nielsr HF Staff commited on
Commit
e059846
·
verified ·
1 Parent(s): 8f9e381

Add task category and sample usage

Browse files

This PR adds the `text-generation` task category to the dataset metadata so the dataset is discoverable when filtering by task. It also adds a short sample usage section based on the official evaluation instructions from the associated GitHub repository.

Files changed (1) hide show
  1. README.md +24 -1
README.md CHANGED
@@ -31,6 +31,8 @@ configs:
31
  path: data/task_no_tie-*
32
  - split: benchmark_tie
33
  path: data/benchmark_tie-*
 
 
34
  ---
35
 
36
  # BigCodeBench
@@ -43,4 +45,25 @@ configs:
43
  - **Homepage:** https://bigcode-bench.github.io/
44
  - **Repository:** https://github.com/bigcode-project/bigcodebench
45
  - **Paper:** [Link](https://arxiv.org/abs/2406.15877)
46
- - **Point of Contact:** contact@bigcode-project.org terry.zhuo@monash.edu
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
31
  path: data/task_no_tie-*
32
  - split: benchmark_tie
33
  path: data/benchmark_tie-*
34
+ task_categories:
35
+ - text-generation
36
  ---
37
 
38
  # BigCodeBench
 
45
  - **Homepage:** https://bigcode-bench.github.io/
46
  - **Repository:** https://github.com/bigcode-project/bigcodebench
47
  - **Paper:** [Link](https://arxiv.org/abs/2406.15877)
48
+ - **Point of Contact:** contact@bigcode-project.org terry.zhuo@monash.edu
49
+
50
+ ## Sample Usage
51
+
52
+ Install the BigCodeBench package:
53
+
54
+ ```bash
55
+ pip install bigcodebench --upgrade
56
+ ```
57
+
58
+ Evaluate a model on the benchmark via the remote evaluation API:
59
+
60
+ ```bash
61
+ bigcodebench.evaluate \
62
+ --model meta-llama/Meta-Llama-3.1-8B-Instruct \
63
+ --execution [e2b|gradio|local] \
64
+ --split [complete|instruct] \
65
+ --subset [full|hard] \
66
+ --backend [vllm|openai|anthropic|google|mistral|hf|hf-inference]
67
+ ```
68
+
69
+ Refer to the [Github repository](https://github.com/bigcode-project/bigcodebench) for advanced usage.