Repository Issue Activity (beta)

#2419 ct.models.utils.bisect_model requires pytest to be installed

#2751 Map MIL ops back to their originating nn.Module for debugging and profiling

#2749 Map MIL ops back to their originating nn.Module for debugging and profiling

#2747 TensorFlow frontend Softplus converter still emits native softplus op (fp16 overflow risk on ANE)

#2745 [PyTorch] Support divisor_override for avg_pool2d and avg_pool3d

#2570 torch.floordivide returns wrong dtype when inputs are integers

#2740 Incomplete public API docstrings, misplaced test file, and grammar inconsistency in setup.py

#2258 RoPE implemetation, Torch.polar op not available in MIL requesting support

#2733 torch.div with rounding_mode returns float for integer inputs (should be int)

#2704 fix(calibration): `_update_tensor_range` should not pick only last batch min/max

#2628 MIL pow op for 2.0 to 0.0 is 0.0

#2730 # Bug Report: EnumeratedShapes / FlexibleShapeInfo models fail at inference, no workaround found

#2204 tf.random.uniform is optimized out when it shouldn't be

#2603 ct.convert(convert_to="mlprogram", compute_precision=ct.precision.FLOAT16) have precision problem for some models

#2625 Large output errors when converting MobileNetV3 in FP16

#2263 ML Program matmul not handling 1D input correctly

#2729 logcumsumexp overflows on Apple Neural Engine in fp16 for inputs > ~11.09 (output collapses to inf then NaN)

#2687 Softplus on Apple Neural Engine has a hard fp16 discontinuity at x ≈ 10.4 (output drops to 0)

#2690 Channel-reduce logsumexp on Apple Neural Engine has a hard fp16 overflow at x ≈ 7.63 (output drops to 0)

#2728 log_softmax produces -inf on Apple Neural Engine in fp16 when one class dominates

#2359 High Numerical Errors in Mish Activation with FLOAT16 Precision on Neural Engine

#2377 `pad = None` in `ops.py` throws error for checks that require iterables

#2568 torch.randint generates out of range values in CoreML

#2716 [ANE] palettize_weights(nbits=4) stateful mlpackage built with cm 9.0 crashes after ~100-200 sequential predicts; cm 8.3.0 build of same source does not

#2456 Error Converting max_pool2d