---
base_model: unsloth/gemma-3n-e4b-it-unsloth-bnb-4bit
tags:
- text-generation-inference
- transformers
- unsloth
- gemma3n
license: apache-2.0
language:
- en
datasets:
- chimbiwide/RolePlay-NPC
---

# Gemma3NPC-it-beta

#### A test model with less convervative training parameters

As mentioned in our [original article](https://huggingface.co/blog/chimbiwide/gemma3npc), we employed a very conservative training parameters for Gemma3NPC

Ever since then, we have always wanted to test the performance of the model when we make the training parameters less conservative. 

So we present ***Gemma3NPC-it-beta***.

Check out our training notebook [here](https://github.com/chimbiwide/Gemma3NPC/blob/main/Training/Gemma3NPC_Instruct_Beta.ipynb)

---

#### Training parameters compared to `Gemma3NPC-it`

| Parameter | Gemma3NPC-it | Gemma3NPC-it-beta |
| --- | --- | --- | 
| Learning Rate | 2e-5 | 2.5e-5 (+25%) | 
| Warmup Steps | 800 | 100 |
| gradient clipping | 0.4 | 1.0 |

---

Here is a graph of the Step Training Loss, saved every 10 steps:

![chart](https://cdn-uploads.huggingface.co/production/uploads/67d5b5a056a9d31aa0b49687/W3cJ_CPoLp9MZsomaZa3b.png)