pre-commit to format code

.github/workflows/lint.yml

@@ -20,4 +20,4 @@ jobs:
           pip install pre-commit
           pre-commit install
       - name: Linting
-        run: pre-commit run --all-files
+        run: pre-commit run --all-files

.pre-commit-config.yaml

@@ -1,15 +1,15 @@
 exclude: |
   (?x)^(
-      results/|
       scripts/|
-      assets/
+      assets/|
+      vlmeval/config.py
   )
 repos:
   - repo: https://github.com/PyCQA/flake8
     rev: 5.0.4
     hooks:
       - id: flake8
-        args: ["--max-line-length=120", "--ignore=F401,F405,E402"]
+        args: ["--max-line-length=120", "--ignore=F401,F403,F405,E402,E722,E741,W503"]
         exclude: ^configs/
   - repo: https://github.com/pre-commit/mirrors-yapf
     rev: v0.30.0

Custom_Benchmark_and_Model.md

@@ -1,8 +1,8 @@
-# 🛠️ How to implement a new Benchmark / VLM in VLMEvalKit? 
+# 🛠️ How to implement a new Benchmark / VLM in VLMEvalKit?
 
 ## Implement a new benchmark
 
-Currently, we organize a benchmark as one single TSV file. During inference, the data file will be automatically downloaded to `$LMUData` (default path is `$HOME/LMUData`, if not set explicitly). All existing benchmark TSV files are handled by `TSVDataset` implemented in `vlmeval/utils/dataset_config.py`. 
+Currently, we organize a benchmark as one single TSV file. During inference, the data file will be automatically downloaded to `$LMUData` (default path is `$HOME/LMUData`, if not set explicitly). All existing benchmark TSV files are handled by `TSVDataset` implemented in `vlmeval/utils/dataset_config.py`.
 
 | Dataset Name \ Fields  | index | image | image_path | question | hint | multi-choice<br>options | answer | category | l2-category | split |
 | ---------------------- | ----- | ----- | ---------- | -------- | ---- | ----------------------- | ------ | -------- | ----------- | ----- |
@@ -23,14 +23,14 @@ Currently, we organize a benchmark as one single TSV file. During inference, the
 **Intro to some fields:**
 
 - **index:** Integer, Unique for each line in `tsv`
-- **image:** the base64 of the image, you can use APIs implemented in `vlmeval/smp.py` for encoding and decoding: 
+- **image:** the base64 of the image, you can use APIs implemented in `vlmeval/smp.py` for encoding and decoding:
   - Encoding: `encode_image_to_base64 `(for PIL Image) / `encode_image_file_to_base64` (for image file path)
   - Decoding: `decode_base64_to_image`(for PIL Image) / `decode_base64_to_image_file` (for image file path)
 
 Besides, your dataset class **should implement the method `build_prompt(self, line, dataset=None)`**. Given line as a line number or one line in the TSV file, the function yields a dictionary `dict(image=image_path, text=prompt)`, including the image path and the prompt that will be fed to the VLMs.
 
 ## Implement a new model
 
-All existing models are implemented in `vlmeval/vlm`. For a minimal model, your model class **should implement the method** `generate(image_path, prompt, dataset=None)`. In this function, you feed the image and prompt to your VLM and return the VLM prediction (which is a string). The optional argument `dataset` can be used as the flag for the model to switch among various inference strategies. 
+All existing models are implemented in `vlmeval/vlm`. For a minimal model, your model class **should implement the method** `generate(image_path, prompt, dataset=None)`. In this function, you feed the image and prompt to your VLM and return the VLM prediction (which is a string). The optional argument `dataset` can be used as the flag for the model to switch among various inference strategies.
 
-Besides, your model can support custom prompt building by implementing an optional method `build_prompt(line, dataset=None)`. In this function, the line is a dictionary that includes the necessary information of a data sample, while `dataset` can be used as the flag for the model to switch among various prompt building strategies. 
+Besides, your model can support custom prompt building by implementing an optional method `build_prompt(line, dataset=None)`. In this function, the line is a dictionary that includes the necessary information of a data sample, while `dataset` can be used as the flag for the model to switch among various prompt building strategies.

