Finetuning
Here we describe how to finetune LLMs and load previously trained models
Finetuning Models
Once you have a prepared dataset, you are ready to create a new finetune job on the Anarchy platform.
Kickstart the Process
From your dataset page, click on the + New Finetune button at the top right corner of the screen. This will take you to the finetune job page.
Label Your Job
Next, name your finetune job. Choose something unique that you can reference later.
Select Your Base Model
Use the dropdown to choose your model.
Adjust the Advanced Settings
Tweak these settings to hone in your finetuning job even further.
Epochs
The number of times the entire dataset is passed through the model during training.
Learning Rate
The step size used for updating the model’s parameters during training.
Batch Size
The number of training samples used in one iteration to update the model’s parameters.
Checkpoints
Saving the model’s state at regular intervals during training for later use or recovery.
Cost and Credits
Finetuning is a relatively compute-intensive process, and the cost will scale with the size of your dataset and the number of Epochs you will train.
On the default Pay-as-you-go plan, 1 credit is equivalent to $1.00USD. For more information about pricing and volume discounts, contact us below or visit our pricing page.
Contact Sales
Get in touch for custom guidance on productionizing finetuned LLMs.