<![CDATA[fine tuning and LoRA Fine tuning]]>https://fine-tuning-and-lora-fine-tuning.hashnode.devRSS for NodeFri, 19 Jun 2026 21:45:07 GMT60<![CDATA[Fine Tuning And LoRA Fine Tuning #chaiaurcode]]>https://fine-tuning-and-lora-fine-tuning.hashnode.dev/fine-tuning-and-lora-fine-tuning-chaiaurcodehttps://fine-tuning-and-lora-fine-tuning.hashnode.dev/fine-tuning-and-lora-fine-tuning-chaiaurcodeTue, 15 Apr 2025 15:58:55 GMTFine Tuning

Swaad Anusaar

Fine tuning is pretrained model to use for our purpose.

How fine tuning made -

  • Internet Data

  • Knowledge cutoff

In this figure the How actually fine tuning work.

Fine Tuning Types

  1. Full Parameter Fine Tuning

  2. LoRA Fine Tuning

Full Parameter Fine Tuning

In full parameter fine tuning change the weight of the models. See in the diagram

The change the weight of the edges. This numerical values(weight) connection between the neurals. IT’s work

Example:

Visit this website:- to see how implement Full Parameter Fine tuning

https://colab.research.google.com/drive/1XDPhdzxtYgk4ybwU-kUFrsJoNZt9r1v0?usp=sharing

Disadvantages

  1. High Hardware

  2. High GPU (Graphics processing unit)

  3. Inferencing use High parameters

  4. High Cost

Note:- Dev Tools available Don’t use Fine tune Full paramater

LoRA fine Tuning

Low-Rank Adaptation aka LoRA is a technique used to finetuning LLMs in a parameter efficient way. significantly reducing the number of trainable parameters compared to full fine-tuning.

It’s not used GPU because of propagation.

Use extra memory space

Example

]]>