README.md

@@ -16,7 +16,7 @@
 </div>
 
 
-**VLMEvalKit** (the python package name is **vlmeval**) is an **open-source evaluation toolkit** of **large vision-language models (LVLMs)**. It enables **one-command evaluation** of LVLMs on various benchmarks, without the heavy workload of data preparation under multiple repositories. In VLMEvalKit, we adopt **generation-based evaluation** for all LVLMs (obtain the answer via `generate` / `chat`  interface), and provide the evaluation results obtained with both **exact matching** and **LLM(ChatGPT)-based answer extraction**. 
+**VLMEvalKit** (the python package name is **vlmeval**) is an **open-source evaluation toolkit** of **large vision-language models (LVLMs)**. It enables **one-command evaluation** of LVLMs on various benchmarks, without the heavy workload of data preparation under multiple repositories. In VLMEvalKit, we adopt **generation-based evaluation** for all LVLMs (obtain the answer via `generate` / `chat`  interface), and provide the evaluation results obtained with both **exact matching** and **LLM(ChatGPT)-based answer extraction**.
 
 ## 🆕 News
 
@@ -27,7 +27,7 @@
 - **[2024-02-24]** We have supported [**InternVL-Chat Series**](https://github.com/OpenGVLab/InternVL). The models achieve over 80% Top-1 accuracies on MMBench v1.0 [[**Blog**](https://github.com/OpenGVLab/InternVL/blob/main/BLOG.md)]. 🔥🔥🔥
 - **[2024-02-07]** We have supported two new models: [**MiniCPM-V**](https://huggingface.co/openbmb/MiniCPM-V) and [**OmniLMM-12B**](https://huggingface.co/openbmb/OmniLMM-12B). 🔥🔥🔥
 - **[2024-01-30]** We have supported three new models: [**QwenVLMax**](https://huggingface.co/spaces/Qwen/Qwen-VL-Max), [**InternLM-XComposer2-7B**](https://huggingface.co/internlm/internlm-xcomposer2-vl-7b), [**MMAlaya**](https://huggingface.co/DataCanvas/MMAlaya). 🔥🔥🔥
-- **[2024-01-30]** We have merged all performance numbers on our leaderboards into a single json file: [**OpenVLM.json**](http://opencompass.openxlab.space/utils/OpenVLM.json). 
+- **[2024-01-30]** We have merged all performance numbers on our leaderboards into a single json file: [**OpenVLM.json**](http://opencompass.openxlab.space/utils/OpenVLM.json).
 - **[2024-01-27]** We have supported the evaluation of [**MMMU_TEST**](https://mmmu-benchmark.github.io). 🔥🔥🔥
 - **[2024-01-24]** We have supported [**Yi-VL**](https://huggingface.co/01-ai/Yi-VL-6B). 🔥🔥🔥
 
@@ -36,7 +36,7 @@
 
 **The performance numbers on our official multi-modal leaderboards can be downloaded from here!**
 
-[**OpenCompass Multi-Modal Leaderboard**](https://rank.opencompass.org.cn/leaderboard-multimodal): [Download All DETAILED Results](http://opencompass.openxlab.space/utils/OpenVLM.json). 
+[**OpenCompass Multi-Modal Leaderboard**](https://rank.opencompass.org.cn/leaderboard-multimodal): [Download All DETAILED Results](http://opencompass.openxlab.space/utils/OpenVLM.json).
 
