* ci: bench: support sse and fix prompt processing time server: add tokens usage in stream mode * ci: bench: README.md EOL * ci: bench: remove total pp and tg as it is not accurate * ci: bench: fix case when there is no token generated * ci: bench: change to the 95 percentile for pp and tg as it is closer to what the server exports in metrics * ci: bench: fix finish reason rate
4.2 KiB
Server benchmark tools
Benchmark is using k6.
Install k6 and sse extension
SSE is not supported by default in k6, you have to build k6 with the xk6-sse extension.
Example:
go install go.k6.io/xk6/cmd/xk6@latest
xk6 build master \
--with github.com/phymbert/xk6-sse
Download a dataset
This dataset was originally proposed in vLLM benchmarks.
wget https://huggingface.co/datasets/anon8231489123/ShareGPT_Vicuna_unfiltered/resolve/main/ShareGPT_V3_unfiltered_cleaned_split.json
Download a model
Example for PHI-2
../../../scripts/hf.sh --repo ggml-org/models --file phi-2/ggml-model-q4_0.gguf
Start the server
The server must answer OAI Chat completion requests on http://localhost:8080/v1
or according to the environment variable SERVER_BENCH_URL
.
Example:
server --host localhost --port 8080 \
--model ggml-model-q4_0.gguf \
--cont-batching \
--metrics \
--parallel 8 \
--batch-size 512 \
--ctx-size 4096 \
--log-format text \
-ngl 33
Run the benchmark
For 500 chat completions request with 8 concurrent users during maximum 10 minutes, run:
./k6 run script.js --duration 10m --iterations 500 --vus 8
The benchmark values can be overridden with:
SERVER_BENCH_URL
server url prefix for chat completions, defaulthttp://localhost:8080/v1
SERVER_BENCH_N_PROMPTS
total prompts to randomly select in the benchmark, default480
SERVER_BENCH_MODEL_ALIAS
model alias to pass in the completion request, defaultmy-model
SERVER_BENCH_MAX_TOKENS
max tokens to predict, default:512
SERVER_BENCH_DATASET
path to the benchmark dataset fileSERVER_BENCH_MAX_PROMPT_TOKENS
maximum prompt tokens to filter out in the dataset: default1024
SERVER_BENCH_MAX_CONTEXT
maximum context size of the completions request to filter out in the dataset: prompt + predicted tokens, default2048
Note: the local tokenizer is just a string space split, real number of tokens will differ.
Or with k6 options:
SERVER_BENCH_N_PROMPTS=500 k6 run script.js --duration 10m --iterations 500 --vus 8
To debug http request use --http-debug="full"
.
Metrics
Following metrics are available computed from the OAI chat completions response usage
:
llamacpp_tokens_second
Trend ofusage.total_tokens / request duration
llamacpp_prompt_tokens
Trend ofusage.prompt_tokens
llamacpp_prompt_tokens_total_counter
Counter ofusage.prompt_tokens
llamacpp_completion_tokens
Trend ofusage.completion_tokens
llamacpp_completion_tokens_total_counter
Counter ofusage.completion_tokens
llamacpp_completions_truncated_rate
Rate of completions truncated, i.e. iffinish_reason === 'length'
llamacpp_completions_stop_rate
Rate of completions stopped by the model, i.e. iffinish_reason === 'stop'
The script will fail if too many completions are truncated, see llamacpp_completions_truncated_rate
.
K6 metrics might be compared against server metrics, with:
curl http://localhost:8080/metrics
Using the CI python script
The bench.py
script does several steps:
- start the server
- define good variable for k6
- run k6 script
- extract metrics from prometheus
It aims to be used in the CI, but you can run it manually:
LLAMA_SERVER_BIN_PATH=../../../cmake-build-release/bin/server python bench.py \
--runner-label local \
--name local \
--branch `git rev-parse --abbrev-ref HEAD` \
--commit `git rev-parse HEAD` \
--scenario script.js \
--duration 5m \
--hf-repo ggml-org/models \
--hf-file phi-2/ggml-model-q4_0.gguf \
--model-path-prefix models \
--parallel 4 \
-ngl 33 \
--batch-size 2048 \
--ubatch-size 256 \
--ctx-size 4096 \
--n-prompts 200 \
--max-prompt-tokens 256 \
--max-tokens 256