BigEval

Warning

The project is under active development. Please check back later for more updates.

Warning

Please use WildCode with caution. Different from EvalPlus, WildCode has a much less constrained execution environment to support tasks with diverse library dependencies. This may lead to security risks. We recommend using a sandbox such as Docker to run the evaluation.

🌳About • 🔥Quick Start • 💻LLM code • 🔍Failure inspection • 🐞Known issues • 📜Citation • 🙏Acknowledgement

About

WildCodeBench

WildCodeBench is a rigorous benchmark for code generation with realistic constraints in the wild. It aims to evaluate the true programming capabilities of large language models (LLMs) in a more realistic setting. The benchmark is designed for HumanEval-like function-level code generation tasks, but with much more fine-grained descriptions and diverse tool use.

WildCode

To facilitate the evaluation of LLMs on WildCodeBench, we provide a Python package wild-code that includes the dataset, generation scripts, and evaluation scripts. The package is built on top of the EvalPlus framework, which is a flexible and extensible evaluation framework for code generation tasks.

Why WildCode?

WildCode is a rigorous evaluation framework for LLM4Code, with:

✨ Precise evaluation & ranking: See our leaderboard for latest LLM rankings before & after rigorous evaluation.
✨ Pre-generated samples: WildCode accelerates code intelligence research by open-sourcing LLM-generated samples for various models -- no need to re-run the expensive benchmarks!

Main Differences from EvalPlus

We inherit the design of the EvalPlus framework, which is a flexible and extensible evaluation framework for code generation tasks. However, WildCode has the following differences:

Execution Environment: The execution environment in WildCode is less bounded than EvalPlus to support tasks with diverse library dependencies.
Test Evaluation: WildCode relies on unittest for evaluating the generated code, which is more suitable for the test harness in WildCodeBench.

🔥 Quick Start

Tip

WildCode ❤️ bigcode-evaluation-harness! WildCodeBench will be integrated to bigcode-evaluation-harness, and you can also run it there!

To get started, please first set up the environment:

pip install wild-code --upgrade

⏬ Install nightly version :: click to expand ::

pip install "git+https://github.com/bigcode-project/wild-code.git" --upgrade

⏬ Using WildCode as a local repo? :: click to expand ::

git clone https://github.com/bigcode-project/wild-code.git
cd wild-code
export PYTHONPATH=$PYTHONPATH:$(pwd)
pip install -e .

Code Generation

You are suggested to use flash-attn for generating code samples.

pip install -U flash-attn

To generate code samples from a model, you can use the following command:

wildcode.generate \
    --model [model_name] \
    --dataset [wildcodebench] \
    --nl2code [False|True] \
    --greedy \
    --bs [bs] \
    --temperature [temp] \
    --n_samples [n_samples] \
    --resume \
    --backend [vllm|hf|openai|mistral|anthropic|google] \
    --tp [gpu_number]

The generated code samples will be stored in a file named [model_name]--wildcodebench-[nl2c|c2c]--[backend]-[temp]-[n_samples].jsonl. Alternatively, you can use the following command to utilize our pre-built docker images for generating code samples:

docker run --gpus '"device=$CUDA_VISIBLE_DEVICES"' -v $(pwd):/wildcode -t codeeval/code-eval-generate-cu11:25052024 --model [model_name] \ 
    --dataset [wildcodebench] \
    --nl2code [False|True] \
    --greedy \
    --bs [bs] \   
    --temperature [temp] \
    --n_samples [n_samples] \
    --resume \
    --backend [vllm|hf|openai|mistral|anthropic|google] \
    --tp [gpu_number]

We make available cuda 11.8.0 and cuda 12.1.1 pre-built docker images with the Dockerfiles available in the Docker directory.

If you wish to use gated or private HuggingFace models and datasets, you need to build the container yourself with --build-arg flags as follows:

docker build --build-arg HF_TOKEN=<YOUR_HF_TOKEN> -t codeeval/code-eval-generate-cu11:latest - < Docker/Generate_Cuda11.Dockerfile

Following which, you can run the built container as shown in above.