 **Supported Dataset**
 
@@ -60,7 +60,7 @@
 | [**OCRBench**](https://github.com/Yuliang-Liu/MultimodalOCR) | OCRBench                                               | ✅         | ✅          | **TBD.**                                                     |
 | [**Core-MM**](https://github.com/core-mm/core-mm)            | CORE_MM                                                | ✅         |            | **N/A**                                                      |
 
-**There are some known issues with VQA tasks like OCRVQA, TextVQA, ChartQA, etc. We will fix them asap.** 
+**There are some known issues with VQA tasks like OCRVQA, TextVQA, ChartQA, etc. We will fix them asap.**
 
 **Supported API Models**
 
@@ -77,9 +77,9 @@
 | [**Monkey**](https://github.com/Yuliang-Liu/Monkey)🚅         | [**EMU2 / EMU2-Chat**](https://github.com/baaivision/Emu)🚅🎞️  | [**Yi-VL-[6B/34B]**](https://huggingface.co/01-ai/Yi-VL-6B)  | [**MMAlaya**](https://huggingface.co/DataCanvas/MMAlaya)🚅    |
 | [**InternLM-XComposer2-7B**](https://huggingface.co/internlm/internlm-xcomposer2-vl-7b)🚅🎞️ | [**MiniCPM-V**](https://huggingface.co/openbmb/MiniCPM-V)🚅   | [**OmniLMM-12B**](https://huggingface.co/openbmb/OmniLMM-12B) | [**InternVL-Chat Series**](https://github.com/OpenGVLab/InternVL)🚅 |
 
-🎞️: Support multiple images as inputs, via the `interleave_generate` interface. 
+🎞️: Support multiple images as inputs, via the `interleave_generate` interface.
 
-🚅: Model can be used without any additional configuration / operation. 
+🚅: Model can be used without any additional configuration / operation.
 
 **Transformers Version Recommendation: ** Note that some VLMs may not be able to run under certain transformer versions, we recommend the following settings to evaluate each VLM:
 
@@ -100,9 +100,9 @@ print(ret)  # There are two apples in the provided images.
 
 ## 🏗️ QuickStart
 
-Before running the evaluation script, you need to **configure** the VLMs and set the model_paths properly. 
+Before running the evaluation script, you need to **configure** the VLMs and set the model_paths properly.
 
-After that, you can use a single script `run.py` to inference and evaluate multiple VLMs and benchmarks at a same time. 
+After that, you can use a single script `run.py` to inference and evaluate multiple VLMs and benchmarks at a same time.
 
 ### Step0. Installation
 
@@ -118,18 +118,18 @@ pip install -e .
 
 Following VLMs require the configuration step:
 
-**Code Preparation & Installation**: InstructBLIP ([LAVIS](https://github.com/salesforce/LAVIS)), LLaVA ([LLaVA](https://github.com/haotian-liu/LLaVA)), MiniGPT-4 ([MiniGPT-4](https://github.com/Vision-CAIR/MiniGPT-4)), mPLUG-Owl2 ([mPLUG-Owl2](https://github.com/X-PLUG/mPLUG-Owl/tree/main/mPLUG-Owl2)), OpenFlamingo-v2 ([OpenFlamingo](https://github.com/mlfoundations/open_flamingo)), PandaGPT-13B ([PandaGPT](https://github.com/yxuansu/PandaGPT)), TransCore-M ([TransCore-M](https://github.com/PCIResearch/TransCore-M)). 
+**Code Preparation & Installation**: InstructBLIP ([LAVIS](https://github.com/salesforce/LAVIS)), LLaVA ([LLaVA](https://github.com/haotian-liu/LLaVA)), MiniGPT-4 ([MiniGPT-4](https://github.com/Vision-CAIR/MiniGPT-4)), mPLUG-Owl2 ([mPLUG-Owl2](https://github.com/X-PLUG/mPLUG-Owl/tree/main/mPLUG-Owl2)), OpenFlamingo-v2 ([OpenFlamingo](https://github.com/mlfoundations/open_flamingo)), PandaGPT-13B ([PandaGPT](https://github.com/yxuansu/PandaGPT)), TransCore-M ([TransCore-M](https://github.com/PCIResearch/TransCore-M)).
 
