Results for “transformer-engine”

Validate and use CUDA graph capture in Megatron Bridge, including local full-iteration graphs and Transformer Engine scoped graphs for attention, MLP, and MoE modules.

Load pre-trained models from Hugging Face Hub, run pipeline inference, generate text, and fine-tune models on NLP, vision, audio, and multimodal tasks using the Transformers library.

Resolves legacy references to the conversion-engineer capability by routing to the current runtime agent or skill.

Inspect and manipulate transformer internals via HookPoints and activation caching for mechanistic interpretability research.

Turbine — small Kotlin testing library for kotlinx.coroutines Flows. Provides ergonomic API to test Flow emissions deterministically: awaitItem, expectMostRecentItem, awaitComplete, awaitError. Works with StateFlow, SharedFlow, Channel-backed flows, combine/map/debounce. KMP-friendly. USE WHEN: user mentions "Turbine", "app.cash.turbine", ".test {}", "awaitItem", "Flow testing", "StateFlow test", "SharedFlow test", "expectMostRecentItem", "cancelAndIgnoreRemainingEvents" DO NOT USE FOR: Mobile E2E - use `testing/maestro` DO NOT USE FOR: Compose snapshot - use `testing/compose-snapshot` DO NOT USE FOR: Generic Kotlin testing - use `testing/kotest` DO NOT USE FOR: Suspend function (non-Flow) testing - use `kotlinx-coroutines-test` directly

Train, evaluate, export, distill, quantize, or run inference for a TAO DINO 2D object detector using transformer-based detection with denoising training and multi-scale features.

Trains vision transformers via self-distillation without labels for self-supervised visual representation learning, and supports export and inference of NVDINOv2 backbones.

Run state-of-the-art machine learning models directly in JavaScript/TypeScript across browsers and server-side runtimes using Transformers.js.

Train, evaluate, export, and run inference for Masked Auto-Encoder (MAE) models for self-supervised pretraining and fine-tuning of visual representations.

Train, evaluate, distill, quantize, export, and run inference for RT-DETR object detection models using NVIDIA TAO.

Trains, evaluates, exports, quantizes, and runs inference for SegFormer semantic segmentation models using NVIDIA TAO.

Train or fine-tune sentence-transformers models for retrieval, similarity, clustering, classification, and reranking, with support for bi-encoders, cross-encoders, and sparse encoders.

Train, evaluate, export, quantize, and run inference for a Deformable DETR 2D object detection model using TAO, with deformable attention for efficient multi-scale feature processing.

Train, evaluate, export, quantize, and run inference on Mask2Former models for panoptic, instance, and semantic segmentation using NVIDIA TAO.

Guides feature extraction and preprocessing for time series data using aeon transformers, covering collection and series transformers with code examples.

Educational GPT implementation in ~300 lines. Reproduces GPT-2 (124M) on OpenWebText. Clean, hackable code for learning transformers. By Andrej Karpathy. Perfect for understanding GPT architecture from scratch. Train on Shakespeare (CPU) or OpenWebText (multi-GPU).

Automates repetitive infrastructure and deployment tasks with Terraform, OpenTofu, Ansible, GitHub Actions, and n8n, enforcing safe practices like plan review and confirmation before applying changes.

Provides the shared STORM methodology, artifact layout, stage-gating contract, citation hygiene, and retrieval fallback. Use when executing any /storm:* skill (generate, research, outline, write, polish). Internal knowledge; never user-invocable.

Train Mixture of Experts (MoE) models using DeepSpeed or HuggingFace. Use when training large-scale models with limited compute (5× cost reduction vs dense models), implementing sparse architectures like Mixtral 8x7B or DeepSeek-V3, or scaling model capacity without proportional compute increase. Covers MoE architectures, routing mechanisms, load balancing, expert parallelism, and inference optimization.

Educational GPT implementation in ~300 lines. Reproduces GPT-2 (124M) on OpenWebText. Clean, hackable code for learning transformers. By Andrej Karpathy. Perfect for understanding GPT architecture from scratch. Train on Shakespeare (CPU) or OpenWebText (multi-GPU).

Train Mixture of Experts (MoE) models using DeepSpeed or HuggingFace. Use when training large-scale models with limited compute (5× cost reduction vs dense models), implementing sparse architectures like Mixtral 8x7B or DeepSeek-V3, or scaling model capacity without proportional compute increase. Covers MoE architectures, routing mechanisms, load balancing, expert parallelism, and inference optimization.

Orchestrate altering an existing, working feature to new desired behavior by updating tests first, then changing implementation, reviewing, and committing with gated approval.

Large Language Model development, training, fine-tuning, and deployment best practices.

Train Mixture of Experts (MoE) models using DeepSpeed or HuggingFace, covering architectures, routing, load balancing, and expert parallelism.

Educational GPT implementation in ~300 lines. Reproduces GPT-2 (124M) on OpenWebText. Clean, hackable code for learning transformers. By Andrej Karpathy. Perfect for understanding GPT architecture from scratch. Train on Shakespeare (CPU) or OpenWebText (multi-GPU).

PyTorch, TensorFlow, neural networks, CNNs, transformers, and deep learning for production

Parameter-efficient fine-tuning for LLMs using LoRA, QLoRA, and 25+ methods. Use when fine-tuning large models (7B-70B) with limited GPU memory, when you need to train <1% of parameters with minimal accuracy loss, or for multi-adapter serving. HuggingFace's official library integrated with transformers ecosystem.

Configure Unreal Engine core settings — rendering, physics, audio, garbage collection, console variables (cvars), and scalability levels (EngineSettingsService). Use when the user asks to change engine/rendering settings, set a console variable (r.*, etc.), adjust scalability/quality levels, or read engine config categories.

Expert skill for building web-based game engines and games using HTML5, Canvas, WebGL, and JavaScript. Use when asked to create games, build game engines, implement game physics, handle collision detection, set up game loops, manage sprites, add game controls, or work with 2D/3D rendering. Covers techniques for platformers, breakout-style games, maze games, tilemaps, audio, multiplayer via WebRTC, and publishing games.

Multimodal Learning with Transformers: A Survey

Train, evaluate, export, prune, and run inference for PointPillars 3D object detection models from LiDAR point clouds using NVIDIA TAO.

Train and fine-tune transformer language models using TRL (Transformers Reinforcement Learning) with support for SFT, DPO, GRPO, KTO, RLOO, and reward model training via CLI commands.

Add, remove, edit, list, or change the boot default of nvpmodel power modes on Jetson/Tegra platforms (Orin, Thor) by modifying the BSP-side configuration file.

Train LLMs with reinforcement learning using verl (Volcano Engine RL), supporting RLHF, GRPO, PPO, and other algorithms for scalable post-training with flexible infrastructure backends.