cfregly

SLA-first GPU capacity sizing for a serving deployment: given a tokens-per-minute (TPM) target AND the interactivity SLA (output tokens/s/user), compute the pods and GPUs needed from a model's measured tok/s/user-vs-concurrency curve. Sizing MUST start from…

Measure the low-concurrency (c=1..c=8) decode hot-path of a live vLLM pod FAST and CORRECTLY: where each token's time actually goes (GPU-busy vs host-idle gap vs comm), whether the workload is kernel-bound / host-bound / comm-bound, and what the addressable…

Capture per-kernel CUDA hardware-counter data (occupancy, achieved warps active, regs/thread, smem/block, DRAM throughput, arithmetic intensity, branch divergence %, warp-stall reason %) from a live vLLM inference pod via an `ncu` debug sidecar (no production…

Capture per-kernel CUDA profile data from a live vLLM inference pod via an nsys debug sidecar (no production image rebuild). Outputs `.nsys-rep` + summary CSV + top-kernels table that joins with the zymtrace per-kernel breakdown and the inference-perf-bench…

End-to-end orchestrator that takes a NEW model from a bare HuggingFace id to a perf-lake-published, validated CROSS-ENGINE (vLLM + SGLang) champion on B200/GB300. Scaffolds a per-model harness + evidence bundle (run-id == experiment-id), then drives…

Canonical inference perf-bench skill (formal name. The colloquial alias is `ai-bench` - identical behaviour). Drives NVIDIA AIPerf + the replay-playback dataset against an in-cluster vLLM endpoint to measure TTFT, ITL, throughput, tok/s/user, request latency,…

Applies Jeff Dean and Sanjay Ghemawat's Performance Hints to GPU inference. Builds a back-of-the-envelope cost ledger, identifies the hot-path class, bounds each optimization with profile share, and routes the highest-value hypothesis to the repo's benchmark,…

Search a vLLM serving model's config space for its best inference performance: a `perftunereport campaign_run` matrix sweep over concurrency x max_num_batched_tokens x kv-cache-dtype x cudagraph_mode, plus a same-node controlled A/B (`run-controlled-ab.sh`)…

Scaffold a new evidence bundle directory ready for reproducibility-grade evidence capture: SOURCE.md (operator + cluster + git SHA + UTC timestamp), summary.md (verdict skeleton), commands/ (for the four-file .cmd/.stdout/.stderr/.exit tuple capture per shell…

SLA-first GPU capacity sizing for a serving deployment: given a tokens-per-minute (TPM) target AND the interactivity SLA (output tokens/s/user), compute the pods and GPUs needed from a model's measured tok/s/user-vs-concurrency curve. Sizing MUST start from…

Measure the low-concurrency (c=1..c=8) decode hot-path of a live vLLM pod FAST and CORRECTLY: where each token's time actually goes (GPU-busy vs host-idle gap vs comm), whether the workload is kernel-bound / host-bound / comm-bound, and what the addressable…

Capture per-kernel CUDA hardware-counter data (occupancy, achieved warps active, regs/thread, smem/block, DRAM throughput, arithmetic intensity, branch divergence %, warp-stall reason %) from a live vLLM inference pod via an `ncu` debug sidecar (no production…

Capture per-kernel CUDA profile data from a live vLLM inference pod via an nsys debug sidecar (no production image rebuild). Outputs `.nsys-rep` + summary CSV + top-kernels table that joins with the zymtrace per-kernel breakdown and the inference-perf-bench…

End-to-end orchestrator that takes a NEW model from a bare HuggingFace id to a perf-lake-published, validated CROSS-ENGINE (vLLM + SGLang) champion on B200/GB300. Scaffolds a per-model harness + evidence bundle (run-id == experiment-id), then drives…

Bridge between the inference-perf-bench / ai-bench output bundle and the profile-and-optimize perf-baseline registry. Knows the canonical inference perf metric shape (TTFT p50/p95/p99, ITL p50, throughput, tok/s/user, request latency p50, prefix cache hit…

Canonical inference perf-bench skill (formal name. The colloquial alias is `ai-bench` - identical behaviour). Drives NVIDIA AIPerf + the replay-playback dataset against an in-cluster vLLM endpoint to measure TTFT, ITL, throughput, tok/s/user, request latency,…

Human-first editing standard for README.md files and user-facing Markdown docs. Use when Codex needs to rewrite, review, or gate public docs so humans see the value story first, links are easy to scan, modern GitHub Markdown components are used tastefully,…

Review an agent's tool inventory, tool calls, reasoning summaries, tool outputs, final answers, and value-bar proof using this repo's trace-review harness. Use when asked to audit, test, score, or review agent traces, Claude Messages API transcripts, tool…

Evidence-first performance review and optimization workflow adapted from Jeff Dean and Sanjay Ghemawat's Performance Hints for this PyTorch, CUDA, Triton, and distributed benchmark repository. Use for hot-path reviews, latency or throughput regressions,…

Closed-loop speculative-decoding-as-a-service orchestrator -- a self-hosted analog of Fireworks FireOptimizer's adaptive speculative execution. Composes the per-phase skills into one profile-matched loop: profile live traffic (inference-workload-profile) ->…