFunction for RVV architecture #4426)
• @Sherlockzhangjinge — RVV Int8 函数修复 ([CPU: Bugfix] fix the problem mentioned in #4454  #4552)
• @xuhaonan999 — RVV Macro.h 包含修复 ([CPU:Bugfix] Add missing Macro.h includes for RISC-V RVV sources #4588)
• @pingji2016 — Android 构建源文件修复 ([Android:Bugfix] Add missing diffusion_session and sana_session sourc… #4569)
• @RinZ27 — Shader 构建脚本现代化 ([Infra:Refact] Modernize shader build scripts to handle complex paths #4550)
• @devin-lai — decoder embedding causal mask 修复 ([Embedding:Bugfix] Use causal mask for decoder embedding export #4548)

---

MNN 3.6.1 Release Notes

Release Date: July 2026

📦 Install: pip install mnn==3.6.1 ｜ PyPI

📌 Overview

MNN 3.6.1 focuses on three core themes: Hexagon NPU Backend, Transformer C4 Fuse Optimization Across Backends, and Stability Fixes:

• Hexagon NPU Backend: New Qualcomm Hexagon NPU direct-programming backend. Qwen3-0.6B (W4 symmetric, block64) reaches 2667 tok/s prefill on Snapdragon 8 Elite (Hexagon v79), 7.9× faster than CPU.
• Transformer C4 Fuse Optimization Across Backends: New fused LLM ops (RoPE / Attention / LayerNorm C4 paths) with Transformer C4 graphs enabled across CPU / Metal / OpenCL / CUDA; optimized C4 export and packed RoPE (RoPE fuse) paths; Metal adds fused Q4/Q8 dequant+GEMM (M_TILE=64) and reworked LLM prefill/decode kernels; OpenCL fixes NC4HW4 raster L2 cache-set thrashing and optimizes int4 GEMM.
• Stability Fixes & Tooling: New MNN-aware QLoRA finetuning; fixes for mmap allocation failure crashes, speculative-decoding KVCache indexing, Arm82 LinearAttention fp16 accuracy, and more.

---

🚀 Highlights

• Hexagon Backend (Qualcomm NPU Direct Programming): Qwen3-0.6B pp512 reaches 2667 tok/s on Snapdragon 8 Elite (7.9× vs CPU, 1.8× vs OpenCL GPU)
• Transformer C4 Graphs Across Backends: Fused LLM ops + C4 backend paths across CPU / Metal / OpenCL / CUDA, including packed RoPE fuse optimization
• Metal LLM Prefill Optimization: Fused Q4/Q8 dequant+GEMM kernel, M_TILE=64 variant (for M5), reworked prefill/decode kernels with shape/RoPE infrastructure, plus a new Metal Op Profile tool
• MNN-aware QLoRA Finetuning: New QLoRA finetuning targeting MNN deployment

---

✨ New Features

Backend: Hexagon (New)

• Hexagon NPU Backend: New Qualcomm Hexagon NPU direct-programming backend (HexagonBackend / HexagonRuntime / DSP kernels) with LLM key ops such as Attention; ships with skills/hexagon development docs

LLM / Inference Acceleration

• Fused LLM Ops & C4 Backend Paths: New fused ops (CPURoPE, SharedGather, etc.) and CPU/Metal C4 compute paths
• C4 Export & Packed RoPE Optimization: Optimized C4 graph export; packed RoPE (RoPE fuse) paths across CPU / Metal / OpenCL
• Transformer C4 Graphs Enabled Across Backends: CPU / Metal / OpenCL / CUDA (new RoPEExecution.cu, LayerNorm/LinearAttention C4 adaptation)
• MNN-aware QLoRA Finetuning

GPU Backends

• Metal LLM prefill/decode Kernel Optimization: Reworked prefill/decode kernels, shape/RoPE infrastructure, MetalConvolution1x1 SIMD-group paths, and a new Metal Op Profile tool
• Metal Fused Q4/Q8 dequant+GEMM: New fused dequant+GEMM kernel with M_TILE=64 variant for LLM prefill on M5
• OpenCL Raster & int4 GEMM Optimization: Fixed NC4HW4 L2 cache-set thrashing by transposing traversal order; optimized int4 GEMM; split binary_layernorm_c4 kernel
• Vulkan Range Op: New image implementation
• QNN Interp/Resize Ops: Added to QNN backend

CPU Backend

• RVV CommonOptFunction Adaptation: Adapted CommonOptFunction for RISC-V RVV

Tools

• llm_bench -fa Option: New flash attention toggle (default 1, flash attention enabled)