🤔 Structure of `problem`? :: click to expand ::

task_id is the identifier string for the task
entry_point is the name of the function
prompt is the function signature with docstring
instruction is the instruction for the task completion

canonical_solution is the ground-truth implementation
test is the unittest test case

Note

Expected Schema of [model_name]--wildcodebench-[task]--[backend]-[temp]-[n_samples].jsonl

task_id: Task ID, which are the keys of get_wildcodebench()
solution (optional): Self-contained solution (usually including the prompt)
- Example: {"task_id": "WildCodeBench/?", "solution": "def f():\n return 1"}

Code Post-processing

LLM-generated text may not be compilable code for including natural language lines or incomplete extra code. We provide a tool namely wildcode.sanitize to clean up the code:

# 💡 If you are storing codes in jsonl:
wildcode.sanitize --samples samples.jsonl
# Sanitized code will be produced to `samples-sanitized.jsonl`

# 💡 If you are storing codes in directories:
wildcode.sanitize --samples /path/to/vicuna-[??]b_temp_[??]
# Sanitized code will be produced to `/path/to/vicuna-[??]b_temp_[??]-sanitized`

🔎 Checking the compilability of post-processed code:: click to expand ::

To double-check the post-processing results, you can use wildcode.syncheck to check the code validity before and after sanitization, which will print erroneous code snippets and why they are wrong:

# 💡 If you are storing codes in jsonl:
wildcode.syncheck --samples samples.jsonl --dataset [wildcodebench]

# 💡 If you are storing codes in directories:
wildcode.syncheck --samples /path/to/vicuna-[??]b_temp_[??] --dataset [wildcodebench]

Code Evaluation

You are strongly recommended to use a sandbox such as docker:

# mount the current directory to the container
docker run -v $(pwd):/wildcode codeeval/code-eval-evaluate:latest --dataset wildcodebench --samples samples.jsonl
# ...Or locally ⚠️
wildcode.evaluate --dataset wildcodebench --samples samples.jsonl

...Or if you want to try it locally regardless of the risks ⚠️:

First, install the dependencies for WildCodeBench:

pip install -r https://raw.githubusercontent.com/bigcode-project/wildcodebench-annotation/main/requirements.txt

Then, run the evaluation:

wildcode.evaluate --dataset [wildcodebench] --samples samples.jsonl

Tip

Do you use a very slow machine?

LLM solutions are regarded as failed on timeout (and OOM etc.). Specifically, we set the dynamic timeout based on the ground-truth solution's runtime.

Additionally, you are NOT encouraged to make your test-bed over stressed while running evaluation. For example, using --parallel 64 on a 4-core machine or doing something else during evaluation are bad ideas...

⌨️ More command-line flags :: click to expand ::

--parallel: by default half of the cores

The output should be like (below is GPT-4 greedy decoding example):

Asserting the groundtruth...
Expected outputs computed in 1200.0 seconds
Reading samples...
1140it [00:00, 1901.64it/s]
Evaluating samples...
100%|██████████████████████████████████████████| 1140/1140 [19:53<00:00, 6.75it/s]
wildcodebench
{'pass@1': 0.568}

The "k" includes [1, 5, 10] where k values <= the sample size will be used
A cache file named like samples_eval_results.jsonl will be cached. Remove it to re-run the evaluation

🤔 How long it would take? :: click to expand ::

If you do greedy decoding where there is only one sample for each task, the evaluation should take just a few seconds. When running 1 sample x 964 tasks x all tests, it can take around ??-?? minutes by using --parallel 64 and --test-details. Here are some tips to speed up the evaluation:

Use --parallel $(nproc)
Use our pre-evaluated results (see LLM-generated code)

Failure Inspection

You can inspect the failed samples by using the following command:

wildcode.inspect --dataset [wildcodebench] --eval-results sample-sanitized_eval_results.json --in-place

Full script

We provide a sample script to run the full pipeline:

bash run.sh

💻 LLM-generated Code

We will share pre-generated code samples from LLMs we have evaluated:

Known Issues

We notice that some tasks heavily use memory for scientific modeling during testing. It will lead to timeout issues on some machines. If you get an error message like Check failed: ret == 0 (11 vs. 0)Thread creation via pthread_create() failed. in Tensorflow, it is very likely due to the memory issue. Try to allocate more memory to the process or reduce the number of parallel processes.
Due to the flakes in the evaluation, the execution results may vary slightly (~0.5%) between runs. We are working on improving the evaluation stability.

📜 Citation

🙏 Acknowledgement

EvalPlus

License

iNeil77/code-eval

ErrorLooks like something went wrong!

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases 3

Packages 0

Languages

Error
Looks like something went wrong!

Packages