change default quantization calibration iteration to larger value

Summary:
Pull Request resolved: https://github.com/facebookresearch/d2go/pull/366

Users often time just use the default value without knowing they need to change  this config, and getting low accuracy after PTQ. Therefore increase the default value.

Reviewed By: miqueljubert

Differential Revision: D39331680

fbshipit-source-id: 6b05773c3c4cd48e6298d341d77a0e373d5439f6

d2go/quantization/modeling.py

@@ -145,7 +145,7 @@ def add_quantization_default_configs(_C):
 
     # post-training quantization
     _C.QUANTIZATION.PTQ = CfgNode()
-    _C.QUANTIZATION.PTQ.CALIBRATION_NUM_IMAGES = 1
+    _C.QUANTIZATION.PTQ.CALIBRATION_NUM_IMAGES = 16  # NOTE: this is actually iterations
     _C.QUANTIZATION.PTQ.CALIBRATION_FORCE_ON_GPU = False
 
     # register deprecated and renamed keys