skills.rest
doanchienthangdev
doanchienthangdevCOMMUNITY
Explore all skills by this author →
GitHub

ai-accelerators

Optimizes machine learning workloads on GPUs, TPUs, and edge accelerators with CUDA, TensorRT, and TFLite.

4|1|Updated Dec 30, 2025
Search Tags:#gpu optimization#tensorrt#hardware acceleration#edge ai#tpu#ai hardware#ml performance
One-click install
npx skills add https://github.com/doanchienthangdev/omgkit --skill ai-accelerators
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ai-accelerators
Source: https://github.com/doanchienthangdev/omgkit/tree/main/plugin/skills/ml-systems/ai-accelerators
Command: npx skills add https://github.com/doanchienthangdev/omgkit --skill ai-accelerators

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve?

This Skill addresses the performance bottlenecks in machine learning by detailing and optimizing the use of specialized hardware like GPUs, TPUs, and edge accelerators.

Core Features & Use Cases

  • Hardware Overview: Understand the landscape of AI accelerators (GPUs, TPUs, NPUs, FPGAs, ASICs, Edge).
  • Optimization Strategies: Learn techniques for GPU (CUDA, Tensor Cores, Multi-GPU) and TPU (JAX, PyTorch/XLA) optimization.
  • Edge Deployment: Explore optimization for edge devices using TensorRT, DeepStream, and TFLite.
  • Hardware-Aware Tuning: Implement auto-tuning with TVM and hardware selection guidance.
  • Use Case: Optimize a deep learning model for faster inference on an NVIDIA Jetson device using TensorRT, or select the most cost-effective GPU cluster for training a large language model based on budget and performance requirements.

Quick Start

Use the ai-accelerators skill to find optimization strategies for running PyTorch models on NVIDIA GPUs.

Frequently Asked Questions about ai-accelerators

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I optimize deep learning models for faster inference on edge devices?▼

Edge deployment optimization uses tools like TensorRT, DeepStream, and TFLite to accelerate machine learning inference on edge devices. These frameworks convert and tune models to leverage specific edge hardware capabilities for lower latency execution.

What's the best way to run PyTorch models on NVIDIA GPUs using Tensor Cores?▼

GPU optimization for PyTorch models involves CUDA memory management and Tensor Core utilization to maximize throughput on NVIDIA GPUs. Multi-GPU strategies distribute workloads across multiple cards to scale performance for demanding training tasks.

How does TPU programming with JAX and PyTorch/XLA differ from GPU acceleration?▼

TPU programming uses JAX and PyTorch/XLA to execute workloads on tensor processing units, differing from GPU acceleration by utilizing specialized matrix compute units. This approach optimizes large-scale tensor operations differently than standard CUDA pipelines.

How do I select the most cost-effective GPU cluster for training a large language model?▼

Hardware selection matrices evaluate budget and performance requirements to identify the most cost-effective GPU clusters for large language model training. Hardware-aware auto-tuning with TVM further refines model execution to match specific cluster capabilities.

Can I use auto-tuning to improve ML performance across different AI hardware accelerators?▼

Auto-tuning improves ML performance by dynamically adjusting model parameters to match the specific capabilities of AI hardware accelerators. Hardware-aware optimization through TVM ensures efficient execution across diverse architectures like GPUs, TPUs, and FPGAs.

What AI accelerators are available for specialized machine learning workloads?▼

AI accelerators for specialized machine learning workloads include GPUs, TPUs, NPUs, FPGAs, and ASICs. Understanding this hardware landscape enables targeted optimization strategies for maximum inference and training efficiency.

Related Skills

Explore More Software Engineering Skills→
agentic-inagentic-in
577

Inference

Select inference paths, serving stacks, and benchmarking plans for ML models.

Official
Intermediate
michaelalbermichaelalber
2

model-optimization

Quantize, prune, and convert ML models to TensorRT, TFLite, and ONNX.

Community
Advanced
alibabaalibaba
71

gpu-kernel-bottleneck-analysis

Analyzes GPU kernel performance using Roofline models and profiling data.

Official
Advanced
tangzheng202202tangzheng202202
1

modal-serverless-gpu

Deploy GPU-based ML workloads on a serverless cloud platform.

Community
Advanced