Skip to main content

Fine-Tuning Guide

This step-by-step tutorial walks through configuring and launching a training job.

Prerequisites

  • Workspace with at least one dataset.
  • Role permission: fine-tune.launch.
  • Understanding of dataset structure (see Datasets).

Step 1: Choose Base Model

  1. Open Fine-Tune Studio.
  2. Review recommended models. The panel shows context length, latency expectations, and suitable strategies.
  3. Select the foundation model. Record the model ID for future reference.

Step 2: Dataset Options

  1. Select Uploaded Dataset or connect external sources.
  2. Optionally blend multiple datasets:
    • Assign weights (e.g., support_chat:0.7, policy_docs:0.3).
    • Preview the combined distribution.
  3. Confirm the final dataset configuration.

Step 3: Configure Training

Key settings:
SettingDescription
Training methodLoRA, QLoRA, or full fine-tune
Learning rateDefault is 2e-4 for LoRA; adjust for specific datasets
Batch sizeAdaptive; suggested ranges appear based on tokens per row
EpochsTypically 2–5 for LoRA; validate with evaluation metrics
Evaluation cadenceFrequency of validation runs
Safety guardrailsEnable automatic redaction and safety classifiers
Logging integrationsOptional: send metrics to Slack, DataDog, or webhook

Step 4: Launch and Monitor

  • Confirm the cost estimate.
  • Click Launch Training.
  • Monitor progress in real time:
    • Loss curves
    • Tokens/sec
    • Epoch completion
    • GPU allocation events
  • You can pause or cancel if necessary.

Step 5: Review Results

When the run completes:
  1. Inspect evaluation metrics and sample outputs.
  2. Use the Evaluate feature to test your model:
    • Single prompt evaluation
    • Comparison with base model
    • Batch evaluation for multiple prompts
    • Results dashboard for comprehensive analysis
  3. Compare with previous runs using the comparison view.
  4. Archive the run or mark as ready for deployment.
For detailed evaluation instructions, see the Evaluate Guide.

Troubleshooting

  • Job stuck in pending: Check resource availability or contact support if it exceeds expected queue times.
  • Spiking loss: Inspect dataset balance; consider re-weighting or cleaning data.
  • Unexpected outputs: Revisit dataset quality and ensure conversation turns are labeled correctly.