Address Raghu's static quantization comments

Seth Weidman · Seth Weidman · commit ba0a7a1f1160 · 2019-10-10T11:43:13.000-07:00
diff --git a/advanced_source/static_quantization_tutorial.py b/advanced_source/static_quantization_tutorial.py
@@ -371,7 +371,7 @@ def prepare_data_loaders(data_path):
 print('\n Inverted Residual Block: Before fusion \n\n', float_model.features[1].conv)
 float_model.eval()
 
-# Fusion is optional
+# Fuses modules
 float_model.fuse_model()
 
 # Note fusion of Conv+BN+Relu and Conv+Relu
@@ -567,6 +567,10 @@ def train_one_epoch(model, criterion, optimizer, data_loader, device, ntrain_bat
     print('Epoch %d :Evaluation accuracy on %d images, %2.2f'%(nepoch, num_eval_batches * eval_batch_size, top1.avg))
 
 #####################################################################
+# Here, we just perform quantization-aware training for a small number of epochs. Nevertheless,
+# quantization-aware training yields an accuracy of over 71% on the entire imagenet dataset,
+# which is close to the floating point accuracy of 71.9%.
+#
 # More on quantization-aware training:
 #
 # - QAT is a super-set of post training quant techniques that allows for more debugging.
