Handle output from TOSA reference model being int8

Move TOSA reference model. This add .npy array now is
aligned to output int8 for BI models.
Make sure the dequantization not overflows by adding a cast
to int32.

Signed-off-by: Per Åstrand <[email protected]>
Change-Id: I2274955bc02a6f3a75687bb2f731b2058834c44f

Author: per

backends/arm/test/runner_utils.py

@@ -439,6 +439,9 @@ def run_tosa_ref_model(
 
             if self.is_quantized:
                 # Need to dequant back to FP32 for comparison with torch output
+                # Convert to int32 prior to dequantize the output
+                if tosa_ref_output.dtype == np.int8:
+                    tosa_ref_output = tosa_ref_output.astype(np.int32)
                 quant_param = self.qp_output
                 assert (
                     quant_param is not None

examples/arm/setup.sh

@@ -88,7 +88,7 @@ ethos_u_base_rev="24.08"
 
 # tosa reference model
 tosa_reference_model_url="https://review.mlplatform.org/tosa/reference_model"
-tosa_reference_model_rev="444eb365d92774430006e56a8c20161be2f2674f"
+tosa_reference_model_rev="f9ea4ab7da19318fe36b1c34d68a3e40fd6e56c5"
 
 ########
 ### Mandatory user args