update readme

This commit is contained in:
2025-09-08 18:13:58 +08:00
parent 4363025bde
commit b4fd32ac74
3 changed files with 28 additions and 1 deletions

View File

@@ -3,4 +3,4 @@ FROM git.modelhub.org.cn:9443/enginex-ascend/vllm-ascend:v0.10.0rc1
WORKDIR /workspace
RUN pip install sentence-transformers
COPY main.py dataset.json /workspace/
COPY main.py test.sh dataset.json /workspace/

23
README.md Normal file
View File

@@ -0,0 +1,23 @@
## Quickstart
### 构建镜像
```bash
docker build -t feature:v0.1 .
```
### 模型下载
模型地址https://modelscope.cn/models/BAAI/bge-large-zh-v1.5
并放到目录:`/mnt/contest_ceph/zhanghao/models/BAAI/bge-large-zh-v1.5`(如更改目录,请修改后面的执行脚本中的模型路径)
### 测试程序
1. 准备输入数据集,可以参考示例`dataset.json`
2. 在docker镜像里运行测试程序会根据`dataset.json`内容计算每个句子的embedding同时计算所有句子的两两相似度结果保存在`output.json`
```bash
./run_in_docker.sh
```
## 测试结果
| | A100 平均生成时间(秒) | 昇腾910B 平均生成时间(秒) |
|------|-------------------------|----------------------------|
| 时间 | 0.0032 | 0.0138 |

4
run_in_docker.sh Executable file
View File

@@ -0,0 +1,4 @@
#! /usr/bin/env bash
image=feature:v0.1
device=1
docker run -v `pwd`:/workspace -e ASCEND_VISIBLE_DEVICES=$device -e NPU_VISIBLE_DEVICES=${device} --device /dev/davinci$device:/dev/davinci0 --device /dev/davinci_manager --device /dev/devmm_svm --device /dev/hisi_hdc -v /mnt:/mnt -v /usr/local/dcmi:/usr/local/dcmi -v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi -v /usr/local/Ascend/driver/lib64/:/usr/local/Ascend/driver/lib64/ -v /usr/local/Ascend/driver/version.info:/usr/local/Ascend/driver/version.info -v /etc/ascend_install.info:/etc/ascend_install.info --privileged --entrypoint bash $image test.sh