bert rte glue torchdistill nlp int8 neural-compressor Intel® Neural Compressor text-classfication PostTrainingDynamic

INT8 bert-large-uncased-rte-int8-dynamic

Post-training dynamic quantization

PyTorch

This is an INT8 PyTorch model quantized with Intel® Neural Compressor.

The original fp32 model comes from the fine-tuned model yoshitomo-matsubara/bert-large-uncased-rte.

Test result

INT8 FP32
Accuracy (eval-f1) 0.7076 0.7401
Model size (MB) 766 1349

Load with Intel® Neural Compressor:

from optimum.intel.neural_compressor.quantization import IncQuantizedModelForSequenceClassification

int8_model = IncQuantizedModelForSequenceClassification.from_pretrained(
    "Intel/bert-large-uncased-rte-int8-dynamic",
)