Files
sglang/benchmark/multi_document_qa/README.md
Lianmin Zheng 22085081bb release initial code
Co-authored-by: Ying Sheng <sqy1415@gmail.com>
Co-authored-by: Liangsheng Yin <hnyls2002@gmail.com>
Co-authored-by: Zhiqiang Xie <xiezhq@stanford.edu>
Co-authored-by: parasol-aser <3848358+parasol-aser@users.noreply.github.com>
Co-authored-by: LiviaSun <33578456+ChuyueSun@users.noreply.github.com>
Co-authored-by: Cody Yu <hao.yu.cody@gmail.com>
2024-01-08 04:37:50 +00:00

48 lines
929 B
Markdown

## Run benchmark
### Benchmark sglang
```
python3 -m sglang.launch_server --model-path codellama/CodeLlama-7b-instruct-hf --port 30000
```
```
python3 bench_sglang.py --num-questions 10 --parallel 1
```
### Benchmark vllm
```
python3 -m vllm.entrypoints.api_server --tokenizer-mode auto --model codellama/CodeLlama-7b-instruct-hf --disable-log-requests --port 21000 --gpu 0.97
```
```
python3 bench_other.py --backend vllm --num-questions 64
```
### Benchmark guidance
```
python3 bench_other.py --backend guidance --num-questions 32 --parallel 1
```
### Build dataset
```
pip install PyPDF2
python3 build_dataset.py
```
```python
import PyPDF2
with open('llama2.pdf', 'rb') as file:
reader = PyPDF2.PdfReader(file)
text = ''
for page_num in range(len(reader.pages)):
text += reader.pages[page_num].extract_text()
with open('output.txt', 'w') as text_file:
text_file.write(text)
```