jinhuang415 commented on a change in pull request #10433: [MXNET-290] MKLDNN 
support for model quantization
URL: https://github.com/apache/incubator-mxnet/pull/10433#discussion_r193411184
 
 

 ##########
 File path: tests/python/quantization/test_quantization.py
 ##########
 @@ -409,11 +441,19 @@ def test_quantize_sym_with_calib():
     sym = get_fp32_sym()
     offline_params = [name for name in sym.list_arguments()
                       if not name.startswith('data') and not 
name.endswith('label')]
-    qsym = mx.contrib.quant._quantize_symbol(sym, 
offline_params=offline_params)
-    requantize_op_names = ['requantize_conv', 'requantize_fc']
-    th_dict = {'conv_output': (np.random.uniform(low=100.0, high=200.0), 
np.random.uniform(low=100.0, high=200.0)),
-               'fc_output': (np.random.uniform(low=100.0, high=200.0), 
np.random.uniform(low=100.0, high=200.0))}
-    op_name_to_th_name = {'requantize_conv': 'conv_output', 'requantize_fc': 
'fc_output'}
+    if is_test_for_mkldnn():
+        dtype = 'uint8'
+        requantize_op_names = ['requantize_conv']
+        th_dict = {'conv_output': (np.random.uniform(low=100.0, high=200.0), 
np.random.uniform(low=100.0, high=200.0))}
+        op_name_to_th_name = {'requantize_conv': 'conv_output'}
+    else:
+        dtype = 'int8'
+        requantize_op_names = ['requantize_conv', 'requantize_fc']
+        th_dict = {'conv_output': (np.random.uniform(low=100.0, high=200.0), 
np.random.uniform(low=100.0, high=200.0)),
+                   'fc_output': (np.random.uniform(low=100.0, high=200.0), 
np.random.uniform(low=100.0, high=200.0))}
+        op_name_to_th_name = {'requantize_conv': 'conv_output', 
'requantize_fc': 'fc_output'}
 
 Review comment:
   Currently mkldnn doesn't support int8 for fc layer so we made separate logic 
for mkldnn to reflect real MKLDNN use case, once it is supported we can use the 
same logic.

----------------------------------------------------------------
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

Reply via email to