|
--- |
|
pipeline_tag: text-generation |
|
inference: true |
|
widget: |
|
- text: 'def print_hello_world():' |
|
example_title: Hello world |
|
group: Python |
|
license: bigcode-openrail-m |
|
datasets: |
|
- bigcode/commitpackft |
|
- bigcode/oasst-octopack |
|
metrics: |
|
- code_eval |
|
library_name: transformers |
|
tags: |
|
- code |
|
model-index: |
|
- name: OctoCoder |
|
results: |
|
- task: |
|
type: text-generation |
|
dataset: |
|
type: bigcode/humanevalpack |
|
name: HumanEvalSynthesize Python |
|
metrics: |
|
- name: pass@1 |
|
type: pass@1 |
|
value: 46.2 |
|
verified: false |
|
- task: |
|
type: text-generation |
|
dataset: |
|
type: bigcode/humanevalpack |
|
name: HumanEvalSynthesize JavaScript |
|
metrics: |
|
- name: pass@1 |
|
type: pass@1 |
|
value: 39.2 |
|
verified: false |
|
--- |
|
|
|
 |
|
|
|
# OctoCoder |
|
|
|
Play with the model on the [TODO Playground](https://huggingface.co/spaces/bigcode/bigcode-playground). |
|
<style> |
|
table{ |
|
border-collapse: collapse; |
|
} |
|
</style> |
|
<table> |
|
<tr> |
|
<th>Model (↓)</th> |
|
<th>Python</th> |
|
<th>JavaScript</th> |
|
<th>Java</th> |
|
<th>Go</th> |
|
<th>C++</th> |
|
<th>Rust</th> |
|
<th>Avg.</th> |
|
</tr> |
|
</table> |
|
<hr style="background-color: black;"> |
|
<center><strong>HumanEvalFix</strong></center> |
|
<hr style="background-color: black;"> |
|
<center>Non-permissive models</center> |
|
<hr style="background-color: black;"> |
|
<table> |
|
<tr> |
|
<td>WizardCoder</td> |
|
<td>31.8</td> |
|
<td>29.5</td> |
|
<td>12.7</td> |
|
<td>30.4</td> |
|
<td>18.7</td> |
|
<td>13.0</td> |
|
<td>22.7</td> |
|
</tr> |
|
<tr> |
|
<td>GPT-4</td> |
|
<td>47.0 </td> |
|
<td>48.2</td> |
|
<td>50.0</td> |
|
<td>50.6</td> |
|
<td>47.6</td> |
|
<td>43.3</td> |
|
<td><u>47.8</u></td> |
|
</tr> |
|
</table> |
|
<hr style="background-color: black;"> |
|
<center>Permissive models</center> |
|
<hr style="background-color: black;"> |
|
<table> |
|
<tr> |
|
<td>InstructCodeT5+<sup>‡</sup></td> |
|
<td>2.7</td> |
|
<td>1.2</td> |
|
<td>4.3</td> |
|
<td>2.1</td> |
|
<td>0.2</td> |
|
<td>0.5</td> |
|
<td>1.8</td> |
|
</tr> |
|
<tr> |
|
<td>BLOOMZ<sup>+</sup></td> |
|
<td>16.6</td> |
|
<td>15.5</td> |
|
<td>15.2</td> |
|
<td>16.4</td> |
|
<td>6.7</td> |
|
<td>5.7</td> |
|
<td>12.5</td> |
|
</tr> |
|
<tr> |
|
<td>StarChat-β</td> |
|
<td>18.1</td> |
|
<td>18.1</td> |
|
<td>24.1</td> |
|
<td>18.1</td> |
|
<td>8.2</td> |
|
<td>3.6</td> |
|
<td>11.2</td> |
|
</tr> |
|
<tr> |
|
<td>CodeGeeX2<sup>*</sup></td> |
|
<td>15.9</td> |
|
<td>14.7</td> |
|
<td>18.0</td> |
|
<td>13.6</td> |
|
<td>4.3</td> |
|
<td>6.1</td> |
|
<td>12.1</td> |
|
</tr> |
|
<tr> |
|
<td>StarCoder</td> |
|
<td>8.7</td> |
|
<td>15.7</td> |
|
<td>13.3</td> |
|
<td>20.1</td> |
|
<td>15.6</td> |
|
<td>6.7</td> |
|
<td>13.4</td> |
|
</tr> |
|
<tr> |
|
<td>OctoGeeX<sup>*</sup></td> |
|
<td>28.1</td> |
|
<td>27.7</td> |
|
<td>30.4</td> |
|
<td>27.6</td> |
|
<td>22.9</td> |
|
<td>9.6</td> |
|
<td>24.4</td> |
|
</tr> |
|
<tr> |
|
<td>OctoCoder</td> |
|
<td><strong>30.2</strong></td> |
|
<td><strong>28.4</strong></td> |
|
<td><strong>30.6</strong></td> |
|
<td><strong>30.2</strong></td> |
|
<td><strong>26.1</strong></td> |
|
<td><strong>16.5</strong></td> |
