Skills

Chapter 31 of 37

Finetuning

Fine-tune models on Microsoft Foundry using SFT (supervised), DPO (preference), or RFT (reinforcement with graders).

2 minutes · 368 words · 8 sections

Fine-Tuning on Microsoft Foundry

Fine-tune models using SFT (supervised), DPO (preference), or RFT (reinforcement with graders). Covers dataset prep, training, deployment, and evaluation.

When to Use

Use this sub-skill when the user asks about:

Fine-tuning a model (SFT, DPO, or RFT)
Preparing, validating, or formatting training data
Submitting, monitoring, or diagnosing training jobs
Calibrating graders or pass thresholds for RFT
Deploying or evaluating a fine-tuned model
Choosing between training types (SFT vs DPO vs RFT)
Distillation, synthetic data generation, or dataset quality scoring
Large file uploads for training data
Cleaning up fine-tuning resources (files, deployments)

Do NOT use for: General model deployment without fine-tuning (use deploy-model), agent creation (use agents), prompt optimization without training (use prompt-optimizer).

Workflows

Stage	Guide
Quick start	workflows/quickstart.md (opens in a new tab)
Full pipeline	workflows/full-pipeline.md (opens in a new tab)
Create data	workflows/dataset-creation.md (opens in a new tab)
Iterate	workflows/iterative-training.md (opens in a new tab)
Diagnose	workflows/diagnose-poor-results.md (opens in a new tab)

References

Topic	File
SFT vs DPO vs RFT	references/training-types.md (opens in a new tab)
Hyperparameters	references/hyperparameters.md (opens in a new tab)
Data formats	references/dataset-formats.md (opens in a new tab)
Grader design (RFT)	references/grader-design.md (opens in a new tab)
Reward hacking	references/reward-hacking.md (opens in a new tab)
Agentic RFT (tools)	references/agentic-rft.md (opens in a new tab)
Deployment	references/deployment.md (opens in a new tab)
Training curves	references/training-curves.md (opens in a new tab)
Evaluation	references/evaluation.md (opens in a new tab)
Vision fine-tuning	references/vision-fine-tuning.md (opens in a new tab)
Large file uploads	references/large-file-uploads.md (opens in a new tab)
Platform gotchas	references/platform-gotchas.md (opens in a new tab)

Scripts

Script	Purpose
`scripts/submit_training.py`	Submit SFT/DPO/RFT jobs
`scripts/monitor_training.py`	Poll job until completion
`scripts/calibrate_grader.py`	Find optimal RFT pass_threshold
`scripts/check_training.py`	Analyze curves, list checkpoints
`scripts/deploy_model.py`	Deploy via ARM REST API
`scripts/evaluate_model.py`	LLM judge evaluation
`scripts/convert_dataset.py`	Convert between SFT/DPO/RFT formats
`scripts/generate_distillation_data.py`	Generate synthetic training data
`scripts/score_dataset.py`	Quality scoring on training data
`scripts/cleanup.py`	Delete old files and deployments
`scripts/validate/`	Data validators (SFT, DPO, RFT) + stats

Rules

Always baseline first — evaluate the base model before fine-tuning
Validate data before submitting — run scripts/validate/validate_sft.py
Calibrate RFT graders — target 25-50% failure rate on the base model
Evaluate checkpoints — don’t blindly deploy the final one
Measure token cost alongside accuracy when comparing models

Quick Reference

Task	Command
Validate SFT data	`python scripts/validate/validate_sft.py data.jsonl`
Submit SFT job	`python scripts/submit_training.py --model gpt-4.1-mini --training-file train.jsonl --validation-file val.jsonl --type sft`
Monitor job	`python scripts/monitor_training.py --job-id ftjob-xxx`
Analyze curves	`python scripts/check_training.py --job-id ftjob-xxx`
Deploy model	`python scripts/deploy_model.py --model-id ft:gpt-4.1-mini:... --name my-eval`
Evaluate model	`python scripts/evaluate_model.py --deployment-name my-eval --test-file test.jsonl`

Error Handling

Error	Cause	Fix
“API version not supported”	Older `openai` SDK on `/v1/` endpoint	Upgrade to `openai>=1.0`
“does not support fine-tuning with Standard TrainingType”	OSS model needs `globalStandard`	Use `--use-rest` flag or script auto-falls back
Job stuck in post-training eval	Under-provisioned tool endpoint (RFT)	Scale to S2+, enable Always On
“DeploymentNotReady” after ARM succeeds	ARM/data-plane race condition	Delete and recreate deployment, wait 5 min
Content safety block at deployment	PII-dense training data	Remove problematic document types

Install this repository

$ npx skills add microsoft/azure-skills

» /plugin marketplace add microsoft/azure-skills

Skills install per repository, not per chapter — the CLI has no documented per-skill form, so we do not print one.

