Automatic Parallelism with LLM#
Source NVIDIA/TensorRT-LLM.
1### Automatic Parallelism with LLM
2from tensorrt_llm import SamplingParams
3from tensorrt_llm._tensorrt_engine import LLM
4
5
6def main():
7 llm = LLM(
8 model="TinyLlama/TinyLlama-1.1B-Chat-v1.0",
9
10 # Enable auto parallelism
11 auto_parallel=True,
12 auto_parallel_world_size=2)
13
14 prompts = [
15 "Hello, my name is",
16 "The president of the United States is",
17 "The capital of France is",
18 "The future of AI is",
19 ]
20
21 sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
22
23 for output in llm.generate(prompts, sampling_params):
24 print(
25 f"Prompt: {output.prompt!r}, Generated text: {output.outputs[0].text!r}"
26 )
27
28 # Got output like
29 # Prompt: 'Hello, my name is', Generated text: '\n\nJane Smith. I am a student pursuing my degree in Computer Science at [university]. I enjoy learning new things, especially technology and programming'
30 # Prompt: 'The president of the United States is', Generated text: 'likely to nominate a new Supreme Court justice to fill the seat vacated by the death of Antonin Scalia. The Senate should vote to confirm the'
31 # Prompt: 'The capital of France is', Generated text: 'Paris.'
32 # Prompt: 'The future of AI is', Generated text: 'an exciting time for us. We are constantly researching, developing, and improving our platform to create the most advanced and efficient model available. We are'
33
34
35if __name__ == '__main__':
36 main()