Ainisa AI documentation

Appearance

⚙️ Fine-tuning ​

After preparing your datasets and creating your AI file, the final step is to fine-tune your model.
Fine-tuning allows you to train an existing OpenAI model (like GPT-4 or GPT-3.5) on your own data to create a custom AI model optimized for your business.

📋 Overview ​

The Fine-tuning section displays all your previous training jobs and their statuses.
Each job represents a model training process that uses one of your .jsonl AI files.

You can access this section here:
👉 https://ainisa.com/business/fine-tunes

📊 Fine-tuning Jobs List ​

In this section, you’ll see a list of all fine-tuning jobs you’ve created.
Each job has a status that shows its current progress:

StatusDescription
Validating FileThe system is checking your .jsonl file before training starts.
RunningThe fine-tuning process is in progress.
SucceededThe training has completed successfully, and your new model is ready.
FailedThe training failed due to a data or model issue.
CancelledThe process was manually stopped.

Click any Job ID to see detailed information about that job.

Fine-tuning list and job details

🔍 Job Details ​

When you click a Job ID, you’ll see full details, including:

  • Status – current state of the training job.
  • Model – the base model used for fine-tuning.
  • Fine-tuned Model – the name of your newly created model.
  • Created / Finished At – timestamps for when the job started and ended.
  • Trained Tokens – total number of tokens used during training.
  • AI File for Training – the .jsonl file used for this fine-tune.
  • Organization ID – your OpenAI organization identifier.
  • Hyperparameters – training parameters like Epochs, Batch size, and Learning rate multiplier.

🚀 Creating a New Fine-tuning Job ​

To create a new fine-tuning job:

  1. Go to Create Fine-tuning Job
  2. Select Fine-tuning Method
    • Choose between Supervised or Direct Preference Optimized (DPO).
  3. Select AI File for Training
    • Pick the .jsonl file you created earlier in the AI Files section.
  4. Select Base Model
    • Choose the model you want to fine-tune (for example, gpt-4o-mini, gpt-4, etc.).
    • You can also select one of your previously trained models as the base.
  5. Click Save to start the fine-tuning process.

Create fine-tuning job

📈 After Fine-tuning ​

  • Once the job succeeds, a new fine-tuned model will appear in your AI Models section.
  • You can assign that custom model to your assistants and use it in chats directly.
  • The fine-tuned model name will look like:
    ft:gpt-4o-2024-08-06:ainisa:yourmodelname

🧠 Notes ​

  • The number of monthly fine-tuning jobs is limited based on your subscription plan.
  • Only OpenAI models currently support fine-tuning on Ainisa.
  • Users cannot delete fine-tuning jobs from the platform directly.
    To remove a job, please contact Ainisa support and request deletion.
  • Make sure you have enough datasets and a valid .jsonl file before starting a fine-tune.
  • Each fine-tuning job consumes API tokens from your connected OpenAI account.

✅ Tip: Learn more about OpenAI fine-tuning best practices: ✅ Tip: For more details about fine-tuning methods, check out: