Installing on Grace Hopper#

  1. Install TensorRT-LLM (tested on Ubuntu 24.04).

    pip3 install torch==2.7.1 torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128

sudo apt-get -y install libopenmpi-dev && pip3 install --upgrade pip setuptools && pip3 install tensorrt_llm

    If using the PyTorch NGC Container image, the prerequisite steps for installing CUDA-enabled PyTorch package and libopenmpi-dev are not required.

  2. Sanity check the installation by running the following in Python (tested on Python 3.12):

     1from tensorrt_llm import SamplingParams
 2from tensorrt_llm._tensorrt_engine import LLM
 3
 4
 5def main():
 6
 7    prompts = [
 8        "Hello, my name is",
 9        "The president of the United States is",
10        "The capital of France is",
11        "The future of AI is",
12    ]
13    sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
14
15    llm = LLM(model="TinyLlama/TinyLlama-1.1B-Chat-v1.0")
16
17    outputs = llm.generate(prompts, sampling_params)
18
19    # Print the outputs.
20    for output in outputs:
21        prompt = output.prompt
22        generated_text = output.outputs[0].text
23        print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")
24
25
26# The entry point of the program need to be protected for spawning processes.
27if __name__ == '__main__':
28    main()