Loading...
Loading...
Found 115 Skills
Fine-tune and serve Physical Intelligence OpenPI models (pi0, pi0-fast, pi0.5) using JAX or PyTorch backends for robot policy inference across ALOHA, DROID, and LIBERO environments. Use when adapting pi0 models to custom datasets, converting JAX checkpoints to PyTorch, running policy inference servers, or debugging norm stats and GPU memory issues.
Guidance for implementing PyTorch pipeline parallelism for distributed model training. This skill should be used when tasks involve implementing pipeline parallelism, distributed training with model partitioning across GPUs/ranks, AFAB (All-Forward-All-Backward) scheduling, or inter-rank tensor communication using torch.distributed.
Battle-tested PyTorch training recipes for all domains — LLMs, vision, diffusion, medical imaging, protein/drug discovery, spatial omics, genomics. Covers training loops, optimizer selection (AdamW, Muon), LR scheduling, mixed precision, debugging, and systematic experimentation. Use when training or fine-tuning neural networks, debugging loss spikes or OOM, choosing architectures, or optimizing GPU throughput.
Provides PyTorch-native distributed LLM pretraining using torchtitan with 4D parallelism (FSDP2, TP, PP, CP). Use when pretraining Llama 3.1, DeepSeek V3, or custom models at scale from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing.
Bump the NVIDIA PyTorch base image (`nvcr.io/nvidia/pytorch:<YY.MM>-py3`) used by Megatron-LM CI. Covers the two pin sites (GitHub CI in `docker/.ngc_version.dev` and GitLab CI in `.gitlab/stages/01.build.yml`), the post-bump CI loop (re-run functional tests, refresh golden values, mark broken tests), and the gotchas that bit PRs
Fine-tune any HuggingFace CV / VLM / LLM model on local NVIDIA GPUs inside an NGC PyTorch container. Use when the user wants to fine-tune a HuggingFace model (full or LoRA), train a vision / VLM / LLM model end-to-end, generate a reproducible HF training pipeline, smoke-test a HuggingFace model locally before scale-up, push a fine-tuned model to the HF Hub with a model card, or emit a self-contained rerun skill for an existing HuggingFace finetune. Supports image classification, object detection, semantic / instance / panoptic segmentation, depth estimation, image-text-to-text VLM (SFT / LoRA), and LLM SFT / DPO / GRPO. Six-step workflow: inspect and qualify, hardware and NGC image, research, generate and smoke, train + eval + infer, push and emit rerun skill.
Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction. Use when training/running transformers with long sequences (>512 tokens), encountering GPU memory issues with attention, or need faster inference. Supports PyTorch native SDPA, flash-attn library, H100 FP8, and sliding window attention.
Simplest distributed training API. 4 lines to add distributed support to any PyTorch script. Unified API for DeepSpeed/FSDP/Megatron/DDP. Automatic device placement, mixed precision (FP16/BF16/FP8). Interactive config, single launch command. HuggingFace ecosystem standard.
Provides guidance for interpreting and manipulating neural network internals using nnsight with optional NDIF remote execution. Use when needing to run interpretability experiments on massive models (70B+) without local GPU resources, or when working with any PyTorch architecture.
Fine-tune Gemma 4 and 3n models with audio, images, and text on Apple Silicon using PyTorch and Metal Performance Shaders.
Generate PyTorch-style interface documentation (README.md) for AscendC operators. Trigger scenarios: Use this when interface documentation needs to be generated after compilation and debugging are completed, or when the user mentions "generate operator documentation", "create README", "document operator", "help me write documentation" (in operator context), "operator documentation".
Guides systematic PyTorch recommender-system model development across compact data facts, existing source code, configs, focused tests, and training loops without overloading context from broad research archives. Use when building, debugging, or refactoring torch/nn.Module RecSys models with Transformer/HSTU/attention blocks, sparse/dense/list feature fusion, pCVR/CTR heads, ablation axes, or competition codebases where many model ideas exist but bugs and interface drift must be controlled. 用来指导推荐系统 PyTorch 模型开发、Transformer/HSTU 建模、关键数据事实、特征交互、shape/debug、训练闭环和已有模型结构的系统化推进。