Rajjeshwar

icon indicating a website link https://medium.com/@gangulyraj3

icon location Master's student in CS (ML/DL) at Mila and UdeM. Author at FAUN publication.

Results 1 comments of Rajjeshwar

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

I tried this with ViTb and convnext, I setup a convnext pipeline in this notebook https://drive.google.com/file/d/1LTfJsAcTgJ3Rb8BXiAuC66OD-_9zEWSy/view?usp=drive_link This is actually using PTQ. Int8 does not seem to give any boosts in...