quantize-rs 0.7.0

Neural network quantization toolkit for ONNX models
Documentation
1
2
3
4
5
6
7
8
9
10
11
12
# Core — always required
onnxruntime>=1.17.0
transformers>=4.35.0
numpy>=1.24.0

# Required for --export only
# Any torch 2.x works; the export uses torch.jit.trace + legacy ONNX exporter
# (opset 14) so onnxscript is NOT required.
torch>=2.0.0

# Optional — enables real WikiText-2 perplexity instead of built-in texts
# datasets>=2.14.0