About this skill

Trigger

Fine-tune models on Microsoft Foundry using SFT (supervised), DPO (preference), or RFT (reinforcement with graders). Covers dataset preparation, training job submission, deployment, and evaluation. USE FOR: fine-tune, SFT, DPO, RFT, training data, grader, distillation, fine-tuned model, training job, large file upload, calibrate grader, deploy fine-tuned model, evaluate fine-tuned model. DO NOT USE FOR: general model deployment without fine-tuning (use deploy-model), agent creation (use agents), prompt optimization without training (use prompt-optimizer).

The verbatim description from this skill’s front matter — the string an agent matches on to decide whether to load it.

Licence: MIT
Also published for: GitHub — collapsed into this chapter because the copies are the same skill.
Source: skills/microsoft-foundry/finetuning/SKILL.md

Colophon for this issue

Colophon

How this issue was made

Source

github.com/microsoft/azure-skills, branch main, last pushed 7 August 2026.

Licence

MIT — the text of every chapter is reproduced unmodified, frontmatter included, under the upstream licence.

Discovery

37 chapters found by walking the repository tree for SKILL.md, not by matching a directory convention. 6 distinct layouts observed: skills/*/SKILL.md, skills/azure-app-onboard/*/SKILL.md, skills/azure-kubernetes/*/SKILL.md, skills/microsoft-foundry/*/SKILL.md, skills/microsoft-foundry/models/deploy-model/*/SKILL.md, skills/microsoft-foundry/models/*/SKILL.md.

Duplicates collapsed

37 mirror copies folded into their canonical chapter — republished for GitHub. Copies match on content hash and on position once a leading per-agent prefix is stripped, because the same skill is routinely shipped under a dozen agent directories with a dozen different hashes.

Issue colours

Resolved from a curated brand profile — hue 251°, chroma 0.167. Two accent tones are generated per issue and each is proven against its own ground before it ships: a single accent that passes AA on both light and dark paper is arithmetically impossible.

Heading repairs

1 repair applied to this chapter so the document has one h1 and no skipped levels:

Shifted “8 headings” from h1 to h2 so the chapter title is the only h1.

Spec compliance

6 editorial notes across 3 of 37 chapters. They are printed in the margin of each chapter rather than as errors here.

Deploy — 2 notes
Prepare — 2 notes
Scaffold — 2 notes

Images

Images inside a chapter come from the upstream repository. Where the author gave no alternative text we mark the image decorative rather than inventing a description — a plausible caption we made up is worse than none for the reader who depends on it.

Marketplace

A plugin manifest is published at .claude-plugin/marketplace.json by Microsoft, declaring 1 plugin. It is read for editorial metadata only — never as the skill index, which is always the repository tree.

Signal

Install counts come from skills.sh. They measure downloads, not quality, and an unranked repository is not an unread one.

Agent surfaces

The whole issue is available as one markdown document at /microsoft/azure-skills.md, and each chapter at its own .md URL.

Publication

Set by Skills Docs from the source repository. Body text is Literata at the reader’s chosen size and measure; code is Geist Mono. Nothing on this page was written by us except this paragraph.

Fine-Tuning on Microsoft Foundry

Fine-tune models using SFT (supervised), DPO (preference), or RFT (reinforcement with graders). Covers dataset prep, training, deployment, and evaluation.

When to Use

Use this sub-skill when the user asks about:

Fine-tuning a model (SFT, DPO, or RFT)
Preparing, validating, or formatting training data
Submitting, monitoring, or diagnosing training jobs
Calibrating graders or pass thresholds for RFT
Deploying or evaluating a fine-tuned model
Choosing between training types (SFT vs DPO vs RFT)
Distillation, synthetic data generation, or dataset quality scoring
Large file uploads for training data
Cleaning up fine-tuning resources (files, deployments)

Do NOT use for: General model deployment without fine-tuning (use deploy-model), agent creation (use agents), prompt optimization without training (use prompt-optimizer).

