Cortex_M backend: Remove int8 requirement for zero_points (#16061)

The zero point would sometimes get out of range in the hardswish
decomposition (depending on the range of the quantized output) as it
adjusts the zero point. Since the zero_point is applied in int32
accumulation there is no actual reason for the limitation, so we can
just remove it.

Additionally change the input to linear_hardswish to reliably trigger
the behaviour causing the crash before the fix was introduced to catch
future regressions.

cc @freddan80 @per @zingo @oscarandersson8218 @digantdesai

Signed-off-by: Adrian Lundell <[email protected]>

Author: AdrianLundell

backends/cortex_m/ops/cortex_m_ops_common.h

@@ -82,13 +82,6 @@ inline void validate_single_quant_params(
   int64_t mult_val = multiplier.to<int64_t>();
   int64_t shift_val = shift.to<int64_t>();
 
-  ET_CHECK_MSG(
-      zp_val >= std::numeric_limits<int8_t>::min() &&
-          zp_val <= std::numeric_limits<int8_t>::max(),
-      "%s zero point must be in int8 range [Value: %d]",
-      param_name,
-      zp_val);
-
   ET_CHECK_MSG(
       mult_val >= std::numeric_limits<int32_t>::min() &&
           mult_val <= std::numeric_limits<int32_t>::max(),

backends/cortex_m/test/ops/test_activation.py

@@ -485,7 +485,7 @@ def forward(self, x):
     ),
     "linear_hardswish": McuTestCase(
         model=CortexMLinearHardswish(in_features=12, out_features=6),
-        example_inputs=(ramp_tensor(-12, 12, (1, 12)),),
+        example_inputs=(ramp_tensor(-2, 0, (1, 12)),),
     ),
     "conv2d_hardsigmoid_inplace": McuTestCase(
         model=CortexMConv2DHardsigmoid(),