MoMo

Keras argmin function returns incorrect index when handling subnormal float values.

3

When using the Keras backend `argmin` function on an input array containing subnormal float values, Keras consistently returns the index of `0.0` as the minimum value, even though a smaller...

Description: When using keras.backend.argmax with an input array containing -0.0, the result is incorrect. Specifically, the function returns 1 (the index of -0.0) as the position of the maximum value,...

Arcsinh Error Report

5

### Description Issue: JAX's arcsinh results are inconsistent with those of PyTorch, TensorFlow, and Keras. Description: When using the same input data, the results returned by JAX's arcsinh function exhibit...

argsort incorrectly handles very small floating-point numbers and -0.0 compared to PyTorch

9

### Description Description: When using JAX's argsort function on an array containing small floating-point numbers, as well as 0.0 and -0.0, the sorting order is incorrect compared to other libraries...