transformers/pr_33915/en/quantization/quanto.md

Optimum Quanto

Quanto is a PyTorch quantization backend for Optimum. It features linear quantization for weights (float8, int8, int4, int2) with accuracy very similar to full-precision models. Quanto is compatible with any model modality and device, making it simple to use regardless of hardware.

Quanto is also compatible with torch.compile for faster generation.

Install Quanto with the following command.

pip install optimum-quanto accelerate transformers

Quantize a model by creating a QuantoConfig and specifying the weights parameter to quantize to. This works for any model in any modality as long as it contains torch.nn.Linear layers.

The Transformers integration only supports weight quantization. Use the Quanto library directly if you need activation quantization, calibration, or QAT.

from transformers import AutoModelForCausalLM, AutoTokenizer, QuantoConfig

quant_config = QuantoConfig(weights="int8")
model = transformers.AutoModelForCausalLM.from_pretrained(
    "meta-llama/Llama-3.1-8B", 
    dtype="auto", 
    device_map="auto", 
    quantization_config=quant_config
)

torch.compile

Wrap a Quanto model with torch.compile for faster generation.

import torch
from transformers import AutoModelForSpeechSeq2Seq, QuantoConfig

quant_config = QuantoConfig(weights="int8")
model = AutoModelForSpeechSeq2Seq.from_pretrained(
  "openai/whisper-large-v2",
  dtype="auto",
  device_map="auto",
  quantization_config=quant_config
)

model = torch.compile(model)

Resources

Read the Quanto: a PyTorch quantization backend for Optimum blog post to learn more about the library design and benchmarks.

For more hands-on examples, take a look at the Quanto notebook.