skilld
All skills
microsoft avatar

/finetuning

@04d245b
by microsoftmicrosoft/skills3.1k stars
351

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).

Ask your Agent

Use this Skill: https://skilld.dev/gh/microsoft/skills/finetuning

This session only. Nothing lands on disk.

referenceslarge-file-uploads.md

≈232 tokens on demand. Your agent reads this file only when SKILL.md points to it.

Large File Uploads

The standard client.files.create() silently fails on JSONL files >~150MB (Azure returns 500 during job execution). Use the chunked Uploads API:

upload = client.uploads.create(filename="data.jsonl", purpose="fine-tune", bytes=file_size, mime_type="application/jsonl")
part_ids = []
with open(filepath, "rb") as f:
    while chunk := f.read(64 * 1024 * 1024):  # 64MB chunks
        part = client.uploads.parts.create(upload_id=upload.id, data=chunk)
        part_ids.append(part.id)
completed = client.uploads.complete(upload_id=upload.id, part_ids=part_ids)
file_id = completed.file.id

Important: Requires openai.AzureOpenAI() client, NOT openai.OpenAI() with /v1/ URL. The project endpoint returns 404 for upload operations.

File Size Method
< 100MB Standard files.create()
100MB–5GB Chunked Uploads API
> 5GB Split dataset

Source: SKILL.md on GitHub

Third-party checks

2 warnings1mo3 checks · Risk SAFE
  • Gen Agent Trust Hub1mo

    This skill provides a robust toolkit for fine-tuning and evaluating models on Microsoft Foundry. It includes administrative scripts for job management and data processing. There are security considerations regarding the dynamic execution of user-supplied scripts and the invocation of the Azure CLI, which are standard for the skill's intended developer use-case.

  • Socket1mo

    1 alert: gptSecurity

  • Snyk1mo

    Risk: MEDIUM · 1 issue

Provenance

Signed by skilld at 04d245b. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub yesterday.

Activeupdated 2 months ago

Capability

metadata
{
  "author": "Microsoft",
  "version": "0.0.0-placeholder"
}

README badge

README badge for microsoft/skills/finetuning