XiaHan
Community@jstzwj · SuZhou,China
Learning to learn
Agent Skills by XiaHan
Showing 23 vetted skills indexed across 1 GitHub repositories.
jstzwjxla
Compile and optimize machine learning models for CPUs, GPUs, and TPUs.
jstzwjdeepspeed
Optimize large-scale deep learning training with ZeRO and parallelism techniques.
jstzwjnsight
Profile GPU and CPU performance with timeline visualization and hardware counters.
jstzwjtile-ir
Generate tile-centric GPU kernels for NVIDIA tensor core execution.
jstzwjnccl
Provide collective operations for multi-GPU data communication in CUDA applications.
jstzwjonnxruntime
Run ONNX model inference and training with hardware acceleration.
jstzwjcuda
Write CUDA kernels for NVIDIA GPU computing with memory management.
jstzwjtriton
Create and autotune GPU kernels in Python with MLIR-based compilation.
jstzwjvllm
Deploy large language models across GPU clusters with distributed serving and OpenAI-compatible APIs.
jstzwjmlir
Create, transform, and optimize MLIR modules for AI frameworks and compiler backends.
jstzwjtensorflow
Creates ML models with TensorFlow including optimization and deployment workflows.
jstzwjsglang
Orchestrate large model deployment with GPU optimization and distributed serving.
jstzwjtilelang
Generates GPU kernels for CUDA, HIP, Metal, and CPU via Python.
jstzwjflash-attention
Implement optimized attention kernels for PyTorch and CUDA GPUs.
jstzwjpytorch
Create, optimize, and deploy PyTorch deep learning models.
jstzwjjax
Run numerical routines, automatic differentiation, and hardware acceleration with JAX.
jstzwjMegatron-LM
Train and deploy large transformer models with tensor, pipeline, and expert parallelism.
jstzwjcuTile
Define immutable tiles for GPU kernels with automatic memory management.
jstzwjcutlass
Execute architecture-tuned GPU matrix multiplication kernels via NVIDIA CUTLASS.
jstzwjtvm
Import models from PyTorch, ONNX, and TFLite into Relax IR for optimization.
jstzwjray
Scale AI workflows across distributed systems with Ray.
jstzwjxformers
Provide optimized attention and sparse tensor operations for Transformer models.
jstzwjbitsandbytes
Quantize weights and optimizer states for PyTorch model training.