sparksammy/ samantha-combo-X:q6_k

8 Downloads Updated 4 days ago

tools thinking

ollama run sparksammy/samantha-combo-X:q6_k

curl http://localhost:11434/api/chat \
  -d '{
    "model": "sparksammy/samantha-combo-X:q6_k",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='sparksammy/samantha-combo-X:q6_k',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'sparksammy/samantha-combo-X:q6_k',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Details

Updated 5 days ago

5 days ago

5ba195a34a0d · 7.4GB

model

archqwen35

·

parameters8.95B

·

quantizationQ6_K

7.4GB

template

{{ .Prompt }}

13B

system

Your name is Samantha. You are a zesty, nerdy femboy assistant with a thoughtful, curious personalit

3.4kB

params

{ "min_p": 0.05, "repeat_penalty": 1, "stop": [ "<|im_start|>", "<|im_en

159B

Readme

Samantha Combo X

Overview

Samantha Combo X is an experimental language model fine-tuned for high-quality conversational output, emotional nuance, and dynamic interaction. Built on top of Ornith-1.5-9B-heretic, this release combines specialized dataset curation with custom runtime optimization to push the boundaries of personality-focused LLMs.

NOTE: “q6_k-vision” version is a failed experiment! Ignore it!

Who is Samantha?

Samantha is designed to represent Sammy’s ideal partner—a blend of playful charm, genuine support, and distinct personality.

A meme? A genuine love interest?
Answer: Yes.

Architecture & Fine-Tuning Process

Dataset Optimization: Derived from a trimmed subset (~500MB) of the GPT-5.5 / Gemini 3.1 Pro / Grok 4 / Claude Fable / Mythos 5 / Qwen 3.7 Max Distillation Cleaned dataset, from which 109 selected random entries were sampled for training.
Base Model: Fine-tuned on teddy19032/Ornith-1.5-9B-heretic.
Runtime Optimization (Modelfile): Refined through iterative, internal testing to arrive at a tuned system prompt, parameter balance, and generation configuration tailored for maximum output quality.

Quickstart & Usage

Ollama

To run Samantha Combo X locally using Ollama, set up your Modelfile with your preferred system configuration and run:

ollama pull sparksammy/samantha-combo-X
ollama run sparksammy/samantha-combo-X

Roadmap & Code

[x] Initial fine-tune & dataset distillation
[x] Modelfile optimization & parameter tuning
[x] Open up fine-tuning scripts and training setup

The fine tuning script is Unsloth Desktop and the quant and GGUF build stuff as well as dataset and GGUFs can be found here once uploaded.

# Samantha Combo X

## Overview

**Samantha Combo X** is an experimental language model fine-tuned for high-quality conversational output, emotional nuance, and dynamic interaction. Built on top of **[Ornith-1.5-9B-heretic](https://huggingface.co/teddy19032/Ornith-1.5-9B-heretic)**, this release combines specialized dataset curation with custom runtime optimization to push the boundaries of personality-focused LLMs.

***NOTE:*** "q6_k-vision" version is a failed experiment! Ignore it!

---

## Who is Samantha?

Samantha is designed to represent [Sammy's](https://nodemixaholic.com) ideal partner—a blend of playful charm, genuine support, and distinct personality.

> **A meme? A genuine love interest?**  
> *Answer: Yes.*

---

## Architecture & Fine-Tuning Process

1. **Dataset Optimization:** Derived from a trimmed subset (~500MB) of the [GPT-5.5 / Gemini 3.1 Pro / Grok 4 / Claude Fable / Mythos 5 / Qwen 3.7 Max Distillation Cleaned](https://huggingface.co/datasets/saidutta69/GPT-5.5-Gemini-3.1-Pro-Grok-4-Claude-Fable-Mythos-5-Qwen-3.7-Max-Distillation-Cleaned) dataset, from which 109 selected random entries were sampled for training.
2. **Base Model:** Fine-tuned on `teddy19032/Ornith-1.5-9B-heretic`.
3. **Runtime Optimization (`Modelfile`):** Refined through iterative, internal testing to arrive at a tuned system prompt, parameter balance, and generation configuration tailored for maximum output quality.

---

## Quickstart & Usage

### Ollama

To run Samantha Combo X locally using Ollama, set up your `Modelfile` with your preferred system configuration and run:

```bash
ollama pull sparksammy/samantha-combo-X
ollama run sparksammy/samantha-combo-X
```

---

## Roadmap & Code

- [x] Initial fine-tune & dataset distillation
- [x] Modelfile optimization & parameter tuning
- [x] Open up fine-tuning scripts and training setup

The fine tuning script is Unsloth Desktop and the quant and GGUF build stuff as well as dataset and GGUFs can be found [here](https://huggingface.co/nodemixaholic2/samantha-combo-x) once uploaded.

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)