LLM Inference

Source https://github.com/NVIDIA/TensorRT-LLM/tree/main/examples/llm-api/llm_inference.py.

 1### Generate text
 2import tempfile
 3
 4from tensorrt_llm import LLM, SamplingParams
 5
 6# Model could accept HF model name or a path to local HF model.
 7llm = LLM(model="TinyLlama/TinyLlama-1.1B-Chat-v1.0")
 8
 9# You can save the engine to disk and load it back later, the LLM class can accept either a HF model or a TRT-LLM engine.
10llm.save(tempfile.mkdtemp())
11
12# Sample prompts.
13prompts = [
14    "Hello, my name is",
15    "The president of the United States is",
16    "The capital of France is",
17    "The future of AI is",
18]
19
20# Create a sampling params.
21sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
22
23for 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'