LLM API with TensorRT Engine#

A simple inference example with TinyLlama using the LLM API:

 1from tensorrt_llm import BuildConfig, SamplingParams
 2from tensorrt_llm._tensorrt_engine import LLM  # NOTE the change
 3
 4
 5def main():
 6
 7    build_config = BuildConfig()
 8    build_config.max_batch_size = 256
 9    build_config.max_num_tokens = 1024
10
11    # Model could accept HF model name, a path to local HF model,
12    # or TensorRT Model Optimizer's quantized checkpoints like nvidia/Llama-3.1-8B-Instruct-FP8 on HF.
13    llm = LLM(model="TinyLlama/TinyLlama-1.1B-Chat-v1.0",
14              build_config=build_config)
15
16    # Sample prompts.
17    prompts = [
18        "Hello, my name is",
19        "The capital of France is",
20        "The future of AI is",
21    ]
22
23    # Create a sampling params.
24    sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
25
26    for output in llm.generate(prompts, sampling_params):
27        print(
28            f"Prompt: {output.prompt!r}, Generated text: {output.outputs[0].text!r}"
29        )
30
31    # Got output like
32    # 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'
33    # 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'
34    # Prompt: 'The capital of France is', Generated text: 'Paris.'
35    # 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'
36
37
38if __name__ == '__main__':
39    main()

For more advanced usage including distributed inference, multimodal, and speculative decoding, please refer to this README.