Skip to main content

大模型性能压测可视化

Project description

Model inference performance stress test

evalscope-perf help

evalscope-perf --help
 Usage: evalscope-perf [OPTIONS] URL MODEL DATASET_PATH                                                                     
                                                                                                                            
╭─ Arguments ──────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ *    url               TEXT  OpenAI URL [default: None] [required]
│ *    model             TEXT  模型名称 [default: None] [required]
│ *    dataset_path      TEXT  数据集路径 [default: None] [required]
╰──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ Options ────────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ --max-prompt-length         INTEGER  最大提示长度 [default: 256]
│ --stop                      TEXT     停止标记 [default: <|im_end|>]
│ --read-timeout              INTEGER  读取超时 [default: 30]
│ --parallels                 INTEGER  并行数 [default: 1]
│ --n                         INTEGER  请求数 [default: 1]
│ --install-completion                 Install completion for the current shell.
│ --show-completion                    Show completion for the current shell, to copy it or customize the installation.
│ --help                               Show this message and exit.
╰──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

Example

Nvidia T4X4 - XInference(vLLM) - Qwen1.5-7B-Chat

evalscope-perf http://172.16.33.66:9997/v1/chat/completions gpt-4-32k \
    ./datasets/open_qa.jsonl \
    --parallels 4 \
    --parallels 8 \
    --parallels 16 \
    --n 20

Ascend 910B4X8 - MindIE - DeepSeek-Coder-6.7B-Instruct

evalscope-perf http://127.0.0.1:1025/v1/chat/completions qwen \
    ./datasets/Codefuse-Evol-Instruct-Clean-data.jsonl \
    --parallels 32 \
    --parallels 64 \
    --parallels 100 \
    --parallels 128 \
    --parallels 150 \
    --parallels 200 \
    --parallels 256 \
    --parallels 300 \
    --parallels 400 \
    --parallels 500 \
    --parallels 600 \
    --parallels 700 \
    --parallels 800 \
    --parallels 900 \
    --parallels 1000 \
    --n 2000

evalscope-perf http://127.0.0.1:11434/v1/chat/completions qwen2.5:7b
./datasets/open_qa.jsonl
--parallels 1
--parallels 2
--n 4

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

evalscope_perf-1.0.0.tar.gz (21.1 MB view details)

Uploaded Source

File details

Details for the file evalscope_perf-1.0.0.tar.gz.

File metadata

  • Download URL: evalscope_perf-1.0.0.tar.gz
  • Upload date:
  • Size: 21.1 MB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.1.0 CPython/3.10.9

File hashes

Hashes for evalscope_perf-1.0.0.tar.gz
Algorithm Hash digest
SHA256 c04cbf4f5985f5b3b0bedc22a1e82ccd7a864180cf4c9d081628d424660965b3
MD5 bb957b0e5ef3bce3facb31f799d2cfcb
BLAKE2b-256 804b0d944b37437aa4bc16043aff735337b914c6fb23679620d8b073f9765e27

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page