who who who
Community@fsx950223 · China
interest in deep learning
Agent Skills by who who who
Showing 12 vetted skills indexed across 1 GitHub repositories.
fsx950223build-rocm-image
Build a custom Docker image with ROCm tools on a remote host via SSH.
fsx950223FlyDSL Kernel Authoring Skill
Compile Python DSL GPU kernels for AMD hardware via MLIR.
fsx950223build-flydsl
Build and install FlyDSL with LLVM/MLIR on remote hosts or Docker containers.
fsx950223bisect-perf-regression
Automate Git bisect to find GPU kernel performance regression commits.
fsx950223Commit and Push Staged Changes
Commit staged Git changes and push to the current branch with rebase conflict resolution.
fsx950223review-skills
Review SKILL.md files and usage logs to correct inaccuracies and inconsistencies.
fsx950223kernel-trace-analysis
Profile GPU kernels with rocprofv3 and generate optimization plans.
fsx950223lds-optimization
Diagnose LDS bank conflicts and latency stalls from GPU kernel trace data.
fsx950223set-git-account
Configure global Git user name and email for commits.
fsx950223prefetch-data-load
Apply software prefetching to Triton/Gluon kernel loops to overlap data load latency with compute.
fsx950223optimize-pa-decode-gluon
Optimize Triton/Gluon paged attention decode kernels for AMD GPUs.
fsx950223agent-status-report
Generate a timestamped .txt status report of all agent tasks.