Summary

Add module_error_inputs_func for torch.nn.Embedding in common_modules.py to enable regression testing for error messages.

Motivation

Currently, torch.nn.Embedding does not have module_error_inputs_func defined in common_modules.py. This means error message regressions are not tested. Adding error inputs will:

1. Ensure error messages remain consistent across releases
2. Test edge cases like out-of-range indices, wrong dtype, etc.
3. Follow the pattern already established for other modules (BatchNorm, GroupNorm, RNN cells, etc.)

Error Conditions to Test

1. Out of range indices: Index exceeds num_embeddings

   • Error: IndexError: index out of range in self

2. Float indices: Float tensor passed instead of integer indices

   • Error: RuntimeError: Expected tensor for argument #1 'indices' to have one of the following scalar types: Long, Int

3. Negative num_embeddings: Constructor called with negative dimension

   • Error: RuntimeError: Trying to create tensor with negative dimension

Implementation

Add module_error_inputs_torch_nn_Embedding function and wire it to the ModuleInfo entry.

cc @albanD @mruberry @jbschlosser @walterddr @mikaylagawarecki