README.md

---
license: apache-2.0
datasets:
- Intel/orca_dpo_pairs
- Locutusque/Hercules-v3.0
language:
- en
inference:
  parameters:
    do_sample: true
    temperature: 0.8
    top_p: 0.95
    top_k: 40
    min_new_tokens: 2
    max_new_tokens: 250
    repetition_penalty: 1.1
---
# NeuralReyna-Mini-1.8B-v0.2
![Reyna image](https://th.bing.com/th/id/OIG3.8IBxuT77hh6Y_r1DZ6WK?dpr=2.6&pid=ImgDetMain)
# Description
Taken aloobun/Reyna-Mini-1.8B-v0.2 and further fine-tuned it using DPO using the Intel/orca_dpo_pairs dataset. 

This model has capabilities in coding, math, science, roleplay, and function calling. 

This model was trained on OpenAI's ChatML prompt format.
# Evaluation
AGIEval:
![eval](https://media.discordapp.net/attachments/1190343662400458884/1209814747864764426/Screenshot_from_2024-02-21_16-21-09.png?ex=65e84ac9&is=65d5d5c9&hm=b3aca22a8add5510bce55fa648c5d204e495ceb54cbe2bc9f63b752a6bcb6c9e&)

GPT4ALL:
|    Tasks    |Version|Filter|n-shot| Metric |Value |   |Stderr|
|-------------|------:|------|-----:|--------|-----:|---|-----:|
|arc_challenge|      1|none  |     0|acc     |0.3208|±  |0.0136|
|             |       |none  |     0|acc_norm|0.3336|±  |0.0138|
|arc_easy     |      1|none  |     0|acc     |0.6035|±  |0.0100|
|             |       |none  |     0|acc_norm|0.5833|±  |0.0101|
|boolq        |      2|none  |     0|acc     |0.6526|±  |0.0083|
|hellaswag    |      1|none  |     0|acc     |0.4556|±  |0.0050|
|             |       |none  |     0|acc_norm|0.6076|±  |0.0049|
|openbookqa   |      1|none  |     0|acc     |0.2600|±  |0.0196|
|             |       |none  |     0|acc_norm|0.3460|±  |0.0213|
|piqa         |      1|none  |     0|acc     |0.7236|±  |0.0104|
|             |       |none  |     0|acc_norm|0.7307|±  |0.0104|
|winogrande   |      1|none  |     0|acc     |0.6062|±  |0.0137|


# Disclaimer

This model may have overfitted to the DPO training data, and may not perform well. 

# Contributions

Thanks to @aloobun and @Locutusque for their contributions to this model.