lucidate

//ArunMalani//IrvineHacks25

Clean, contextual error reports.

Traceback (most recent call last):
  File "example.py", line 16, in <module>
    output_2 = model(wrong_data)
  File ".../keras/engine/base_layer.py", line 1037, in __call__
    outputs = call_fn(inputs, *args, **kwargs)
  File ".../keras/engine/functional.py", line 485, in call
    return self._run_internal_graph(inputs, training=training)
  File ".../keras/engine/functional.py", line 617, in _run_internal_graph
    outputs = node.layer(*args, **kwargs)
  File ".../keras/layers/core/dense.py", line 239, in call
    outputs = tf.linalg.matmul(inputs, self.kernel)
tensorflow.python.framework.errors_impl.InvalidArgumentError: 
Incompatible shapes: [32,128] vs. [64,128]
 [[{{node MatMul}}]] [Op:__inference_model_call_12345]

→

The model was fed data with a batch size of 32 and
then 64 in the same execution. TensorFlow cannot perform 
matrix multiplication with tensors of incompatible shapes.

Try it yourself:

> pip install lucidate
> lucidate python test.py