41 lines
1.1 KiB
Markdown
41 lines
1.1 KiB
Markdown
---
|
|
license: llama3.2
|
|
datasets:
|
|
- ChaoticNeutrals/Gryphes-Sonnet3.5-Charcard-Roleplay-Heavy_Filtering
|
|
language:
|
|
- en
|
|
base_model:
|
|
- meta-llama/Llama-3.2-3B-Instruct
|
|
pipeline_tag: text-generation
|
|
tags:
|
|
- rp
|
|
- roleplay
|
|
- creative
|
|
- writer
|
|
- llama3.2
|
|
- lora
|
|
- finetune
|
|
- experimental
|
|
---
|
|
|
|
# What is it?
|
|
Its a roleplay tuned Llama 3.2 3B instruct, trained on 5M~ tokens of high quality roleplay data, it preserves original model's foundation and capabilities while at the same time refining its writing style.
|
|
Just as any other model created by me, this one is my attempt at making local roleplay possible for any hardware.
|
|
|
|
# How good is it?
|
|
## Advantages:
|
|
* Noticeably better writing
|
|
* Noticeably less slop
|
|
* Small size
|
|
* Improved punctuation
|
|
## Disadvantages:
|
|
* Its still a 3b model, even if I did my best to make it better at roleplay
|
|
|
|
# A word on the quants for this model:
|
|
* BF16- Mostly overkill, however, highest quality
|
|
* Q8_0- Amazing quality, near lossless
|
|
* Q6_K- High quality, fast
|
|
* Q5_K_M- Mid to high quality, small and fast
|
|
* Q4_K_M- Mid quality, very small and very fast
|
|
|
|
Quants are located in the repo, along with safetensors |