opensite-ai/opensite-skills / aws-ai-ml
aws-ai-ml
VerifiedSelects, deploys, and customizes AI models on Amazon SageMaker. Training or Processing jobs, fine-tuning (SFT/DPO/RLVR/RLAIF), model selection, dataset preparation, evaluation, SageMaker or Bedrock deployment, inference optimization and endpoint diagnostics. Covers the full lifecycle from planning through production. Use when fine-tuning models on SageMaker, choosing which base model to customize, fine-tune, or deploy from SageMaker JumpStart or Hub, SageMakerPublicHub, or the SageMaker public model catalog, transforming or validating training data, evaluating model quality, deploying or optimizing endpoints, configuring IAM/S3 for training, or managing SageMaker Managed MLflow. Use for endpoint health, failures, latency, logs, metrics, errors. Covers Serverless Model Customization, Nova and OSS deployment paths, and PySDK v3. NOT for Ground Truth labeling, Feature Store, or general-purpose AWS infrastructure.
references/training-jobs/references/configuration.md
View on GitHub# Training Template Configuration
All templates support notebook and script mode. Notebook-only dependency lines are marked `# NOTEBOOK_ONLY`. Fill every bracketed placeholder before running.
## Common values
| Placeholder | Expected value | Example |
|---|---|---|
| `[PROJECT_DIR]` | Confirmed project root | `./fraud-training` |
| `[ROLE_ARN]` | SageMaker execution role ARN; leave empty only when `get_execution_role()` works | `arn:aws:iam::123456789012:role/SageMakerExecutionRole` |
| `[TRAINING_S3_URI]` | Same-region S3 prefix/object for the training channel | `s3://my-sagemaker-bucket/train/` |
| `[OUTPUT_S3_URI]` | S3 prefix for model artifacts | `s3://my-sagemaker-bucket/output/` |
| `[INSTANCE_TYPE]` | Training instance type verified against quota | `ml.m5.4xlarge` |
| `[BASE_JOB_NAME]` | Human-readable job prefix; the template adds a UUID fragment | `fraud-xgboost` |
Optional verified settings shared by the templates:
- `OUTPUT_KMS_KEY_ID`: KMS key ID/ARN for output artifacts; execution role and caller need the corresponding permissions.
- `CHECKPOINT_S3_URI`: managed checkpoint destination. Required for the custom Spot example; optional for recipe/JumpStart when the selected algorithm supports checkpoints.
- `SUBNETS` and `SECURITY_GROUP_IDS`: provide both to place training containers in a VPC.
- `ENABLE_NETWORK_ISOLATION`: disables container network access; verify required downloads/endpoints remain available.
- `ENABLE_INTER_CONTAINER_TRAFFIC_ENCRYPTION`: enables encryption between distributed training containers and can affect throughput.
These are a bounded, SDK-verified subset rather than an attempt to expose every `CreateTrainingJob` field.
## Recipe template
| Placeholder | Expected value |
|---|---|
| `[RECIPE_NAME_OR_YAML]` | Published SageMaker recipe name or confirmed local YAML path |
Find published names and schemas in the [SageMaker HyperPod recipes catalog](https://github.com/aws/sagemaker-hyperpod-recipes/). `Compute.instance_type` is required. The base recipe plus `RECIPE_OVERRIDES` derives source/image/distributed/compute/hyperparameter settings; explicit networking/output/checkpoint/stopping inputs are applied by `ModelTrainer`. Review `trainer.get_resolved_recipe()` before approval.
`TRAINING_PLAN_ARN` is optional. Populate it only after verifying that the plan matches the requested compute and schedule. `INSTANCE_TYPE` is the required user-confirmed compute input. Leave `INSTANCE_COUNT` and `VOLUME_SIZE_GB` as `None` to preserve recipe/SDK defaults; populate them only when the user explicitly requests an override. Set `TRAINING_IMAGE_URI` only when the selected recipe/provider documentation requires it, using an exact supported tag or digest; recipe requirements may change. Recipe handling has no `accept_eula` parameter; confirm any recipe/provider license prerequisites conversationally.
## JumpStart template
| Placeholder | Expected value |
|---|---|
| `[JUMPSTART_MODEL_ID]` | Exact Hub model ID returned by `model-selection`/Hub discovery |
Pin `MODEL_VERSION` when reproducibility matters. Leave both compute override fields unset to use JumpStart defaults. Set either or both deliberately to override only those values; `OVERRIDE_INSTANCE_COUNT=None` avoids silently forcing a count of one. Set `ACCEPT_EULA=True` only after the user explicitly accepts the gated model license.
## Custom template
| Placeholder | Expected value |
|---|---|
| `[TRAINING_IMAGE_URI]` | Versioned or digest-pinned ECR training image URI; clear when using `ALGORITHM_NAME` |
| `[ALGORITHM_NAME]` | Marketplace algorithm name/ARN; set only when `TRAINING_IMAGE` is empty |
| `[SOURCE_DIR]` | Local/S3 source directory; may be empty for a self-contained image/algorithm |
| `[ENTRY_SCRIPT]` | Script path relative to source directory; required when source directory is set |
| `[REQUIREMENTS]` | Requirements path relative to source directory; optional |
Exactly one of `TRAINING_IMAGE` and `ALGORITHM_NAME` must be non-empty. `SOURCE_DIR` and `ENTRY_SCRIPT` must be provided together. Never silently substitute an unverified `latest` image tag.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deployment | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy. | Stars: 119k | Updated | Claude CodeCodex | Verified | Install |
| childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release. | Stars: 118k | Updated | Claude CodeCodex | Verified | Install |