|
<td><strong>27.0</strong></td> |
|
</tr> |
|
</table> |
|
<hr style="background-color: black;"> |
|
<center><h4>HumanEvalExplain</h4></center> |
|
<hr style="background-color: black;"> |
|
<center>Non-permissive models</center> |
|
<hr style="background-color: black;"> |
|
<table> |
|
<tr> |
|
<td>WizardCoder</td> |
|
<td>32.5</td> |
|
<td>33.0</td> |
|
<td>27.4</td> |
|
<td>26.7</td> |
|
<td>28.2</td> |
|
<td>16.9</td> |
|
<td>27.5</td> |
|
</tr> |
|
<tr> |
|
<td>GPT-4</td> |
|
<td>64.6</td> |
|
<td>57.3</td> |
|
<td>51.2</td> |
|
<td>58.5</td> |
|
<td>38.4</td> |
|
<td>42.7</td> |
|
<td><u>52.1</u></td> |
|
</tr> |
|
</table> |
|
<hr style="background-color: black;"> |
|
<center>Permissive models</center> |
|
<hr style="background-color: black;"> |
|
<table> |
|
<tr> |
|
<td>InstructCodeT5+<sup>‡</sup></td> |
|
<td>20.8</td> |
|
<td>0.0</td> |
|
<td>0.0</td> |
|
<td>0.0</td> |
|
<td>0.1</td> |
|
<td>0.0</td> |
|
<td>3.5</td> |
|
</tr> |
|
<tr> |
|
<td>BLOOMZ<sup>+</sup></td> |
|
<td>14.7</td> |
|
<td>8.8</td> |
|
<td>12.1</td> |
|
<td>8.5</td> |
|
<td>0.6</td> |
|
<td>0.0</td> |
|
<td>7.5</td> |
|
</tr> |
|
<tr> |
|
<td>StarChat-β</td> |
|
<td>25.4</td> |
|
<td>21.5</td> |
|
<td>24.5</td> |
|
<td>18.4</td> |
|
<td>17.6</td> |
|
<td>13.2</td> |
|
<td>20.1</td> |
|
</tr> |
|
<tr> |
|
<td>CodeGeeX2<sup>*</sup></td> |
|
<td>0.0</td> |
|
<td>0.0</td> |
|
<td>0.0</td> |
|
<td>0.0</td> |
|
<td>0.0</td> |
|
<td>0.0</td> |
|
<td>0.0</td> |
|
</tr> |
|
<tr> |
|
<td>StarCoder</td> |
|
<td>0.0</td> |
|
<td>0.0</td> |
|
<td>0.0</td> |
|
<td>0.0</td> |
|
<td>0.0</td> |
|
<td>0.0</td> |
|
<td>0.0</td> |
|
</tr> |
|
<tr> |
|
<td>OctoGeeX<sup>*</sup></td> |
|
<td>30.4</td> |
|
<td>24.0</td> |
|
<td>24.7</td> |
|
<td><strong>21.7</strong></td> |
|
<td>21.0</td> |
|
<td><strong>15.9</strong></td> |
|
<td>22.9</td> |
|
</tr> |
|
<tr> |
|
<td>OctoCoder</td> |
|
<td><strong>35.1</strong></td> |
|
<td><strong>24.5</strong></td> |
|
<td><strong>27.3</strong></td> |
|
<td>21.1</td> |
|
<td><strong>24.1</strong></td> |
|
<td>14.8</td> |
|
<td><strong>24.5</strong></td> |
|
</tr> |
|
</table> |
|
<hr style="background-color: black;"> |
|
<center><h4>HumanEvalSynthesize</h4></center> |
|
<hr style="background-color: black;"> |
|
<center>Non-permissive models</center> |
|
<hr style="background-color: black;"> |
|
<table> |
|
<tr> |
|
<td>WizardCoder</td> |
|
<td>57.3</td> |
|
<td>49.5</td> |
|
<td>36.1</td> |
|
<td>36.4</td> |
|
<td>40.9</td> |
|
<td>20.2</td> |
|
<td>40.1</td> |
|
</tr> |
|
<tr> |
|
<td>GPT-4</td> |
|
<td>86.6</td> |
|
<td>82.9</td> |
|
<td>81.7</td> |
|
<td>72.6</td> |
|
<td>78.7</td> |
|
<td>67.1</td> |
|
<td><u>78.3</u></td> |
|
</tr> |
|
</table> |
|
<hr style="background-color: black;"> |
|
<center>Permissive models</center> |
|
<hr style="background-color: black;"> |
|
<table> |
|
<tr> |
|
<td>InstructCodeT5+<sup>‡</sup></td> |
|
<td>37.0</td> |
|
<td>18.9</td> |
|
<td>17.4</td> |
|
<td>9.5</td> |
|
<td>19.8</td> |
|
<td>0.3</td> |
|
<td>17.1</td> |
|
</tr> |
|
<tr> |
|
<td>BLOOMZ<sup>+</sup></td> |
|
<td>15.6</td> |
|
<td>14.8</td> |
|
<td>18.4</td> |
|
<td>8.4</td> |
|
<td>6.5</td> |
|
<td>5.5</td> |
|
<td>11.5</td> |
|
</tr> |
|
<tr> |
|
<td>StarChat-β</td> |
