README.md

---
base_model: Qwen/Qwen3-1.7B
library_name: transformers
model_name: Vex_Amber_mini_2.5
tags:
- generated_from_trainer
- trl
- sft
- code
- reasoning
- 2B
licence: license
license: cc-by-nc-4.0
language:
- en
- fa
- fr
metrics:
- code_eval
new_version: Arioron/Vex-Amber-Mini-1.2
pipeline_tag: text-generation
num_parameters: 2000000000
---





      type: text-generation
      name: Mathematical Reasoning
    dataset:
      name: MATH
      type: math
      split: test
    metrics:
    - name: Accuracy
      type: accuracy
      value: 55.0
---

# Amber Fable 1.0

## Model Description
**Amber Fable 1.0** is a **1.7B parameter** specialized language model, fine-tuned using **LoRA (Low-Rank Adaptation)** on the powerful **Qwen3-1.7B** base model. 

This model is engineered specifically for **mathematical reasoning** and **algorithmic logic**. It achieves remarkable performance on math benchmarks (75% on GSM8K) for its size class, making it a highly efficient solution for educational tools and logic-based tasks, although it trades off some general world knowledge (MMLU) to achieve this peak reasoning capability.

- **Developed by:** Arioron
- **Model type:** Decoder-only Transformer (LoRA Adapter)
- **Language(s):** English
- **License:** Apache 2.0
- **Finetuned from model:** Qwen/Qwen3-1.7B

### Model Sources
- **Repository:** https://huggingface.co/Arioron/Amber-Fable-1.0
- **Documentation:** Arioron Model Docs

## Performance

Amber Fable 1.0 demonstrates state-of-the-art efficiency in mathematical tasks.

| Benchmark | Metric | Score | Description |
| :--- | :--- | :--- | :--- |
| **GSM8K** | Accuracy | **75.0%** | Grade School Math |
| **MATH** | Accuracy | **55.0%** | Advanced Math Problems |
| **HumanEval**| Pass@1 | **42.0%** | Python Coding Capability |
| MMLU | Accuracy | 22.0% | General World Knowledge |



## Quick Start

```python
from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

model_name = "Arioron/Amber-Fable-1.0"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.float16,
    device_map="auto"
)

# Math reasoning example
messages = [
    {"role": "user", "content": "Natalia sold clips to 48 of her friends in April, and then she sold half as many clips in May. How many clips did Natalia sell altogether in April and May?"},
]

input_text = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(input_text, return_tensors="pt").to(model.device)

outputs = model.generate(
    **inputs,
    max_new_tokens=512,
    temperature=0.6,
    do_sample=True,
    top_p=0.9,
    pad_token_id=tokenizer.eos_token_id
)

print(tokenizer.decode(outputs[0], skip_special_tokens=True))
```
### Model Summary
- **Model:** Amber Fable 1.0 (1.7B)
- **Specialty:** Advanced Math Reasoning
- **Logic:** Chain-of-Thought (CoT)
- **Coding:** Python & Algorithms (42%)
- **Tuning:** LoRA on Synthetic/Textbooks
- **Base:** Qwen3-1.7B (PyTorch/PEFT)
- **Usage:** Tutoring, Puzzles & Scripts
- **Caution:** Verify all calculations
- **Author:** Arioron (2025)
If you use this model in your research, please cite:
code
Bibtex
@misc{amberfable1.0,
  title = {Amber Fable 1.0: A Specialized 1.7B Math Model},
  author = {Arioron},
  year = {2025},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/Arioron/Amber-Fable-1.0}}

  contact
  Email: inquiry@arioron.com
Website: https://arioron.com
Documentation: https://docs.arioron.com
}