KellenSunderland commented on a change in pull request #13697: [MKLDNN] Enable
signed int8 support for convolution.
URL: https://github.com/apache/incubator-mxnet/pull/13697#discussion_r244627698
##########
File path: example/quantization/imagenet_gen_qsym_mkldnn.py
##########
@@ -140,8 +140,8 @@ def save_params(fname, arg_params, aux_params,
logger=None):
' thresholds. This mode is expected to produce
the best inference accuracy of all three'
' kinds of quantized models if the calibration
dataset is representative enough of the'
' inference dataset.')
- parser.add_argument('--quantized-dtype', type=str, default='uint8',
- choices=['int8', 'uint8'],
+ parser.add_argument('--quantized-dtype', type=str, default='auto',
Review comment:
Thanks for the explanation. I'm still curious about why uint8 actually
requires positive inputs. Wouldn't an input array get the same affine
transformation (scale and zero shift) applied to it just as the weights have
before a fused kernel call? I'll take a look at the FC gemm you're calling and
see if I can better understand why we need int8 for negative inputs. In any
case not blocking for this PR.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
