Manu Mathew

Applying quantization and sparsity together (or applying quantization on sparsified net) should work - but I did face some issues recently and have not looked carefully into that. In test,...

Keeping this issue open for some more time, as this is an interesting conversation and provides help to someone who wants to try the same.

No problem. Documentation is missing for some of these aspects - so these conversations help. Feel free to ask more questions if you have.

If you see that sparsity is vanishing when you apply quantization, try setting these two flags to false and let me know: apply_offset_activations: false apply_offset_weights: false

Is this fixed in NVIDIA/caffe?

A simple merge/rebase failed. We will have to manually look at the conflicts and resolve them. If you have an urgent need, I can try to do this by next...

Hi, NVIDIA/caffe 0.17 is in the making with support for SSD object detection. You can follow the development and the PR that I submitted, here: https://github.com/NVIDIA/caffe/pull/446 I would like to...

We have upgraded to caffe-0.17 branch now. Let us know if you find it useful.

Hi, it's similar, but not exactly same. Caffe-jacito does not do quantitied inference, it only simulates quantization with floating point operations - and the simulation is not perfect. However, I...

CC: @kumardesappan 8 bit quantized values are supposed to be in the range 0-255 or -128-127 depending on whether they are unsigned or signed respectively. This is a quantized representation...