Implementation of Positional Interpolation (PI) Feature

[andy-yang-1]

This Pull Request aims to introduce the Positional Interpolation (PI) feature to the vllm library. Positional Interpolation is a novel encoding technique that enhances the effects of Rotary Position Encoding (RoPE) by providing more precise positional information. The code now supports long-context models.

Try with:

from vllm import LLM, SamplingParams

# Sample prompts.
prompts = [
    "Hello, my name is",
    "The president of the United States is",
    "The capital of France is",
    "The future of AI is",
]
# Create a sampling params object.
sampling_params = SamplingParams(temperature=0.8, top_p=0.95)

# Create an LLM.
llm = LLM(model="lmsys/vicuna-7b-v1.5-16k")
# Generate texts from the prompts. The output is a list of RequestOutput objects
# that contain the prompt, generated text, and other information.
outputs = llm.generate(prompts, sampling_params)
# Print the outputs.
for output in outputs:
    prompt = output.prompt
    generated_text = output.outputs[0].text
    print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")

[zhuohan123]

@andy-yang-1 Does this PR have overlaps with #555? Can you take a look at that as well?

[andy-yang-1]

@zhuohan123 Yes, it overlaps with #555

[WoosukKwon]

HI @andy-yang-1, we recently merged #555. This feature is now supported. Thanks for the contribution!