shuvo881/Astra-LLM

Check Python version

N.B: python version must be required up to 3.10.0

!python --version

Install dependencies of Astra LLM

pip install astra-llm==0.1.19

Load Model

Selection of device for load model

import torch

# Set device to GPU if available
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

Load model

from astra.model import Model

model = Model('/kaggle/input/astra/pytorch/1.4b/2', map_location=device)

Generate text

response = model.generate(
    messages="hello",
    max_new_tokens=50,
    top_k=25,
    temperature=1.4
)