Dataset Card: Blind Spots of Qwen2.5-1.5B (Base)
Model Tested
Model: Qwen/Qwen2.5-1.5B
Type: Base Model (Not instruction-tuned)
Parameters: 1.5 Billion
How the Model was Loaded
The model was loaded using a Google Colab T4 GPU instance via the transformers library, utilizing torch.float16 precision and device_map="auto" to fit comfortably within the 16GB VRAM constraint.
Python
from transformers import AutoModelForCausalLM, AutoTokenizerimport torch
model_id = "Qwen/Qwen2.5-1.5B"
tokenizer =… See the full description on the dataset page:
https://huggingface.co/datasets/SA-R/Qwen-1.5B-Base-Blindspots.
