Performance skills
Free performance skills for AI coding agents, part of our developer tools collection.
40 skills
Customize Clocks
nvidia
Control CPU, GPU, and EMC clock settings for Jetson devices.
DeepStream Profiling
nvidia
Efficiently profile and tune DeepStream pipelines.
Accelerated Computing cuDF
nvidia
Efficient GPU DataFrame operations for pandas users.
Adaptive Tuning for NeMo Relay
nvidia
Optimize adaptive plugin behavior for NeMo Relay.
TP DP PP Communication Overlap
nvidia
Optimize Megatron-Bridge for efficient communication overlap.
Megatron FSDP
nvidia
Optimize distributed training with Megatron FSDP.
Jetson Speculative Decoding
nvidia
Enhance token generation speed on Jetson vLLM servers.
Jetson Memory Audit
nvidia
Monitor and verify memory usage on Jetson devices.
Jetson LLM Benchmark
nvidia
Benchmark LLM performance on Jetson devices with JSON output.
DOCA GPUNetIO ib_write_lat
nvidia
Measure GPU kernel RDMA WRITE latency accurately.
DOCA GPUNetIO ib_write_bw
nvidia
Measure sustained RDMA WRITE bandwidth with GPU support.
DOCA Flow Tune
nvidia
Optimize and analyze your DOCA Flow pipelines effectively.
DOCA Flow Perf
nvidia
Measure the performance of DOCA Flow pipelines accurately.
DOCA Flow DPA Perf
nvidia
Measure DPA-offloaded Flow performance on supported hardware.
DOCA Bench Extension
nvidia
Enhance doca-bench with custom plug-ins for specific workloads.
DOCA Bench
nvidia
Measure DOCA library performance reliably and reproducibly.