Qwen/Qwen2.5-3B — a 3 billion parameter base language model released by the Qwen team in late 2024. This is the base (pre-trained) model, not the instruction-tuned variant.
How the Model Was Loaded
The model was loaded in Google Colab on a T4 GPU (free tier) using the transformers library with float16 precision:
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
