Valentin Buchner commited on
Commit
797106d
·
1 Parent(s): 03b54a3

remove markdown leaderboard, add huggingface space as remote

Browse files
.github/scripts/generate-leaderboard.py DELETED
@@ -1,63 +0,0 @@
1
- import json
2
-
3
-
4
- def generate_markdown(data):
5
- markdown = """<div align="center">\n\n"""
6
- markdown += "# 🔥🏅️GenCeption Leaderboard 🏅️🔥\n\n"
7
- markdown += """\n\n</div>\n\n"""
8
- markdown += "#### GC@3 scores for different models and categories:\n"
9
- markdown += "| Model | **Mean** | Exist. | Count | Posi. | Col. | Post. | Cel. | Sce. | Lan. | Art. | Comm. | **Vis Mean** | Code | Num. | Tran. | OCR | **Text Mean** |\n"
10
- markdown += (
11
- "|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|\n"
12
- )
13
- for model in data["models"]:
14
- scores = model["scores"]
15
- markdown += f"| [{model['name']}]({model['url']}) "
16
- for score_key in [
17
- "Mean",
18
- "Exist",
19
- "Count",
20
- "Posi",
21
- "Col",
22
- "Post",
23
- "Cel",
24
- "Sce",
25
- "Lan",
26
- "Art",
27
- "Comm",
28
- "VisMean",
29
- "Code",
30
- "Num",
31
- "Tran",
32
- "OCR",
33
- "TextMean",
34
- ]:
35
- if "Mean" in score_key:
36
- markdown += f"| **{scores[score_key]}** "
37
- else:
38
- markdown += f"| {scores[score_key]} "
39
- markdown += "|\n"
40
- markdown += """\n\nLegend:
41
- - Exist.: Existence
42
- - Count: Count
43
- - Posi.: Position
44
- - Col.: Color
45
- - Post.: Poster
46
- - Cel.: Celebrity
47
- - Sce.: Scene
48
- - Lan.: Landmark
49
- - Art.: Artwork
50
- - Com. R.: Commonsense Reasoning
51
- - Code: Code Reasoning
52
- - Num.: Numerical Calculation
53
- - Tran.: Text Translation
54
- - OCR: OCR"""
55
- return markdown
56
-
57
-
58
- if __name__ == "__main__":
59
- with open("leaderboard/leaderboard.json", "r") as f:
60
- data = json.load(f)
61
- markdown_content = generate_markdown(data)
62
- with open("leaderboard/Leaderboard.md", "w") as f:
63
- f.write(markdown_content)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
.github/workflows/update-leaderboard.yaml DELETED
@@ -1,31 +0,0 @@
1
- name: Update Markdown from JSON
2
-
3
- on:
4
- push:
5
- paths:
6
- - 'leaderboard/leaderboard.json'
7
-
8
- jobs:
9
- update-markdown:
10
- runs-on: ubuntu-latest
11
- steps:
12
- - uses: actions/checkout@v4 # Updated to the latest version at the time of writing
13
- - name: Set up Python
14
- uses: actions/setup-python@v5
15
- with:
16
- python-version: '3.10'
17
- - name: Install dependencies
18
- run: |
19
- python -m pip install --upgrade pip
20
- pip install pyyaml
21
- - name: Generate Markdown
22
- run: python .github/scripts/generate-leaderboard.py
23
- - name: Commit and push if changed
24
- env:
25
- GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
26
- run: |
27
- git config --global user.email "[email protected]"
28
- git config --global user.name "GitHub Action"
29
- git add leaderboard/Leaderboard.md
30
- git commit -m "Update Leaderboard.md from JSON data" -a || echo "No changes to commit"
31
- git push
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
README.md CHANGED
@@ -2,10 +2,9 @@
2
 
3
  <div>
4
  <p align="center">
5
- <a href="https://github.com/EQTPartners/GenCeption/blob/main/leaderboard/Leaderboard.md">🔥🏅️Leaderboard🏅️🔥</a>&emsp;•&emsp;
6
  <a href="#contribute">Contribute</a>&emsp;•&emsp;
7
  <a href="https://arxiv.org/abs/2402.14973">Paper</a>&emsp;•&emsp;
8
- <a href="">HF Space 🤗</a>&emsp;•&emsp;
9
  <a href="#cite-this-work">Citation</a>
10
  </p>
11
 