 **Manual Weight Preparation & Configuration**: InstructBLIP, LLaVA-v1-7B, MiniGPT-4, PandaGPT-13B
 
-### Step2. Evaluation 
+### Step2. Evaluation
 
 We use `run.py` for evaluation. To use the script, you can use `$VLMEvalKit/run.py` or create a soft-link of the script (to use the script anywhere):
 
 **Arguments**
 
-- `--data (list[str])`: Set the dataset names that are supported in VLMEvalKit (defined in `vlmeval/utils/dataset_config.py`). 
-- `--model (list[str])`: Set the VLM names that are supported in VLMEvalKit (defined in `supported_VLM` in `vlmeval/config.py`). 
+- `--data (list[str])`: Set the dataset names that are supported in VLMEvalKit (defined in `vlmeval/utils/dataset_config.py`).
+- `--model (list[str])`: Set the VLM names that are supported in VLMEvalKit (defined in `supported_VLM` in `vlmeval/config.py`).
 - `--mode (str, default to 'all', choices are ['all', 'infer'])`: When `mode` set to "all", will perform both inference and evaluation; when set to "infer", will only perform the inference.
 - `--nproc (int, default to 4)`: The number of threads for OpenAI API calling.
 
@@ -138,24 +138,24 @@ We use `run.py` for evaluation. To use the script, you can use `$VLMEvalKit/run.
 You can run the script with `python` or `torchrun`:
 
 ```bash
-# When running with `python`, only one VLM instance is instantiated, and it might use multiple GPUs (depending on its default behavior). 
+# When running with `python`, only one VLM instance is instantiated, and it might use multiple GPUs (depending on its default behavior).
 # That is recommended for evaluating very large VLMs (like IDEFICS-80B-Instruct).
 
 # IDEFICS-80B-Instruct on MMBench_DEV_EN, MME, and SEEDBench_IMG, Inference and Evalution
-python run.py --data MMBench_DEV_EN MME SEEDBench_IMG --model idefics_80b_instruct --verbose 
+python run.py --data MMBench_DEV_EN MME SEEDBench_IMG --model idefics_80b_instruct --verbose
 # IDEFICS-80B-Instruct on MMBench_DEV_EN, MME, and SEEDBench_IMG, Inference only
 python run.py --data MMBench_DEV_EN MME SEEDBench_IMG --model idefics_80b_instruct --verbose --mode infer
 
-# When running with `torchrun`, one VLM instance is instantiated on each GPU. It can speed up the inference. 
-# However, that is only suitable for VLMs that consume small amounts of GPU memory. 
+# When running with `torchrun`, one VLM instance is instantiated on each GPU. It can speed up the inference.
+# However, that is only suitable for VLMs that consume small amounts of GPU memory.
 
 # IDEFICS-9B-Instruct, Qwen-VL-Chat, mPLUG-Owl2 on MMBench_DEV_EN, MME, and SEEDBench_IMG. On a node with 8 GPU. Inference and Evaluation.
 torchrun --nproc-per-node=8 run.py --data MMBench_DEV_EN MME SEEDBench_IMG --model idefics_80b_instruct qwen_chat mPLUG-Owl2 --verbose
-# Qwen-VL-Chat on MME. On a node with 2 GPU. Inference and Evaluation. 
+# Qwen-VL-Chat on MME. On a node with 2 GPU. Inference and Evaluation.
 torchrun --nproc-per-node=2 run.py --data MME --model qwen_chat --verbose
 ```
 
-The evaluation results will be printed as logs, besides. **Result Files** will also be generated in the directory `$YOUR_WORKING_DIRECTORY/{model_name}`. Files ending with `.csv` contain the evaluated metrics. 
+The evaluation results will be printed as logs, besides. **Result Files** will also be generated in the directory `$YOUR_WORKING_DIRECTORY/{model_name}`. Files ending with `.csv` contain the evaluated metrics.
 