---

🐛 Bug Fixes

LLM / VLM

• Qwen3.5 fused inference correctness: Fixed
• C4 fuse guard & Metal raster: Added guard conditions for C4 fuse and fixed Metal raster
• RoPE schema compile error: Fixed
• Silent all-zero Q4 weights on MPS: Fixed when quantizing large tensors
• Generic embedding scale export: Fixed missing embedding scale on export
• WikiText calibration dataset loading: Fixed
• QNN export chunk-size: Honor --chunk_size when generating QNN prefill IO
• Decoder embedding export causal mask: Use causal mask for decoder embedding export

GPU Backends

• Metal missing ATTENTION_C4 output branch: Fixed in prefill_qkv_tensor
• Metal attention kv_cache flag: Now respected
• OpenCL attention c4 kernel compile: Fixed
• OpenCL c4 long-prefill attention: Fixed V-buffer OOB in GEMM tune and missing NC4HW4 output layout
• OpenCL maskless attention setArg overflow: Fixed setArg overflow causing wrong SD UNet output on Adreno (issue 3.6.0 OpenCL regression on Adreno 830: SD UNet output wrong (mean-biased) from first inference, diverges to inf — CPU correct, 3.5.x correct #4653), plus LinearAttention OOB writes, shared runtime gpuMode clobbering, and shared_gather build error
• CUDA pooling pads & TopKV2 largest flag: Fixed explicit pooling pads and TopKV2 largest flag handling; reorganized related tests
• QNN mmpose model: Fixed several runtime bugs
• QNN context generation stack limit: Raised

CPU Backend

• Arm82 LinearAttention fp16 accuracy: Use fp32 accumulators
• Speculative decoding KVCache indexing: Fixed CPU KVCache value index degrading Eagle3 output quality (issue CPU KVCache索引问题降低 Eagle3输出质量 #4591)
• ConvInt8 SIGSEGV: Fixed crash when Conv lacks quantized weight
• Missing Macro.h in RVV sources: Added
• RVV Int8 function fixes: Fixed MNNLineDepthWiseInt8AddBiasScaleUnit and related RVV int8 functions (issue [CPU:Bugfix] supplement missing headfile, and fix a bug in MNNLineDep… #4454)

Core / Misc

• mmap allocation failure crashes: Fixed across all backends
• RuntimeManager::setCache null runtime: Guarded
• Android build missing sources: Added diffusion_session and sana_session to CMakeLists

---

📚 Other Improvements

• Development workflow guidelines: Added docs
• CI test scripts consolidation
• Shader build scripts modernization: Handle complex paths
• NeuroPilot implementation removed from public repository
• iOS release package renamed: Removed CoreML suffix
• Configurable Copybara sync script
• PyMNN packaging update: Updated PyMNN packaging for the 3.6.1 release

---

📊 Performance Benchmarks

Compared with 3.6.0 (same devices and quantization config, models re-exported with 3.6.1):

• Major Android recovery: Qwen3-4B OpenCL prefill (pp512) +101.7%, Qwen3-4B CPU decode (tg128) +72.2%, Qwen3-0.6B OpenCL decode +76.1%
• macOS Metal improvements across the board: Qwen3-0.6B prefill (pp1024) +50.1%, decode +36.1%; Qwen3-4B prefill (pp512) +18.8%

Environment: Mac mini M4 (16GB) / OnePlus PJZ110 (Snapdragon 8 Elite) | Models: Qwen3-0.6B / Qwen3-4B | Quantization: W4 (block=64) | Tool: llm_bench (rep=5, median)

Hexagon NPU Backend Performance

Qwen3-0.6B (4-bit symmetric, block64, C4) on SM8750 (Snapdragon 8 Elite, Hexagon v79), three-backend comparison:

Backend	pp512 (tok/s)	tg128 (tok/s)	vs CPU prefill
Hexagon NPU	2667	68.5	7.9×
OpenCL GPU	1491	71.4	4.4×
CPU	336	94.3	1.0×

Hexagon NPU excels at prefill (7.9×); CPU remains fastest for decode. DSP skeletons built with Hexagon SDK 5.x and 6.x perform identically (<2% difference, within noise).