@@ -23,7 +22,7 @@ We demostrate a 5-iteration GenCeption procedure below run on a seed images to e
23
 
24
 
25
  ## Contribute
26
- Please add your model details and results to `leaderboard/leaderboard.json` and **create a PR (Pull-Request)** to contribute your results to the [🔥🏅️**Leaderboard**🏅️🔥](https://github.com/EQTPartners/GenCeption/blob/main/leaderboard/Leaderboard.md). Start by creating your virtual environment:
27
 
28
  ```{bash}
29
  conda create --name genception python=3.10 -y
 
2
 
3
  <div>
4
  <p align="center">
5
+ <a href="https://huggingface.co/spaces/valbuc/GenCeption">🔥🏅️🤗 Leaderboard🏅️🔥</a>&emsp;•&emsp;
6
  <a href="#contribute">Contribute</a>&emsp;•&emsp;
7
  <a href="https://arxiv.org/abs/2402.14973">Paper</a>&emsp;•&emsp;
 
8
  <a href="#cite-this-work">Citation</a>
9
  </p>
10
 
 
22
 
23
 
24
  ## Contribute
25
+ Please add your model details and results to `leaderboard/leaderboard.json` and **create a PR (Pull-Request)** to contribute your results to the [🔥🏅️**Leaderboard**🏅️🔥](https://huggingface.co/spaces/). Start by creating your virtual environment:
26
 
27
  ```{bash}
28
  conda create --name genception python=3.10 -y
leaderboard/Leaderboard.md DELETED
@@ -1,32 +0,0 @@
1
- <div align="center">
2
-
3
- # 🔥🏅️GenCeption Leaderboard 🏅️🔥
4
-
5
-
6
-
7
- </div>
8
-
9
- #### GC@3 scores for different models and categories:
10
- | Model | **Mean** | Exist. | Count | Posi. | Col. | Post. | Cel. | Sce. | Lan. | Art. | Comm. | **Vis Mean** | Code | Num. | Tran. | OCR | **Text Mean** |
11
- |---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
12
- | [ChatGPT-4V](https://cdn.openai.com/papers/GPTV_System_Card.pdf) | **0.351** | 0.422 | 0.404 | 0.408 | 0.403 | 0.324 | 0.332 | 0.393 | 0.353 | 0.421 | 0.471 | **0.393** | 0.193 | 0.24 | 0.157 | 0.393 | **0.246** |
13
- | [mPLUG-Owl2](https://arxiv.org/pdf/2311.04257.pdf) | **0.257** | 0.323 | 0.299 | 0.306 | 0.29 | 0.243 | 0.232 | 0.299 | 0.275 | 0.252 | 0.353 | **0.287** | 0.176 | 0.192 | 0.081 | 0.276 | **0.181** |
14
- | [LLaVA-13B](https://arxiv.org/pdf/2304.08485.pdf) | **0.238** | 0.305 | 0.294 | 0.255 | 0.3 | 0.215 | 0.206 | 0.277 | 0.242 | 0.212 | 0.334 | **0.264** | 0.144 | 0.195 | 0.116 | 0.239 | **0.174** |
15
- | [LLaVA-7B](https://arxiv.org/pdf/2304.08485.pdf) | **0.225** | 0.308 | 0.253 | 0.285 | 0.284 | 0.214 | 0.188 | 0.266 | 0.252 | 0.21 | 0.294 | **0.255** | 0.107 | 0.155 | 0.111 | 0.222 | **0.149** |
16
-
17
-
18
- Legend:
19
- - Exist.: Existence
20
- - Count: Count
21
- - Posi.: Position
22
- - Col.: Color
23
- - Post.: Poster
24
- - Cel.: Celebrity
25
- - Sce.: Scene
26
- - Lan.: Landmark
27
- - Art.: Artwork
28
- - Com. R.: Commonsense Reasoning
29
- - Code: Code Reasoning
30
- - Num.: Numerical Calculation
31
- - Tran.: Text Translation
32
- - OCR: OCR