Train multimodal large models on Huawei Ascend NPU using MindSpeed-MM: set up base environment with CANN, PyTorch/torch_npu, and Megatron-LM; convert HuggingFace weights via mm-convert CLI; run end-to-end pipelines for VLMs like Qwen2.5VL/InternVL/GLM4V and generative models like Wan/HunyuanVideo/CogVideoX/FLUX with Megatron/FSDP2 trainers.
npx claudepluginhub ascend-ai-coding/awesome-ascend-skills --plugin hiascend-forumMindSpeed-MM skill router and model index for Huawei Ascend NPU. Use when the user is uncertain which MindSpeed-MM skill to use, needs to choose between understanding/generative/omni/audio model categories, or wants an overview of the full training pipeline. Routes to the appropriate leaf skill based on model type.
MindSpeed-MM multimodal model suite environment setup guide for Huawei Ascend NPU. Covers CANN activation, PyTorch + torch_npu installation, MindSpeed acceleration library, Megatron-LM core module integration, and MindSpeed-MM installation. Use when setting up MindSpeed-MM multimodal training environment on Ascend NPU.
MindSpeed-MM weight conversion guide using mm-convert CLI tool. Covers HuggingFace to MindSpeed-MM format conversion, reverse conversion, and PP weight resplitting. Supports Qwen2VLConverter, Qwen2_5_VLConverter, InternVLConverter, WanConverter and more. Use when converting multimodal model weights on Ascend NPU.
Universal VLM (vision-language understanding model) training guide for Huawei Ascend NPU using MindSpeed-MM. Covers all three framework patterns (Megatron, FSDP2, Custom trainers), weight conversion, dataset preparation (MLLM JSON format), fine-tuning, inference, and evaluation. Supports Qwen2.5VL, Qwen2VL, Qwen3VL, InternVL2.5/3/3.5, GLM4.1V, GLM4.5V, DeepSeekVL2, DeepSeekOCR, Ming, and more. Use when training or fine-tuning any multimodal understanding model on Ascend NPU.
Universal MindSpeed-MM generative model training guide for Huawei Ascend NPU. Covers all backend patterns (Megatron, Megatron+FSDP2, FSDP2-native, Accelerate+DeepSpeed), feature extraction, weight conversion, and training for ALL supported generative models. Supports Wan2.1/2.2, HunyuanVideo/1.5, CogVideoX, OpenSoraPlan, VACE, LTX2, FLUX, SD3, SDXL, Sana, HiDream, StepVideo, Lumina and more. Use when training multimodal generative models on Ascend NPU.
Share bugs, ideas, or general feedback.
Specialized skills for LLM engineering tasks including model training, evaluation, fine-tuning, and deployment optimization.
Transfer learning adaptation
Agent Skills for Together AI platform — inference, training, embeddings, audio, video, images, function calling, and infrastructure
SkyPilot agent skill for launching cloud VMs, Kubernetes pods, and Slurm jobs across 25+ clouds
Deploy and benchmark vLLM with Claude Code