MLBEDSW-7442: Removed ofm quantization for ArgMax

- Quantization for the OFM was added for the ArgMax operator
as a workaround in order to avoid a crash in the weight compressor.
This quantization is now removed.

- The weight compressor expects that all tensors have a quantization.
Updated code to use scale = 1.0 and zero point = 0 for tensor without
quantization.

Change-Id: I6816dce2db55f7d795d19f88d7fbe7ee419347fc
Signed-off-by: Johan Alfven <johan.alfven@arm.com>

1 files changed, 4 insertions, 2 deletions

diff --git a/ethosu/vela/tensor.py b/ethosu/vela/tensor.py
index 51c7592e..9ba6ab77 100644
--- a/ethosu/vela/tensor.py
+++ b/ethosu/vela/tensor.py
@@ -229,6 +229,8 @@ class QuantizationParameters:
         max: Union[float, np.ndarray, None] = None,
         num_bits=None,
         narrow_range=None,
+        scale_f32: Union[float, np.ndarray, None] = None,
+        zero_point: Union[int, np.ndarray, None] = None,
     ):
         self.min = min
         self.max = max
@@ -240,8 +242,8 @@ class QuantizationParameters:
         # natural rounding to perform rounding away from zero. This only affects the ofm scale and bias tensor, it has
         # no affect on global scaling i.e. the ofm_scale register
         self.next_after = False
-        self.scale_f32: Union[float, np.ndarray, None] = None
-        self.zero_point: Union[int, np.ndarray, None] = None
+        self.scale_f32: Union[float, np.ndarray, None] = scale_f32
+        self.zero_point: Union[int, np.ndarray, None] = zero_point
         self.quant_min: Optional[float] = None
         self.quant_max: Optional[float] = None
         self.quant_dim: Optional[int] = None