!2209 add function for quant_utils that convert float to int

Merge pull request !2209 from chenzhongming/master

mindspore/train/quant/quant_utils.py

@@ -19,6 +19,7 @@ import numpy as np
 
 def cal_quantization_params(input_min,
                             input_max,
+                            data_type,
                             num_bits=8,
                             symmetric=False,
                             narrow_range=False):
@@ -28,6 +29,7 @@ def cal_quantization_params(input_min,
     Args:
         input_min (int, list): The dimension of channel or 1.
         input_max (int, list): The dimension of channel or 1.
+        data_type (numpy type) : Can ben numpy int8, numpy uint8.
         num_bits (int): Quantization number bit, support 4 and 8bit. Default: 8.
         symmetric (bool): Quantization algorithm use symmetric or not. Default: False.
         narrow_range (bool): Quantization algorithm use narrow range or not. Default: False.
@@ -52,7 +54,7 @@ def cal_quantization_params(input_min,
         # scale = 1.0, zp = 0.0
         return np.ones(input_min.shape), np.zeros(input_min.shape)
 
-    if symmetric:
+    if data_type == np.int8:
         quant_min = 0 - 2 ** (num_bits - 1)
         quant_max = 2 ** (num_bits - 1)
     else:
@@ -84,3 +86,33 @@ def cal_quantization_params(input_min,
             zp = np.floor(zp_double + 0.5)
 
     return scale, zp
+
+
+def weight2int(data,
+               scale,
+               zero_point):
+    r"""
+    calculate int8/uint8 weight from fp32. the formula is defined as:
+
+    .. math::
+
+        int8/uint8 = round(float/scale) + offset
+
+    Args:
+        data (int, list): The dimension of channel or 1. Should be NCHW.
+        scale (int, list): The dimension of channel or 1.
+        zero_point (int, list): The dimension of channel or 1.
+
+    Outputs:
+        weight (int, list): The dimension of channel or 1.
+
+    Examples:
+        >>> weight = weight2int([1, 2, 1], 1, 0)
+    """
+    if scale.shape != zero_point.shape:
+        raise ValueError("scale and zero_point should have the same shape.")
+    if scale.shape[0] > 0:
+        scale = scale.reshape(1, -1, 1, 1)
+        zero_point = zero_point.reshape(1, -1, 1, 1)
+
+    return np.round((data/scale) + zero_point)