mlabonne/dummy-llama-2

dummy-llama-2

This is a dummy version of the model based on `meta-llama/Llama-2-7b-hf`.

🧩 Dummy

dummy-llama-2 has a size of 929.07 MB instead of the original 13476.98 MB (compression factor of 14.51) but keeps the base model's functionality.

The purpose of this dummy version is to be used for debugging, so you don't have to download the entire original model. Do not use it for inference.

💻 Usage

# pip install transformers accelerate

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

model = "dummy-llama-2"

tokenizer = AutoTokenizer.from_pretrained(model)
model = AutoModelForCausalLM.from_pretrained(
    model,
    low_cpu_mem_usage=True,
    return_dict=True,
    torch_dtype=torch.float16,
    device_map={"": 0},
)