---
license: apache-2.0
pipeline_tag: text-generation
language:
  - en
tags:
  - pretrained
datasets:
  - Skylion007/openwebtext
  - c4
  - wikimedia/wikipedia
  - tiiuae/falcon-refinedweb
  - izumi-lab/open-text-books
  - togethercomputer/RedPajama-Data-V2
  - databricks/databricks-dolly-15k
  - euclaise/reddit-instruct-curated
  - CohereForAI/aya_dataset
widget:
  - messages:
      - role: user
        content: Specs of a game about trolls and warriors in a fantasy world.
  - messages:
      - role: user
        content: Reducing waste generation is essential to...
  - messages:
      - role: user
        content: Water, planet, resource, future
  - messages:
      - role: user
        content: Background story of an RPG game about wizards and dragons in a sci-fi world. The story takes place in a...
inference:
  parameters:
    max_new_tokens: 250
    do_sample: true
    temperature: 0.65
    top_p: 0.55
    top_k: 35
    repetition_penalty: 1.176
---

# Minueza-32M-Base

## Summary

Minueza-32M-Base is a foundation model with 32 million parameters trained from scratch on a large corpus of text in English.

It's available in the following formats: [Safetensors](https://huggingface.co/Felladrin/Minueza-32M-Base), [GGUF](https://huggingface.co/Felladrin/gguf-Minueza-32M-Base), and [ONNX](https://huggingface.co/Felladrin/onnx-Minueza-32M-Base).

And it's being released alongside these fine-tuned versions:
  - [Minueza-32M-UltraChat](https://huggingface.co/Felladrin/Minueza-32M-UltraChat): Trained in a single conversational dataset.
  - [Minueza-32M-Chat](https://huggingface.co/Felladrin/Minueza-32M-Chat): Trained in a mix of conversational datasets.

## Intended Uses

This model was created with the following objectives in mind:

- Run on mobile web browsers via [Transformers.js](https://huggingface.co/docs/transformers.js).
- Run fast on machines without GPU.
- Serve as a base for fine-tunes using ChatML format, hence the two additional special tokens (`<|im_start|>` and `<|im_end|>`) with `<|im_end|>` as default EOS token.
  - ChatML works great for both instruction and chat models, so if all fine-tunes are made following the ChatML pattern, other users might benefit from the easiness of creating merges.

## Datasets

The model was trained on a subset of each of the following non-synthetic datasets:

- [Skylion007/openwebtext](https://huggingface.co/datasets/Skylion007/openwebtext)
- [c4](https://huggingface.co/datasets/c4)
- [wikimedia/wikipedia - 20231101.simple](https://huggingface.co/datasets/wikimedia/wikipedia/viewer/20231101.simple)
- [tiiuae/falcon-refinedweb](https://huggingface.co/datasets/tiiuae/falcon-refinedweb)
- [izumi-lab/open-text-books](https://huggingface.co/datasets/izumi-lab/open-text-books)
- [togethercomputer/RedPajama-Data-V2](https://huggingface.co/datasets/togethercomputer/RedPajama-Data-V2)
- [databricks/databricks-dolly-15k](https://huggingface.co/datasets/databricks/databricks-dolly-15k)
- [euclaise/reddit-instruct-curated](https://huggingface.co/datasets/euclaise/reddit-instruct-curated)
- [CohereForAI/aya_dataset - original english annotations](https://huggingface.co/datasets/CohereForAI/aya_dataset/viewer/default/train?f[language_code][value]=%27eng%27)

The subsets were interleaved to form the final training corpus of approximately 650 million tokens.

## Model Architecture

This is a transformer model with the Mistral architecture, trained on a context window of 2048 tokens.

| Configuration           | Value |
| :---------------------- | :---- |
| max_position_embeddings | 2048  |
| hidden_size             | 312   |
| intermediate_size       | 1092  |
| num_attention_heads     | 12    |
| num_hidden_layers       | 10    |
| num_key_value_heads     | 4     |
| vocab_size              | 32002 |

The pretraining was made with these hyperparameters and frameworks:

| Hyperparameter              | Value                                         |
| :-------------------------- | :-------------------------------------------- |
| learning_rate               | 5e-05                                         |
| train_batch_size            | 1                                             |
| eval_batch_size             | 1                                             |
| seed                        | 42                                            |
| gradient_accumulation_steps | 8                                             |
| total_train_batch_size      | 8                                             |
| optimizer                   | Adam with betas=(0.9,0.999) and epsilon=1e-08 |
| lr_scheduler_type           | linear                                        |

| Framework    | Version     |
| :----------- | :---------- |
| Transformers | 4.38.0.dev0 |
| Pytorch      | 2.1.2       |
| Datasets     | 2.16.1      |
| Tokenizers   | 0.15.1      |

## Usage

This is just a base model. For your task, you will likely want to perform application-specific fine-tuning as recommended above.

Also note that this model was trained on internet text data, which may contain biases, offensive or inappropriate content, and may produce incorrect or irrelevant responses. No evaluation has been conducted, so use with care.

Having that said, here's how you can run it:

```python
from transformers import pipeline

generate = pipeline("text-generation", "Felladrin/Minueza-32M-Base")

prompt = "The best way to improve your health is"

output = generate(
    prompt,
    max_new_tokens=256,
    do_sample=True,
    temperature=0.72,
    top_p=0.73,
    top_k=50,
    repetition_penalty=1.176,
)

print(output[0]["generated_text"])
```

## License

This model is licensed under the [Apache License 2.0](https://huggingface.co/Felladrin/Minueza-32M-Base/resolve/main/license.txt).