Documentation Index
Fetch the complete documentation index at: https://docs.benchgen.com/llms.txt
Use this file to discover all available pages before exploring further.
Fine-tune a Model
Fine-tuning adapts a base model to your specific task using a LoRA (Low-Rank Adaptation) adapter. Train manages the compute — you just configure the run.Prerequisites
- A dataset uploaded or available in Train (see Upload a dataset)
- A base model selected (see supported base models below)
Steps
- Go to Train → Fine-tune → New Run.
- Select a base model from the supported model list.
- Select a dataset from your uploaded datasets.
- Configure LoRA parameters:
| Parameter | Description | Recommended starting value |
|---|---|---|
rank | LoRA rank — controls adapter capacity | 16 |
alpha | Scaling factor for the adapter | 32 |
target_modules | Which layers to adapt | q_proj, v_proj |
dropout | Regularization | 0.05 |
- Configure training hyperparameters:
| Parameter | Description | Recommended starting value |
|---|---|---|
learning_rate | Step size for gradient updates | 2e-4 |
epochs | Number of full passes over the dataset | 3 |
batch_size | Examples per gradient step | 4 |
max_seq_length | Maximum token length per example | 2048 |
- Click Start run.
- Monitor the run — loss curve and sample outputs update in real time.
- When complete, the adapter is saved and ready to merge.
Next Steps
- Merge the LoRA adapter into the base model for deployment.