Raj-vivid comments

Repositories
Issues
Comments

Results 2 comments of


                                            Raj-vivid

Int8 quantized model performs worse or similar to non quantized fp32 or fp16 model.

I tried a few more things, using onnx opset 17 to convert to onnx allowed for using a fused node for Layernorm but even with qdq before conv operation, before...

Int8 quantized model performs worse or similar to non quantized fp32 or fp16 model.

Hello, I appreciate the help you do for the community to answer all the issues on the thread, could you please have a look at this ? ``` No valid...