Workflows

Stage	Guide
Quick start	workflows/quickstart.md (opens in a new tab)
Full pipeline	workflows/full-pipeline.md (opens in a new tab)
Create data	workflows/dataset-creation.md (opens in a new tab)
Iterate	workflows/iterative-training.md (opens in a new tab)
Diagnose	workflows/diagnose-poor-results.md (opens in a new tab)

References

Topic	File
SFT vs DPO vs RFT	references/training-types.md (opens in a new tab)
Hyperparameters	references/hyperparameters.md (opens in a new tab)
Data formats	references/dataset-formats.md (opens in a new tab)
Grader design (RFT)	references/grader-design.md (opens in a new tab)
Reward hacking	references/reward-hacking.md (opens in a new tab)
Agentic RFT (tools)	references/agentic-rft.md (opens in a new tab)
Deployment	references/deployment.md (opens in a new tab)
Training curves	references/training-curves.md (opens in a new tab)
Evaluation	references/evaluation.md (opens in a new tab)
Vision fine-tuning	references/vision-fine-tuning.md (opens in a new tab)
Large file uploads	references/large-file-uploads.md (opens in a new tab)
Platform gotchas	references/platform-gotchas.md (opens in a new tab)

Scripts

Script	Purpose
`scripts/submit_training.py`	Submit SFT/DPO/RFT jobs
`scripts/monitor_training.py`	Poll job until completion
`scripts/calibrate_grader.py`	Find optimal RFT pass_threshold
`scripts/check_training.py`	Analyze curves, list checkpoints
`scripts/deploy_model.py`	Deploy via ARM REST API
`scripts/evaluate_model.py`	LLM judge evaluation
`scripts/convert_dataset.py`	Convert between SFT/DPO/RFT formats
`scripts/generate_distillation_data.py`	Generate synthetic training data
`scripts/score_dataset.py`	Quality scoring on training data
`scripts/cleanup.py`	Delete old files and deployments
`scripts/validate/`	Data validators (SFT, DPO, RFT) + stats

Rules

Always baseline first — evaluate the base model before fine-tuning
Validate data before submitting — run scripts/validate/validate_sft.py
Calibrate RFT graders — target 25-50% failure rate on the base model
Evaluate checkpoints — don’t blindly deploy the final one
Measure token cost alongside accuracy when comparing models

Quick Reference

Task	Command
Validate SFT data	`python scripts/validate/validate_sft.py data.jsonl`
Submit SFT job	`python scripts/submit_training.py --model gpt-4.1-mini --training-file train.jsonl --validation-file val.jsonl --type sft`
Monitor job	`python scripts/monitor_training.py --job-id ftjob-xxx`
Analyze curves	`python scripts/check_training.py --job-id ftjob-xxx`
Deploy model	`python scripts/deploy_model.py --model-id ft:gpt-4.1-mini:... --name my-eval`
Evaluate model	`python scripts/evaluate_model.py --deployment-name my-eval --test-file test.jsonl`

Error Handling

Error	Cause	Fix
“API version not supported”	Older `openai` SDK on `/v1/` endpoint	Upgrade to `openai>=1.0`
“does not support fine-tuning with Standard TrainingType”	OSS model needs `globalStandard`	Use `--use-rest` flag or script auto-falls back
Job stuck in post-training eval	Under-provisioned tool endpoint (RFT)	Scale to S2+, enable Always On
“DeploymentNotReady” after ARM succeeds	ARM/data-plane race condition	Delete and recreate deployment, wait 5 min
Content safety block at deployment	PII-dense training data	Remove problematic document types

Finetuning

Fine-Tuning on Microsoft Foundry

When to Use

Workflows

References

Scripts

Rules

Quick Reference

Error Handling

About this skill

Trigger

Colophon

Finetuning

Fine-Tuning on Microsoft Foundry

When to Use

Workflows

References

Scripts

Rules

Quick Reference

Error Handling

Appendix 31.1–31.33

ReferencesMarkdown · 12 files

ScriptsPython · 16 files

Also bundledMarkdown · 5 files

About this skill

Trigger

Colophon

Appendix 31.1–31.33

ReferencesMarkdown · 12 files

ScriptsPython · 16 files

Also bundledMarkdown · 5 files