kosmos / quantize.py
strelizi's picture
feat: fresh deployment without LFS artifacts
66ec738
Raw History Blame Contribute Delete
355 Bytes
# quantize.py
from onnxruntime.quantization import quantize_dynamic, QuantType
print("Quantizing bge_m3_onnx/model.onnx to INT8...")
quantize_dynamic(
model_input="bge_m3_onnx/model.onnx",
model_output="bge_m3_onnx/model_quantized.onnx",
weight_type=QuantType.QUInt8
)
print("Quantization complete: bge_m3_onnx/model_quantized.onnx created!")