xiaxinhang/pr-agent
1
0
Fork 0
mirror of https://github.com/qodo-ai/pr-agent.git synced 2025-07-14 17:50:37 +08:00
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #937 from KennyDizi/main

Browse Source 
Optimize document table format
This commit is contained in:
Tal
2024-06-01 10:53:47 +03:00
committed by GitHub
parent 37f6e18953 40658cfb7c
commit 1e1636911f
1 changed files with 28 additions and 138 deletions
Download Patch File Download Diff File Expand all files Collapse all files

166
docs/docs/finetuning_benchmark/index.md
View File

@ -1,8 +1,8 @@
# PR-Agent Code Fine-tuning Benchmark # PR-Agent Code Fine-tuning Benchmark
On coding tasks, the gap between open-source models and top closed-source models such as GPT4 is significant. On coding tasks, the gap between open-source models and top closed-source models such as GPT4 is significant.
<br> <br>
In practice, open-source models are unsuitable for most real-world code tasks, and require further fine-tuning to produce acceptable results. In practice, open-source models are unsuitable for most real-world code tasks, and require further fine-tuning to produce acceptable results.
_PR-Agent fine-tuning benchmark_ aims to benchmark open-source models on their ability to be fine-tuned for a code task. _PR-Agent fine-tuning benchmark_ aims to benchmark open-source models on their ability to be fine-tuned for a code task.
Specifically, we chose to fine-tune open-source models on the task of analyzing a pull request, and providing useful feedback and code suggestions. Specifically, we chose to fine-tune open-source models on the task of analyzing a pull request, and providing useful feedback and code suggestions.
@ -11,144 +11,33 @@ Here are the results:
<br> <br>
<br> <br>
<head> **Model performance:**
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<style>
td, th {
font-size: 16px; /* Adjust this value to your preference */
}
table {
width: 100%;
border-collapse: collapse;
}
th {
background-color: #f2f2f2;
border: 1px solid #dddddd;
text-align: center;
padding: 8px;
}
td {
border: 1px solid #dddddd;
text-align: center;
padding: 8px;
}
tr:nth-child(even) {
background-color: #f9f9f9;
text-align: center;
}
</style>
<title>Model Performance Table</title>
</head>
<body>
<table> | Model name | Model size [B] | Better than gpt-4 rate, after fine-tuning [%] |
<tr> |-----------------------------|----------------|----------------------------------------------|
<th align="center">Model name</th> | **DeepSeek 34B-instruct** | **34** | **40.7** |
<th align="center">Model size [B]</th> | DeepSeek 34B-base | 34 | 38.2 |
<th align="center">Better than gpt-4 rate, after fine-tuning [%]</th> | Phind-34b | 34 | 38 |
</tr> | Granite-34B | 34 | 37.6 |
<tr> | Codestral-22B-v0.1 | 22 | 32.7 |
<td align="center"><b>DeepSeek 34B-instruct</b></td> | QWEN-1.5-32B | 32 | 29 |
<td align="center"><b>34</b></td> | | | |
<td align="center"><b>40.7</b></td> | **CodeQwen1.5-7B** | **7** | **35.4** |
</tr> | Granite-8b-code-instruct | 8 | 34.2 |
<tr> | CodeLlama-7b-hf | 7 | 31.8 |
<td align="center">DeepSeek 34B-base</td> | Gemma-7B | 7 | 27.2 |
<td align="center">34</td> | DeepSeek coder-7b-instruct | 7 | 26.8 |
<td align="center">38.2</td> | Llama-3-8B-Instruct | 8 | 26.8 |
</tr> | Mistral-7B-v0.1 | 7 | 16.1 |
<tr>
<td align="center">Phind-34b</td>
<td align="center">34</td>
<td align="center">38</td>
</tr>
<tr>
<td align="center">Granite-34B</td>
<td align="center">34</td>
<td align="center">37.6</td>
</tr>
<tr>
<td align="center">Codestral-22B-v0.1</td>
<td align="center">22</td>
<td align="center">32.7</td>
</tr>
<tr>
<td align="center">QWEN-1.5-32B</td>
<td align="center">32</td>
<td align="center">29</td>
</tr>
<tr>
<td align="center"></td>
<td align="center"></td>
<td align="center"></td>
</tr>
<tr>
<td align="center"><b>CodeQwen1.5-7B</b></td>
<td align="center"><b>7</b></td>
<td align="center"><b>35.4</b></td>
</tr>
<tr>
<td align="center">Granite-8b-code-instruct</td>
<td align="center">8</td>
<td align="center">34.2</td>
</tr>
<tr>
<td align="center">CodeLlama-7b-hf</td>
<td align="center">7</td>
<td align="center">31.8</td>
</tr>
<tr>
<td align="center">Gemma-7B</td>
<td align="center">7</td>
<td align="center">27.2</td>
</tr>
<tr>
<td align="center">DeepSeek coder-7b-instruct</td>
<td align="center">7</td>
<td align="center">26.8</td>
</tr>
<tr>
<td align="center">Llama-3-8B-Instruct</td>
<td align="center">8</td>
<td align="center">26.8</td>
</tr>
<tr>
<td align="center">Mistral-7B-v0.1</td>
<td align="center">7</td>
<td align="center">16.1</td>
</tr>
</table>
</body>
<br> <br>
**Fine-tuning impact:** **Fine-tuning impact:**
<body>
<table>
<tr>
<th align="center">Model name</th>
<th>Model size [B]</th>
<th>Fine-tuned</th>
<th>Better than gpt-4 rate [%]</th>
</tr>
<tr>
<td align="center">DeepSeek 34B-instruct</td>
<td align="center">34</td>
<td align="center">yes</td>
