Agent
Path C Phase 2 subagent. Probe or install AISBench from source in a target container before PD deploy. Skip if already available. Writes aisbench-install-report/status.
2 11d ago A 44 tokens
Agent
A specialised agent for reproducing serving-performance baselines for a model-serving setup. It uses configuration files and reference documents to produce a usable vLLM server command when needed.
2 11d ago A 85 tokens
Agent
Path C Phase 1 subagent. Validate PD Prefill/Decode launch commands and machine/container environment; ensure one container per host when unspecified; network-only multi-host rewrites; write pd-check-report and rendered scripts.
2 11d ago A 53 tokens
Agent
Path C Phase 3 subagent. Deploy mooncakemaster then Prefill/Decode/Proxy from Phase 1 rendered scripts after AISBench is ready; on failure diagnose and signal rollback to Phase 1.
2 11d ago A 51 tokens
Agent
Path C Phase 4 subagent. Run AISBench P/D QPS tests under SLO, compute best PD ratio, fit to cluster capacity, verify deploy + e2e per-NPU throughput; write pd-ratio-report and status. P checks TTFT only; D checks TPOT only; concurrencies may differ.
2 11d ago A 77 tokens
Agent
A specialized helper for analyzing performance profiles from Ascend NPU tools, using local paths to profiling files such as ascendpt, ascendms, or PROF data.
2 11d ago A 106 tokens
Agent
A specialized agent for the second phase of serving-performance tuning, starting from a Phase 1 baseline summary. The description is in Chinese and says it performs offline parallel-strategy tuning without deployment or load testing.
2 11d ago A 86 tokens
Skill Claude CodeCodex
Probe or install AISBench from source inside a target container for Path C PD-ratio benchmarking. Runs after pd-config-env-check and before pd-deploy. Skip if aisbench is already available. Pin numpy for vLLM Ascend co-install.
2 11d ago A 55 tokens
Skill Claude CodeCodex
Finds the best vLLM-Ascend deployment config (low-latency or high-throughput) from local baseline markdown docs by matching device, model, quantization, NPU count, and deploy strategy; if no local match, fetches high-throughput configs from the official vLLM-Ascend model tutorials. Adjusts context length by…
2 11d ago A 116 tokens
Skill Claude CodeCodex
A diagnostic skill for finding fast and slow devices in Ascend cluster profiling data, where a cluster is a group of computers or accelerators working together.
2 11d ago A 127 tokens
Skill Claude CodeCodex
Analyze Ascend NPU collective communication profiling data with a DB-first workflow. Use when the user provides clusteranalysisoutput/clusteranalysis.db, rank-level analysis.db, rank-level ascendpytorchprofiler{rankid}.db, together with profilerinfo.json, and asks about HCCL or hcom communication cost, collective…
2 11d ago A 119 tokens
Skill Claude CodeCodex
Analyze Ascend NPU computation-side profiling data for single-card runs or a selected rank from multi-card runs. Use this skill when the user asks to diagnose computation bottlenecks, AI Core / AI Vector / AICPU hotspots, dynamic shape overhead, block dim issues, redundant TransData/Transpose/Cast, cross-stream waits…
2 11d ago A 82 tokens
Skill Claude CodeCodex
A tool for collecting, analyzing, and comparing information from Ascend NPU environments and dump files.
2 11d ago A 55 tokens
Skill Claude CodeCodex
A command-line analysis tool for profiling data from AI training and inference. Profiling data records how a program spends time and resources while it runs.
2 11d ago A 72 tokens
Skill Claude CodeCodex
A validation guide for performance data collected by MindStudio profiler tools. Profiling records how software uses time and resources so performance problems can be studied.
2 11d ago A 56 tokens
Skill Claude CodeCodex
A SQL analysis skill for Ascend PyTorch Profiler and msprof database files. It turns questions about operator timing, communication, scheduling, or database tables into safe SQL queries and can look up official schema details.
2 11d ago A 82 tokens
Skill Claude CodeCodex
Analyze Ascend NPU schedule, operator dispatch, operator launch, and Host Bound profiling issues in Ascend profiling data. Use when need to diagnose device Free time, framework/operator dispatch latency, launch latency, PYTORCHAPI/CANNAPI launch gaps, aclrtSynchronizeStream stalls, task queue behavior, CPU scheduling…
2 11d ago A 86 tokens
Skill Claude CodeCodex
Ascend cluster performance analysis and comparison tool. Invoke when user asks to analyze cluster profiling data (DB or TEXT format), generate cluster analysis reports, or compare two cluster datasets.
2 11d ago A 38 tokens
Skill Claude CodeCodex
Parses msprof-analyze compare output xlsx, extracts key insights per sheet, generates HTML report and Chinese xlsx. Pipeline-only for Path B after ascend-msprof-analyze-cli; do not invoke as a standalone tool.
2 11d ago A 50 tokens
Skill Claude CodeCodex
Enumerates legal DP×TP×EP parallel combinations for vLLM-Ascend single-node mixed deployment from model weight size and NPU count. Use when computing min TP, listing parallel strategy candidates, or as sub-skill of serving-parallel-strategy-tuning.
2 11d ago A 63 tokens
Skill Claude CodeCodex
A helper for reading source code, configuration, README files, Markdown, and documentation from GitHub links or repositories.
2 11d ago A 114 tokens
Skill Claude CodeCodex
Reads model feature support table (xlsx) and converts to compact JSON. Invoke when user needs to parse or convert model-feature xlsx to JSON, or query model feature support across versions.
2 11d ago A 42 tokens
Skill Claude CodeCodex
Installation and setup instructions for connecting msprof-mcp, a server that lets coding tools analyse Ascend performance-profile data, to tools such as Cursor or Claude Code.
2 11d ago A 117 tokens
Skill Claude CodeCodex
A calculator for MFU, or Machine FLOP Utilization, which measures how much of a machine's theoretical calculation capacity an operation uses.
2 11d ago A 38 tokens