mGPT-quantized

mGPT-quantized

The concept: 8-bit quantized version of mGPT, a 1.3B param model released by AI-Forever / Sberbank AI in April 2022.

On the GPT scale, it is a similar # of parameters to GPT2-XL, but on 60+ languages.

AI-Forever also released a 13B-parameter model. I made an 8-bit quantized version with weights available here: https://huggingface.co/monsoon-nlp/mGPT-13B-quantized

My goal is to evaluate this on Arabic, Hindi, and Indonesian tasks, where there are fewer autoregressive language models in this size range.

For English: use a GPT model or LLaMa2-7B

In August 2023 AI-Forever added 1.3B-param models for about 1/3 of the model's languages. If your language is Mongolian, for example, use mGPT-1.3B-mongol and not this one.

How was the model created?

Quantization of mGPT 1.3B was done using bitsandbytes library:

from transformers import BitsAndBytesConfig, GPT2LMHeadModel

quantization_config = BitsAndBytesConfig(
    load_in_8bit=True,
    bnb_8bit_compute_dtype=torch.bfloat16,
    bnb_8bit_use_double_quant=True,
    bnb_8bit_quant_type="nf4",
)

qmodel = GPT2LMHeadModel.from_pretrained(
    "ai-forever/mGPT",
    load_in_8bit=True,
    torch_dtype=torch.bfloat16,
    quantization_config=quantization_config,
    device_map="auto"
)

qmodel.save_pretrained("model_name")

Future steps

• mGPT could be further quantized (4-bit), but model.save_pretrained() currently throws a NotImplementedError error.