Konstantin Gulin

icon company Neural Magic icon location Los Angeles icon location Mostly ML and Estimation Theory

Results 2 issues of Konstantin Gulin

Add variable bit width support to ONNXToDeepsparse

1 comment

This PR adds support for variable-bit weight quantization in the ONNXToDeepsparse exporter. This affects two steps: - Conversion of intiailziers to unit8 - Clipping in quantization of weight arrays **Test...

mle-team

Add automatic PR integration tests

This PR adds a workflow which triggers the base integration tests (3 out of the 10 total case) when a PR is opened against `sparsify.alpha` (to be changed to `main`...

mle-team