Fix incorrect dtype default in aten_full causing ONNX shape-inference errors

[Copilot]

aten_full defaulted dtype to FLOAT.dtype. When fill_value is integral and no dtype is supplied, the value was cast to float, yielding type mismatches downstream (e.g. Expand: Inferred elem type differs from existing elem type: (1) vs (7)) and failing strict shape inference.

Change

• onnxscript/function_libs/torch_lib/ops/core.py: default dtype changed from FLOAT.dtype to -1, matching aten_full_like. The existing if dtype != -1: guard now skips the cast when dtype is unspecified, preserving fill_value's original type.

def aten_full(
    size: Union[INT64, INT32],
    fill_value: TensorType,
    dtype: int = -1,  # was FLOAT.dtype
    ...
):
    if dtype != -1:
        fill_value = op.Cast(fill_value, to=dtype)
    size = op.Cast(size, to=INT64.dtype)
    return op.Expand(fill_value, size)

With the prior default, an integral fill_value with no dtype reproduced the Expand TypeInferenceError; the change resolves it while leaving the explicit-dtype path unchanged.

[Copilot]

Pull request overview

This PR fixes aten_full’s default dtype handling in the Torch lib so that, when dtype is not specified, the fill_value’s original element type is preserved (avoiding downstream ONNX type/shape-inference mismatches such as Expand type inference errors).

Changes:

• Update aten_full’s dtype default from FLOAT.dtype to -1 (unspecified), aligning it with aten_full_like.
• Rely on the existing if dtype != -1: guard to skip Cast when dtype is unspecified.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 72.65%. Comparing base (b16ef16) to head (f85e90b).
⚠️ Report is 2 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2926   +/-   ##
=======================================
  Coverage   72.64%   72.65%           
=======================================
  Files         259      259           
  Lines       31655    31655           
  Branches     2981     2981           
=======================================
+ Hits        22997    22998    +1     
  Misses       7649     7649           
+ Partials     1009     1008    -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.