VectorSpaceLab/AREX-Skill - 第 4 页

tracking-sdk

Use Aim's Python SDK to create local repositories and runs, track metrics/params/media/artifacts/logs, query sequences, and validate SDK instrumentation safely.

原文语言：英语

aimet

Route AIMET install, PyTorch/ONNX quantization, GenAILab, model access, cluster/Pod, Qualcomm SDK, optimization, export, and repository tasks to focused operating guidance.

原文语言：英语

cluster-pod-workflows

Operate AIMET cluster and Pod workflows for GPU development, source builds, GenAILab runs, sync, exec, and cleanup.

原文语言：英语

genai-lab

Run and configure AIMET GenAILab LLM/VLM quantization scorecards with Torch or ONNX backends, exports, caches, and summaries.

原文语言：英语

install-and-build

Install, build, repair, and validate AIMET package environments and repository build/test workflows.

原文语言：英语

model-access-and-credentialed-evaluation

Handle AIMET model downloads, Hugging Face tokens, GitHub Actions scorecard runs, AWS/S3 checkpoints, caches, and benchmark-result comparability.

原文语言：英语

onnx-quantization

Use AIMET ONNX for QuantizationSimModel calibration, provider selection, encodings, QDQ export, graph passes, and ONNX PTQ utilities.

原文语言：英语

optimization-analysis-deployment

Use AIMET analysis, mixed precision, compression, debugging, export validation, and deployment artifact workflows after core quantization setup.

原文语言：英语

qualcomm-sdk-deployment

Prepare AIMET exports for Qualcomm AI Hub, QAIRT/QNN conversion, profiling, inference, and SDK command generation.

原文语言：英语

torch-quantization

Use AIMET Torch for model preparation, QuantizationSimModel calibration, PTQ/QAT, encodings, and Torch export workflows.

原文语言：英语

airweave

Route Airweave tasks to the most specific sub-skill: local-development, backend-api, source-connectors, frontend-dashboard, connect-widget, mcp-search, or monke-e2e.

原文语言：英语

backend-api

Operates Airweave's backend API surface for search tiers, collections, source connections, Connect sessions, browse-tree selection, sources, webhooks, usage checks, source rate limits, and streaming search debugging.

原文语言：英语

connect-widget

Operates the embeddable Connect widget iframe, client-side session UX, parent-child postMessage, OAuth popup recovery, theme updates, connection creation, folder selection, and sync progress handling.

原文语言：英语

frontend-dashboard

Operates Airweave's React dashboard: auth, organization context, API client behavior, collections, search UI, billing/admin pages, webhooks, auth providers, usage gates, and cron utilities.

原文语言：英语

local-development

Operate the Airweave local Docker stack safely: start, reuse, restart, recreate, destroy, seed environment, check ports, and recover health failures.

原文语言：英语

mcp-search

Operate the Airweave MCP search server in stdio and Streamable HTTP modes, including tool registration, collection and organization resolution, auth modes, Prometheus metrics, deployment knobs, and search-tool behavior.

原文语言：英语

monke-e2e

Operate Airweave's Monke connector E2E framework safely: discover connectors, understand runner flow, configs, auth, bongos, generated test data, and credentialed-test boundaries.

原文语言：英语

source-connectors

Implement, validate, and troubleshoot Airweave source connectors: decorators, registry metadata, auth/config schemas, browse trees, ACLs, federated search, incremental sync, and source rate limits.

原文语言：英语

aix360

Routes AIX360 explainability tasks across local black-box attribution, counterfactuals and certification, interpretable models, time-series explanations, datasets, and explanation-quality metrics.

原文语言：英语

counterfactual-and-certification

Use AIX360 for contrastive CEM/CEM-MAF explanations, black-box certification, GLANCE recourse, and order-constrained optimal-transport matching.

原文语言：英语

datasets-and-metrics

Guides AIX360 dataset acquisition, local-path validation, preprocessing contracts, explanation-quality metrics, and offline troubleshooting for HELOC, COMPAS, CDC, MEPS, Ford, Sunspots, CIFAR, MNIST, CelebA, eSNLI, and related datasets.

原文语言：英语

interpretable-models

Choose and operate AIX360's directly interpretable models, rule learners, prototypes, model differencing, teaching explanations, and optional neural interpretable methods.

原文语言：英语

local-black-box

Guide local post-hoc explanations for black-box predictors on tabular, text, image, and generic structured inputs with AIX360 0.3.0 wrappers, output checks, and local metrics.

原文语言：英语

time-series

Use AIX360 TSICE, TSLime, and TSSaliency for local numeric explanations of univariate or multivariate time-series models, including temporal perturbations, relevant history, exogenous inputs, and integrated-gradient saliency.

原文语言：英语

aizynthfinder

Use AiZynthFinder for retrosynthetic planning, configuration, route analysis, custom extensions, and focused development workflows.

原文语言：英语

configuration-and-data

Create, validate, and troubleshoot AiZynthFinder configuration files, stocks, model assets, public-data setup commands, and data-tool commands.

原文语言：英语

extension-and-development

Extend AiZynthFinder with custom hooks, stocks, scorers, policies, plugins, search algorithms, and focused development tests.

原文语言：英语

planning-workflows

Run AiZynthFinder retrosynthesis planning and one-step expansion through CLI, Python APIs, and optional notebook interfaces.

原文语言：英语

route-analysis

Inspect AiZynthFinder output files, route collections, reaction trees, route scores, clustering, images, and downstream summaries without rerunning a search.

原文语言：英语

alae

Use the ALAE repository for Adversarial Latent Autoencoder data preparation, CUDA training, checkpoint-backed generation, latent editing, and legacy metrics.

原文语言：英语

data-preparation

Routes ALAE dataset preparation, TFRecord layout validation, sample-image setup, style-mixing image layout, and face-alignment workflows.

原文语言：英语

generation

Use ALAE checkpoints for demo, image generation, reconstructions, style mixing, traversals, and principal-direction management.

原文语言：英语

metrics

Prepare and route safe ALAE FID, reconstruction FID, PPL, and LPIPS metric workflows.

原文语言：英语

training

Train ALAE and StyleALAE with configs, TFRecords, DDP, and checkpoints.

原文语言：英语

albumentations

Build, debug, serialize, and integrate Albumentations 2.x augmentation pipelines for images, masks, bboxes, keypoints, volumes, and PyTorch-style datasets.

原文语言：英语

framework-integration

Integrate Albumentations with PyTorch-style datasets, tensor conversion, optional extras, and DataLoader reproducibility.

原文语言：英语

pipeline-composition

Build, edit, validate, seed, and debug Albumentations augmentation pipelines with Compose and composition combinators.

原文语言：英语

serialization-and-reproducibility

Save, load, replay, inspect, and reproduce Albumentations augmentation pipelines with JSON/YAML serialization, ReplayCompose, applied-parameter tracking, custom Lambda mappings, deterministic seeds, and Hub caveats.

原文语言：英语

targets-and-formats

Apply Albumentations pipelines safely to images, masks, bboxes, keypoints, multiple images, volumes, 3D masks, and labels without corrupting geometry or data formats.

原文语言：英语

transform-catalog

Select and parameterize Albumentations 2D transform families, including pixel, crop, resize, geometric, dropout, mixing/domain adaptation, spectrogram, text, target support, interpolation, and fill behavior.

原文语言：英语