tensor creation error in `math_interface.py`

[suryabulusu]

Line 45 in math_interface.py creates list of tensors as follows:

 x_var = [minitorch.tensor(x, requires_grad=True) for x in xs]

This leads to an error when x is not a list or tuple.

This error can be corrected in tensor_functions.py by adding the following code in def tensor(): before processing ls

if not isinstance(ls, (list, tuple)):
    ls = [ls]

[srush]

Thanks! I will fix in the master repo.

[jchwenger]

Hi @srush, I'm not sure if that's something coming from an error in my implementation, but I still get an issue for scalar tensors, this time because of a mismatch when checking the size attribute of TensorData and the length of the underlying array.
So, starting from the same Line 45:

x_var = [minitorch.tensor(x, requires_grad=True) for x in xs]

We end up in TensorData, where shape is []:

self.size = int(prod(shape))

In my implementation, prod returns 0.0 if len(a) == 0, but of course we do have an underlying array of length 1 (one scalar), which makes the following assert fail (I added a message):

assert len(self._storage) == self.size, f"{self._storage=}, {self.size=} {shape=}"

It seems reasonable to follow the standard to have a shape of () for scalars, which is the case in PyTorch, but then I'm not sure what would be the elegant way to handle those (and not bug-inducing in other parts of the code base)? Simply avoid scalars and do x_var = [minitorch.tensor([x], requires_grad=True) for x in xs] in math_interface.py? Return 1.0 instead in prod for length-zero Iterables?

Apologies if I overlooked something really simple!

Full trace:

AssertionError: self._storage=array([-48.99999]), self.size=0 shape=()
---
File "/Users/jcw/Documents/dl/learning/srush/minitorch/.venv/lib/python3.12/site-packages/streamlit/runtime/scriptrunner/script_runner.py", line 556, in _run_script
    exec(code, module.__dict__)
File "/Users/jcw/Documents/dl/learning/srush/minitorch/minitorch-module-2-jchwenger/project/app.py", line 118, in <module>
    page()
File "/Users/jcw/Documents/dl/learning/srush/minitorch/minitorch-module-2-jchwenger/project/app.py", line 89, in render_m2_sandbox
    return render_math_sandbox(True, True)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/jcw/Documents/dl/learning/srush/minitorch/minitorch-module-2-jchwenger/project/math_interface.py", line 47, in render_math_sandbox
    x_var = [minitorch.tensor(x, requires_grad=True) for x in xs]
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/jcw/Documents/dl/learning/srush/minitorch/minitorch-module-2-jchwenger/minitorch/tensor_functions.py", line 434, in tensor
    return _tensor(cur, tuple(shape2), backend=backend, requires_grad=requires_grad)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/jcw/Documents/dl/learning/srush/minitorch/minitorch-module-2-jchwenger/minitorch/tensor_functions.py", line 399, in _tensor
    tensor = minitorch.Tensor.make(ls, shape, backend=backend)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/jcw/Documents/dl/learning/srush/minitorch/minitorch-module-2-jchwenger/minitorch/tensor.py", line 263, in make
    return Tensor(TensorData(storage, shape, strides), backend=backend)
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/jcw/Documents/dl/learning/srush/minitorch/minitorch-module-2-jchwenger/minitorch/tensor_data.py", line 192, in __init__
    assert len(self._storage) == self.size, f"{self._storage=}, {self.size=} {shape=}"
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^