 ## 🛠️ Custom Benchmark or VLM
 
@@ -169,13 +169,13 @@ To implement a custom benchmark or VLM in **VLMEvalKit**, please refer to [Custo
 **The codebase is designed to:**
 
 1. Provide an **easy-to-use**, **opensource evaluation toolkit** to make it convenient for researchers & developers to evaluate existing LVLMs and make evaluation results **easy to reproduce**.
-2. Make it easy for VLM developers to evaluate their own models. To evaluate the VLM on multiple supported benchmarks, one just need to **implement a single `generate` function**, all other workloads (data downloading, data preprocessing, prediction inference, metric calculation) are handled by the codebase. 
+2. Make it easy for VLM developers to evaluate their own models. To evaluate the VLM on multiple supported benchmarks, one just need to **implement a single `generate` function**, all other workloads (data downloading, data preprocessing, prediction inference, metric calculation) are handled by the codebase.
 
 **The codebase is not designed to:**
 
 1. Reproduce the exact accuracy number reported in the original papers of all **3rd party benchmarks**. The reason can be two-fold:
-   1. VLMEvalKit uses **generation-based evaluation** for all VLMs (and optionally with **LLM-based answer extraction**). Meanwhile, some benchmarks may use different approaches (SEEDBench uses PPL-based evaluation, *eg.*). For those benchmarks, we compare both scores in the corresponding result. We encourage developers to support other evaluation paradigms in the codebase. 
-   2. By default, we use the same prompt template for all VLMs to evaluate on a benchmark. Meanwhile, **some VLMs may have their specific prompt templates** (some may not covered by the codebase at this time). We encourage VLM developers to implement their own prompt template in VLMEvalKit, if that is not covered currently. That will help to improve the reproducibility. 
+   1. VLMEvalKit uses **generation-based evaluation** for all VLMs (and optionally with **LLM-based answer extraction**). Meanwhile, some benchmarks may use different approaches (SEEDBench uses PPL-based evaluation, *eg.*). For those benchmarks, we compare both scores in the corresponding result. We encourage developers to support other evaluation paradigms in the codebase.
+   2. By default, we use the same prompt template for all VLMs to evaluate on a benchmark. Meanwhile, **some VLMs may have their specific prompt templates** (some may not covered by the codebase at this time). We encourage VLM developers to implement their own prompt template in VLMEvalKit, if that is not covered currently. That will help to improve the reproducibility.
 
 ## 🖊️ Citation
 
@@ -195,4 +195,4 @@ If you use VLMEvalKit in your research or wish to refer to the published OpenSou
 - [opencompass](https://github.com/open-compass/opencompass/): An LLM evaluation platform, supporting a wide range of models (LLaMA, LLaMa2, ChatGLM2, ChatGPT, Claude, etc) over 50+ datasets.
 - [MMBench](https://github.com/open-compass/MMBench/): Official Repo of "MMBench: Is Your Multi-modal Model an All-around Player?"
 - [BotChat](https://github.com/open-compass/BotChat/): Evaluating LLMs' multi-round chatting capability.
-- [LawBench](https://github.com/open-compass/LawBench): Benchmarking Legal Knowledge of Large Language Models. 
+- [LawBench](https://github.com/open-compass/LawBench): Benchmarking Legal Knowledge of Large Language Models.

requirements.txt

@@ -1,28 +1,28 @@
+einops
+gradio==4.15.0
+huggingface_hub
+matplotlib
 numpy>=1.23.4
+omegaconf
 openai==1.3.5
-requests
-tqdm
+opencv-python>=4.4.0.46
+openpyxl
 pandas>=1.5.3
-gradio==4.15.0
-tiktoken
-rich
+pillow
 portalocker
-torch>=2.0.1
+pycocoevalcap
+requests
+rich
+seaborn
+sentencepiece
+sty
+tabulate
+tiktoken
 timeout-decorator
+torch>=2.0.1
+tqdm
 transformers
-opencv-python>=4.4.0.46
 typing_extensions==4.7.1
-pillow
-omegaconf
-matplotlib
-einops
-sentencepiece
-sty
-xtuner
-huggingface_hub
 visual_genome
-pycocoevalcap
-openpyxl
-seaborn
-tabulate
-xlsxwriter
+xlsxwriter
+xtuner