<td align="center">40.7</td>
</tr>
<tr>
<td align="center">DeepSeek 34B-instruct</td>
<td align="center">34</td>
<td align="center">no</td>
<td align="center">3.6</td>
</tr>
</table>
</body>
| Model name | Model size [B] | Fine-tuned | Better than gpt-4 rate [%] |
|---------------------------|----------------|------------|----------------------------|
| DeepSeek 34B-instruct | 34 | yes | 40.7 |
| DeepSeek 34B-instruct | 34 | no | 3.6 |
## Results analysis ## Results analysis
@ -159,10 +48,9 @@ Here are the results:
- **The best small model** - For small 7B code-dedicated models, the gaps when fine-tuning are much larger. **CodeQWEN 1.5-7B** is by far the best model for fine-tuning. - **The best small model** - For small 7B code-dedicated models, the gaps when fine-tuning are much larger. **CodeQWEN 1.5-7B** is by far the best model for fine-tuning.
- **Base vs. instruct** - For the top model (deepseek), we saw small advantage when starting from the instruct version. However, we recommend testing both versions on each specific task, as the base model is generally considered more suitable for fine-tuning. - **Base vs. instruct** - For the top model (deepseek), we saw small advantage when starting from the instruct version. However, we recommend testing both versions on each specific task, as the base model is generally considered more suitable for fine-tuning.
## The dataset
## The dataset ### Training dataset
### Training dataset
Our training dataset is comprised of 25,000 pull requests, aggregated from permissive license repos. For each pull request, we generated responses for the three main tools of PR-Agent: Our training dataset is comprised of 25,000 pull requests, aggregated from permissive license repos. For each pull request, we generated responses for the three main tools of PR-Agent:
[Describe](https://pr-agent-docs.codium.ai/tools/describe/), [Review](https://pr-agent-docs.codium.ai/tools/improve/) and [Improve](https://pr-agent-docs.codium.ai/tools/improve/). [Describe](https://pr-agent-docs.codium.ai/tools/describe/), [Review](https://pr-agent-docs.codium.ai/tools/improve/) and [Improve](https://pr-agent-docs.codium.ai/tools/improve/).
@ -171,11 +59,13 @@ On the raw data collected, we employed various automatic and manual cleaning tec
An example input prompt can be found [here](https://github.com/Codium-ai/pr-agent/blob/main/pr_agent/settings/pr_code_suggestions_prompts.toml), and an example output can be found [here](https://github.com/Codium-ai/pr-agent/pull/910#issuecomment-2118761309). An example input prompt can be found [here](https://github.com/Codium-ai/pr-agent/blob/main/pr_agent/settings/pr_code_suggestions_prompts.toml), and an example output can be found [here](https://github.com/Codium-ai/pr-agent/pull/910#issuecomment-2118761309).
### Evaluation dataset ### Evaluation dataset
- For each tool, we aggregated 100 additional examples to be used for evaluation. These examples were not used in the training dataset, and were manually selected to represent diverse real-world use-cases. - For each tool, we aggregated 100 additional examples to be used for evaluation. These examples were not used in the training dataset, and were manually selected to represent diverse real-world use-cases.
- For each test example, we generated two responses: one from the fine-tuned model, and one from the best code model in the world, `gpt-4-turbo-2024-04-09`. - For each test example, we generated two responses: one from the fine-tuned model, and one from the best code model in the world, `gpt-4-turbo-2024-04-09`.
- We used a third LLM to judge which response better answers the prompt, and will likely be perceived by a human as better response. - We used a third LLM to judge which response better answers the prompt, and will likely be perceived by a human as better response.
<br> <br>
We experimented with three model as judges: `gpt-4-turbo-2024-04-09`, `gpt-4o`, and `claude-3-opus-20240229`. All three produced similar results, with the same ranking order. This strengthens the validity of our testing protocol. We experimented with three model as judges: `gpt-4-turbo-2024-04-09`, `gpt-4o`, and `claude-3-opus-20240229`. All three produced similar results, with the same ranking order. This strengthens the validity of our testing protocol.
Here is an example for a judge model feedback: Here is an example for a judge model feedback:
@ -191,4 +81,4 @@ why: |
are practical improvements. In contrast, Response 2 focuses more on general advice and less are practical improvements. In contrast, Response 2 focuses more on general advice and less
actionable suggestions, such as changing variable names and adding comments, which are less actionable suggestions, such as changing variable names and adding comments, which are less
critical for immediate code improvement." critical for immediate code improvement."
``` ```