Open kmn1024 opened 6 months ago
Hi @kmn1024 , sorry for my late response. For the bug, there is something wrong during getting specific tensor. it would be helpful to root cause this cause if you can provide your model. For questions 1, https://github.com/intel/neural-compressor/blob/8e833bdca9085d67c47a86d73917a9975e4f8fc9/neural_compressor/config.py#L202 please follow this sample code to resume. For question 2, you can try to exclude last or first linear from quantization.
When I use "basic" strategy tuning to quantize my model, I ran into this issue during one of the phases:
The config I use looks like this:
Aside from the bug, I have some related questions: