An interactive Ascend-NPU process viewer
-
Updated
Oct 10, 2026 - Python
#
cann
Here are 44 public repositories matching this topic...
RAG for visually rich documents — layout, formula and table structure survive the pipeline. Ported from CUDA to Huawei Ascend NPUs and tested on Ascend 910C.
-
Updated
Oct 8, 2026 - Python
MiniMax-H3 昇腾 NPU 本地部署（910B/910C 真机实测）：一键脚本 + 40MB/s 分片下载 + 性能实测数据 + 17 条踩坑记录 | MiniMax-H3 local deployment on Ascend NPU, tested on 910B
minimax quantization huawei cann uncensored video-generation image-to-video npu ascend text-to-video comfyui comfyui-workflow 910b local-deployment minimax-h3 910c
-
Updated
Oct 13, 2026 - Python
Deploy DeepSeek-V4.1 (dsv4.1) on Ascend 910B / 910B3 (NPU): a hardware-specialized LLM inference engine with W4A8, DSpark speculative decoding, 128K+ context and 8,700+ tok/s prefill. Also supports NVIDIA A100/A800. 昇腾910B 部署 DeepSeek-V4.1-Flash。
cuda quantization cann inference-engine npu ascend kv-cache ai-agent model-inference llm-serving llm-inference ascend-npu speculative-decoding deepseek deepseek-v4 huawei-ascend w4a8 ascend-910b deepseek-v41 dsv41
-
Updated
Oct 27, 2026 - C++
昇腾 NPU 训练与推理故障诊断知识库：按症状命中 case、用先验知识与源码分析定位根因，定位后进待审队列、每周升格回库。
pytorch knowledge-base diagnosis codex cann ai-agents aiops npu root-cause-analysis ascend mindspore llm vllm ascend-npu agent-skills claude-code agentskills skills-sh
-
Updated
Nov 13, 2026 - JavaScript
Native AscendC Mamba2 selective scan / SSD forward-backward custom operator for Huawei Ascend 910B3 and 950PR, with CANN, torch_npu, A100 benchmarks and msprof profiling.
pytorch ssd mamba cann performance-optimization state-space-models ascend-npu selective-scan ai-infrastructure ascendc kernel-optimization mamba2 huawei-ascend torch-npu custom-operator ascend-910b triton-ascend ascend-950
-
Updated
Oct 26, 2026 - Python
Ascend NPU fork of nanochat for LLM training with torch_npu/HCCL (experimental)
-
Updated
Feb 24, 2026 - Python
昇腾 Atlas 310B + YOLOv8n 果园果实检测模型工程化：端到端 p50 ≤30ms、mAP50-95 掉点 ≤1pt。含 AIPP/ATC 硬伤校验、六段时延预算、量化流水线与精度-时延帕累托前沿实验平台。
computer-vision object-detection cann ascend model-quantization edge-inference model-engineering yolov8 atlas310
-
Updated
Oct 30, 2026 - Python
Disaggregated Prefill-Decode serving engine connecting NVIDIA Blackwell (SM120) and Huawei Ascend 910B2 (CANN) across physical nodes.
cuda cann inference-engine heterogeneous-computing kv-cache vllm llm-inference pagedattention deepseek ai-infrastructure distributed-inference nvidia-blackwell disaggregated-serving huawei-ascend ascend-910b pytorch-npu davinci-architecture
-
Updated
Sep 24, 2026 - Python
Ascend NPU linear-attention operator optimization | GDN eager->AscendC on Qwen3.5-0.8B, 3.45x speedup with verified accuracy alignment
-
Updated
Sep 28, 2026 - Python
Out-of-tree HCCL Task DAG and P2P traffic capture studio
-
Updated
Aug 4, 2026 - Python
Qwen3.8-27B on original Ascend 910: custom CANN kernels, six-NPU experiments, measured decode speedups and reproducible build recipes.
-
Updated
Sep 19, 2026 - C++
TidalCache: Sparse KV Cache Offload for DeepSeek Sparse Attention (DSA) on Ascend NPU — only move what the model attends to
-
Updated
Oct 20, 2026 - HTML
Add this topic to your repo
To associate your repository with the cann topic, visit your repo's landing page and select "manage topics."