---

🙏 Acknowledgements

We sincerely thank all contributors for their valuable contributions to this release (listed by GitHub login):

• @jxt1234 — Hexagon backend (Hexagon 后端【高通NPU直接编程支持】 #4633), fused LLM ops & C4 backend paths ([LLM:Feature] Add segment operator runtime support #4554)
• @wangzhaode — Transformer C4 graphs across backends ([LLM:Perf] Enable Transformer C4 graphs across backends #4629), C4 export & packed RoPE optimization ([LLM:Perf] Optimize C4 export and packed RoPE paths #4598), QLoRA finetuning ([LLM:Feature] Add MNN-aware QLoRA finetuning #4560), Qwen3.5 fused inference fix ([LLM:Bugfix] Fix Qwen3.5 fused inference correctness #4631), C4 fuse guard & Metal raster fix ([LLM:Bugfix] Guard C4 fuse and fix Metal raster #4607), Vulkan Range op ([Vulkan:Feature] Add Range op image implementation #4563), Metal kv_cache flag fix ([Metal:Bugfix] Respect attention kv_cache flag #4556), embedding scale export fix ([LLM:Bugfix] Export generic embedding scale #4615)
• @v0jiuqi — Metal LLM prefill/decode kernels & RoPE infrastructure (* [Metal:Speed] LLM prefill/decode kernel + shape/RoPE infrastructure… #4650), fused Q4/Q8 dequant+GEMM ([Metal:Speed] Fused Q4/Q8 dequant+GEMM kernel + M_TILE=64 variant for… #4651), llm_bench -fa option ([Tools:Feature] llm_bench: add '-fa' meaning whether to use flash att… #4658), Metal ATTENTION_C4 branch fix ([Metal:Bugfix] prefill_qkv_tensor missing ATTENTION_C4 output branch #4654), MPS all-zero Q4 weights fix ([LLM:Bugfix] Fix silent all-zero Q4 weights when quantizing large ten… #4649), Arm82 LinearAttention fp16 fix ([CPU:Bugfix] Arm82 LinearAttention fp16: use fp32 accumulators #4634), mmap failure crash fix ([Core:Bugfix] Fix crashes when mmap allocation fails across all backends #4630)
• @Qxinyu — OpenCL NC4HW4 raster / int4 GEMM optimization ([OpenCL:Opt]Fix NC4HW4 L2 cache-set thrashing in raster by transposin… #4606), c4 long-prefill attention fix ([OpenCL:Bugfix] Fix c4 long-prefill attention: V-buffer OOB in gemm t… #4609), maskless attention setArg overflow & related fixes ([OpenCL:Bugfix] Fix maskless attention setArg overflow causing wrong SD UNet output on Adreno (#4653) #4659), attention c4 kernel compile fix ([OpenCL:Bugfix] Fix opencl kernel compile bug for attention c4 #4644), QNN mmpose fixes ([QNN:Bugfix] Fix some bugs when run mmpose model #4626), setCache null guard ([Express:Bugfix] Guard null runtime in RuntimeManager::setCache #4617)
• @Fatemanx (xly) — CUDA pooling pads & TopKV2 largest flag fix ([Backend] Fix explicit pooling pads and CUDA TopKV2 largest flag #4627)
• @LudovicoYIN — QNN export chunk-size fix ([LLM:Bugfix] Fix QNN export chunk-size and per-task workdir handling #4611), WikiText calibration fix ([LLM:Bugfix] Fix WikiText calibration dataset loading #4566), QNN context generation stack limit
• @huluhuluu — Speculative decoding KVCache index fix ([CPU:Bugfix] CPU KVCache value index for speculative decoding(issue#4591) #4600)
• @lanling-47 — QNN Interp/Resize ops ([QNN:Feature] Add Interp/Resize op support for QNN backend #4582)
• @jxgxxx — RVV CommonOptFunction adaptation ([CPU:Perf] Adapt CommonOptFunction for RVV architecture #4426)
• @Sherlockzhangjinge — RVV Int8 function fixes ([CPU: Bugfix] fix the problem mentioned in #4454  #4552)
• @xuhaonan999 — RVV Macro.h include fix ([CPU:Bugfix] Add missing Macro.h includes for RISC-V RVV sources #4588)
• @pingji2016 — Android build sources fix ([Android:Bugfix] Add missing diffusion_session and sana_session sourc… #4569)
• @RinZ27 — Shader build scripts modernization ([Infra:Refact] Modernize shader build scripts to handle complex paths #4550)
• @devin-lai — Decoder embedding causal mask fix ([Embedding:Bugfix] Use causal mask for decoder embedding export #4548)

---

Full Changelog: 3.6.0...3.6.1

---

This discussion was created from the release 3.6.1.

[pingji2016]

这是来自QQ邮箱的自动回复邮件。邮件收到^_^

[Bullobis]

既然mnn支持视频生成了，那aop为啥不支持

[tomas-anderson-ds]

Thank you for your work. You're awesome!!!