@tduncan / Nvidia-Optimization-Stack-VLLM
Code Commits Issues Pulls Cloud Agent CI/CD
main
main
py 102 lines 6.41 KB
Raw
1 # SPDX-License-Identifier: Apache-2.0
2 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
3 """Demonstration script for Automatic Prefix Caching (APC) in vLLM.
4
5 Automatic Prefix Caching (APC) allows the vLLM engine to reuse cached
6 KV (key-value) pairs from previous prompts if a new query shares the same
7 prefix. This reduces redundant computation and improves inference speed.
8
9 To enable APC, set `enable_prefix_caching=True` when initializing the
10 vLLM engine.
11
12 This script uses a long Markdown table as the shared prompt prefix and
13 compares the generation time for two queries that share the same prefix
14 but ask different questions.
15
16 Run:
17 python examples/features/automatic_prefix_caching/automatic_prefix_caching_offline.py
18 """
19
20 import time
21
22 from vllm import LLM, SamplingParams
23
24 # ruff: noqa: E501
25 # A prompt containing a large markdown table. The table is randomly generated by GPT-4.
26 LONG_PROMPT = (
27 "You are a helpful assistant in recognizes the content of tables in markdown format. Here is a table as follows.\n# Table\n"
28 """
29 | ID | Name | Age | Occupation | Country | Email | Phone Number | Address |
30 |-----|---------------|-----|---------------|---------------|------------------------|----------------|------------------------------|
31 | 1 | John Doe | 29 | Engineer | USA | john.doe@example.com | 555-1234 | 123 Elm St, Springfield, IL |
32 | 2 | Jane Smith | 34 | Doctor | Canada | jane.smith@example.com | 555-5678 | 456 Oak St, Toronto, ON |
33 | 3 | Alice Johnson | 27 | Teacher | UK | alice.j@example.com | 555-8765 | 789 Pine St, London, UK |
34 | 4 | Bob Brown | 45 | Artist | Australia | bob.b@example.com | 555-4321 | 321 Maple St, Sydney, NSW |
35 | 5 | Carol White | 31 | Scientist | New Zealand | carol.w@example.com | 555-6789 | 654 Birch St, Wellington, NZ |
36 | 6 | Dave Green | 28 | Lawyer | Ireland | dave.g@example.com | 555-3456 | 987 Cedar St, Dublin, IE |
37 | 7 | Emma Black | 40 | Musician | USA | emma.b@example.com | 555-1111 | 246 Ash St, New York, NY |
38 | 8 | Frank Blue | 37 | Chef | Canada | frank.b@example.com | 555-2222 | 135 Spruce St, Vancouver, BC |
39 | 9 | Grace Yellow | 50 | Engineer | UK | grace.y@example.com | 555-3333 | 864 Fir St, Manchester, UK |
40 | 10 | Henry Violet | 32 | Artist | Australia | henry.v@example.com | 555-4444 | 753 Willow St, Melbourne, VIC|
41 | 11 | Irene Orange | 26 | Scientist | New Zealand | irene.o@example.com | 555-5555 | 912 Poplar St, Auckland, NZ |
42 | 12 | Jack Indigo | 38 | Teacher | Ireland | jack.i@example.com | 555-6666 | 159 Elm St, Cork, IE |
43 | 13 | Karen Red | 41 | Lawyer | USA | karen.r@example.com | 555-7777 | 357 Cedar St, Boston, MA |
44 | 14 | Leo Brown | 30 | Chef | Canada | leo.b@example.com | 555-8888 | 246 Oak St, Calgary, AB |
45 | 15 | Mia Green | 33 | Musician | UK | mia.g@example.com | 555-9999 | 975 Pine St, Edinburgh, UK |
46 | 16 | Noah Yellow | 29 | Doctor | Australia | noah.y@example.com | 555-0000 | 864 Birch St, Brisbane, QLD |
47 | 17 | Olivia Blue | 35 | Engineer | New Zealand | olivia.b@example.com | 555-1212 | 753 Maple St, Hamilton, NZ |
48 | 18 | Peter Black | 42 | Artist | Ireland | peter.b@example.com | 555-3434 | 912 Fir St, Limerick, IE |
49 | 19 | Quinn White | 28 | Scientist | USA | quinn.w@example.com | 555-5656 | 159 Willow St, Seattle, WA |
50 | 20 | Rachel Red | 31 | Teacher | Canada | rachel.r@example.com | 555-7878 | 357 Poplar St, Ottawa, ON |
51 | 21 | Steve Green | 44 | Lawyer | UK | steve.g@example.com | 555-9090 | 753 Elm St, Birmingham, UK |
52 | 22 | Tina Blue | 36 | Musician | Australia | tina.b@example.com | 555-1213 | 864 Cedar St, Perth, WA |
53 | 23 | Umar Black | 39 | Chef | New Zealand | umar.b@example.com | 555-3435 | 975 Spruce St, Christchurch, NZ|
54 | 24 | Victor Yellow | 43 | Engineer | Ireland | victor.y@example.com | 555-5657 | 246 Willow St, Galway, IE |
55 | 25 | Wendy Orange | 27 | Artist | USA | wendy.o@example.com | 555-7879 | 135 Elm St, Denver, CO |
56 | 26 | Xavier Green | 34 | Scientist | Canada | xavier.g@example.com | 555-9091 | 357 Oak St, Montreal, QC |
57 | 27 | Yara Red | 41 | Teacher | UK | yara.r@example.com | 555-1214 | 975 Pine St, Leeds, UK |
58 | 28 | Zack Blue | 30 | Lawyer | Australia | zack.b@example.com | 555-3436 | 135 Birch St, Adelaide, SA |
59 | 29 | Amy White | 33 | Musician | New Zealand | amy.w@example.com | 555-5658 | 159 Maple St, Wellington, NZ |
60 | 30 | Ben Black | 38 | Chef | Ireland | ben.b@example.com | 555-7870 | 246 Fir St, Waterford, IE |
61 """
62 )
63
64
65 def get_generation_time(llm, sampling_params, prompts):
66 # time the generation
67 start_time = time.time()
68 output = llm.generate(prompts, sampling_params=sampling_params)
69 end_time = time.time()
70 # print the output and generation time
71 print("-" * 30)
72 print(f"Output: {output[0].outputs[0].text}")
73 print(f"Generation time: {end_time - start_time} seconds.")
74 print("-" * 30)
75
76
77 def main():
78 # set enable_prefix_caching=True to enable APC
79 llm = LLM(model="lmsys/longchat-13b-16k", enable_prefix_caching=True)
80
81 sampling_params = SamplingParams(temperature=0, max_tokens=100)
82
83 # Querying the age of John Doe
84 get_generation_time(
85 llm,
86 sampling_params,
87 LONG_PROMPT
88 + "Question: what is the age of John Doe? Your answer: The age of John Doe is ",
89 )
90
91 # Querying the age of Zack Blue
92 # This query will be faster since vllm avoids computing the KV cache of LONG_PROMPT again.
93 get_generation_time(
94 llm,
95 sampling_params,
96 LONG_PROMPT
97 + "Question: what is the age of Zack Blue? Your answer: The age of Zack Blue is ",
98 )
99
100
101 if __name__ == "__main__":
102 main()