---
pipeline_tag: text-generation
inference: true
widget:
- text: 'def print_hello_world():'
  example_title: Hello world
  group: Python
license: bigscience-openrail-m
datasets:
- books
- arxiv
- c4
- falcon-refinedweb
- wiki
- github-issues
- stack_markdown
library_name: transformers
tags:
- code
language:
- en
---

![image/png](https://cdn-uploads.huggingface.co/production/uploads/643a9dd0c5f633a7fa7e804a/HkB0QYV0BbmB3ktMugbZy.png)


# Refact-1.6B-base

Finally, the model we started training with our [blog post](https://refact.ai/blog/2023/applying-recent-innovations-to-train-model/) is ready 🎉
The model might contain some problems, especially with the FIM format


# It Works As a Chat

The primary application of this model is code completion (infill) in multiple programming languages.
But it works as a chat quite well.


# Example

Fill-in-the-middle uses special tokens to identify the prefix/middle/suffix part of the input and output:

```python
# pip install -q transformers
from transformers import AutoModelForCausalLM, AutoTokenizer

checkpoint = "smallcloudai/Refact-1_6B-fim"
device = "cuda" # for GPU usage or "cpu" for CPU usage

tokenizer = AutoTokenizer.from_pretrained(checkpoint)
model = AutoModelForCausalLM.from_pretrained(checkpoint, trust_remote_code=True).to(device)

prompt = '<fim_prefix>def print_hello_world():\n    """<fim_suffix>\n    print("Hello world!")<fim_middle>'

inputs = tokenizer.encode(prompt, return_tensors="pt").to(device)
outputs = model.generate(inputs, max_length=100, temperature=0.2)
print("-"*80)
print(tokenizer.decode(outputs[0]))
```

# Chat Format

The same model works as chat (experimental).

```python
prompt_template = "<empty_output>SYSTEM {system}\n" \
                  "<empty_output>USER {query}\n" \
                  "<empty_output>ASSISTANT"
prompt = prompt_template.format(system="You are a programming assistant",
                                query="How do I sort a list in Python?")
```

# Architecture

As described in more detail in the blog post, we used:

- [ALiBi](https://arxiv.org/abs/2108.12409) based attention
- [LayerNorm](https://arxiv.org/abs/1607.06450v1) instead of [RMSNorm](https://arxiv.org/pdf/1910.07467.pdf)
- [Multi Query Attention](https://arxiv.org/abs/1911.02150)

We also used LiON, flash attention, early dropout. It's not that innovative that you can't run it, in fact you can -- see an example below.


# Training

For the base model, we used our own dataset that contains code with permissive licenses only, and open text datasets.
Filtering is the key to success of this model:

- We only used text in English
- Only topics related to computer science
- Applied heavy deduplication

The text to code proportion was 50:50, model trained for 1.2T tokens. 

We don't release the base model, because its Fill-in-the-Middle (FIM) capability likes to repeat itself too much, so
its practical use is limited. But if you still want it, write us a message on Discord.


# Limitations and Bias

The Refact-1.6B model was trained on text in English. But it has seen a lot more languages in
code comments. Its performance on non-English languages is lower, for sure.


# Model Stats

- **Architecture:** LLAMA-like model with multi-query attention
- **Objectives** Fill-in-the-Middle, Chat
- **Tokens context:** 4096
- **Pretraining tokens:** 1.2T
- **Finetuning tokens:** 40B
- **Precision:** bfloat16
- **GPUs** 64 NVidia A5000
- **Training time** 28 days


# License

The model is licensed under the BigScience OpenRAIL-M v1 license agreement


# Citation

If you are using this model, please give a link to this page.