|
<td>33.5</td> |
|
<td>31.4</td> |
|
<td>26.7</td> |
|
<td>25.5</td> |
|
<td>26.6</td> |
|
<td>14.0</td> |
|
<td>26.3</td> |
|
</tr> |
|
<tr> |
|
<td>CodeGeeX2<sup>*</sup></td> |
|
<td>35.9</td> |
|
<td>32.2</td> |
|
<td>30.8</td> |
|
<td>22.5</td> |
|
<td>29.3</td> |
|
<td>18.1</td> |
|
<td>28.1</td> |
|
</tr> |
|
<tr> |
|
<td>StarCoder</td> |
|
<td>33.6</td> |
|
<td>30.8</td> |
|
<td>30.2</td> |
|
<td>17.6</td> |
|
<td>31.6</td> |
|
<td>21.8</td> |
|
<td>27.6</td> |
|
</tr> |
|
<tr> |
|
<td>OctoGeeX<sup>*</sup></td> |
|
<td>44.7</td> |
|
<td>33.8</td> |
|
<td>36.9</td> |
|
<td>21.9</td> |
|
<td>32.3</td> |
|
<td>15.7</td> |
|
<td>30.9</td> |
|
</tr> |
|
<tr> |
|
<td>OctoCoder</td> |
|
<td><strong>46.2</strong></td> |
|
<td><strong>39.2</strong></td> |
|
<td><strong>38.2</strong></td> |
|
<td><strong>30.4</strong></td> |
|
<td><strong>35.6</strong></td> |
|
<td><strong>23.4</strong></td> |
|
<td><strong>35.5</strong></td> |
|
</tr> |
|
</table> |
|
|
|
## Table of Contents |
|
|
|
1. [Model Summary](##model-summary) |
|
2. [Use](##use) |
|
3. [Limitations](##limitations) |
|
4. [Training](##training) |
|
5. [License](##license) |
|
6. [Citation](##citation) |
|
|
|
## Model Summary |
|
|
|
OctoCoder is an instruction tuned model with 15.5B parameters created by finetuning StarCoder on CommitPackFT & OASST as described in the OctoPack paper. |
|
|
|
- **Repository:** [bigcode/octopack](https://github.com/bigcode-project/octopack) |
|
- **Paper:** [TODO]() |
|
- **Languages:** 80+ Programming languages |
|
|
|
## Use |
|
|
|
### Intended use |
|
|
|
The model follows instructions provided in the input. We recommend prefacing your input with "Question: " and finishing with "Answer:", for example: "Question: Please write a function in Python that performs bubble sort.\n\nAnswer:" |
|
|
|
**Feel free to share your generations in the Community tab!** |
|
|
|
### Generation |
|
```python |
|
# pip install -q transformers |
|
from transformers import AutoModelForCausalLM, AutoTokenizer |
|
|
|
checkpoint = "bigcode/octocoder" |
|
device = "cuda" # for GPU usage or "cpu" for CPU usage |
|
|
|
tokenizer = AutoTokenizer.from_pretrained(checkpoint) |
|
model = AutoModelForCausalLM.from_pretrained(checkpoint).to(device) |
|
|
|
inputs = tokenizer.encode("Question: Please write a function in Python that performs bubble sort.\n\nAnswer:", return_tensors="pt").to(device) |
|
outputs = model.generate(inputs) |
|
print(tokenizer.decode(outputs[0])) |
|
``` |
|
|
|
# Training |
|
|
|
## Model |
|
|
|
- **Architecture:** GPT-2 model with multi-query attention and Fill-in-the-Middle objective |
|
- **Steps:** 250k pretraining & 30 instruction tuning |
|
- **Pretraining tokens:** 1 trillion pretraining & 2M instruction tuning |
|
- **Precision:** bfloat16 |
|
|
|
## Hardware |
|
|
|
- **Pretraining:** |
|
- **GPUs:** 512 Tesla A100 |
|
- **Training time:** 24 days |
|
- **Instruction tuning:** |
|
- **GPUs:** 8 Tesla A100 |
|
- **Training time:** 4 hours |
|
|
|
## Software |
|
|
|
- **Orchestration:** [Megatron-LM/Transformers](https://github.com/bigcode-project/octopack#training) |
|
- **Neural networks:** [PyTorch](https://github.com/pytorch/pytorch) |
|
|
|
# Citation |
|
|
|
TODO |
