Skip to content

Commit 7143721

Browse files
committed
fix
Signed-off-by: guangli.bao <[email protected]>
1 parent 11ee11b commit 7143721

File tree

1 file changed

+16
-16
lines changed

1 file changed

+16
-16
lines changed

docs/examples/benchmark_practice_on_drun.md

Lines changed: 16 additions & 16 deletions
Original file line numberDiff line numberDiff line change
@@ -4,16 +4,16 @@
44

55
## Getting Started
66

7-
## 📦 1. Benchmark Testing Environment Setup
7+
### 📦 1. Benchmark Testing Environment Setup
88

9-
### 1.1 Create a Conda Environment (recommended)
9+
#### 1.1 Create a Conda Environment (recommended)
1010

1111
```bash
1212
conda create -n guidellm-bench python=3.11 -y
1313
conda activate guidellm-bench
14-
```
14+
````
1515

16-
### 1.2 Install Dependencies
16+
#### 1.2 Install Dependencies
1717

1818
```bash
1919
git clone https://github.com/vllm-project/guidellm.git
@@ -23,33 +23,33 @@ pip install guidellm
2323

2424
For more detailed instructions, refer to [GuideLLM README](https://github.com/vllm-project/guidellm/blob/main/README.md).
2525

26-
### 1.3 Verify Installation
26+
#### 1.3 Verify Installation
2727

2828
```bash
2929
guidellm --help
3030
```
3131

32-
### 1.4 Apply for account and api_key in D.run
32+
#### 1.4 Apply for Account and API Key in D.run
3333

3434
Firstly, register an account, refer to [D.run Registration](https://docs.d.run/en/#register-account); then, create an API key, refer to [D.run API Key](https://docs.d.run/en/#register-account); finally, charge your account at [D.run Account Management](https://docs.d.run/en/#register-account).
3535

36-
### 1.5 Chat with model in D.run
36+
#### 1.5 Chat with Model in D.run
3737

38-
Check if you can use chat model in D.run
38+
Check if you can use the chat model in D.run.
3939

4040
![alt text](image.png)
4141

42-
### 1.6 Find out the http request URL and body
42+
#### 1.6 Find Out the HTTP Request URL and Body
4343

44-
Use Developer Tool in Chrome browser or F12 to open Network, then chat with LLM model to catch http request URL and body
44+
Use the Developer Tool in Chrome browser or press F12 to open Network, then chat with the LLM model to capture the HTTP request URL and body.
4545

4646
![alt text](image-1.png)
4747

4848
![alt text](image-2.png)
4949

5050
In this request, the vllm backend service URL is `https://chat.d.run`; vllm model is `public/qwen2.5-72b-instruct-awq`. These two pieces of information will be used in the following benchmark command.
5151

52-
### 1.7 Download one chat dataset from modelscope
52+
#### 1.7 Download a Chat Dataset from Modelscope
5353

5454
Download the chat dataset JSON file `Open-Source-Meeseeks-high-quality.json` from [Modelscope - Meeseeks](https://modelscope.cn/datasets/meituan/Meeseeks/files).
5555

@@ -76,10 +76,10 @@ guidellm benchmark \
7676

7777
![alt text](image-4.png)
7878

79-
After benchmark completion, key results are clear and straight, e.g.:
79+
After the benchmark completes, key results are clear and straightforward, such as:
8080

81-
* `TTFT`: time to first token
82-
* `TPOT`: Time Per Output Token
83-
* `ITL`: inter-token latency
81+
* **`TTFT`**: Time to First Token
82+
* **`TPOT`**: Time Per Output Token
83+
* **`ITL`**: Inter-Token Latency
8484

85-
First benchmark testing is complated.
85+
The first benchmark test is complete.

0 commit comments

Comments
 (0)