Fix divide by zero error

[apsonawane]

This pull request improves error handling for the EncodeBatch function in the tokenizer by adding input validation and corresponding unit tests. The main goal is to prevent crashes when the function is called with empty input.

Error handling improvements:

• src/models/model.cpp: Added a check in Tokenizer::EncodeBatch to throw a std::runtime_error if the input string list is empty, preventing a crash when called with no input strings.

Testing enhancements:

• test/c_api_tests.cpp: Added a new test (EncodeBatchEmptyInputThrows) to verify that calling EncodeBatch with zero input strings throws a std::runtime_error instead of causing a crash.

[Copilot]

Pull request overview

This PR hardens the tokenizer batch-encoding path used by the C/C++ API by adding input validation to prevent a divide-by-zero crash when EncodeBatch is called with an empty batch, and adds a regression test to ensure the error is surfaced as an exception.

Changes:

• Add an early strings.empty() guard in Tokenizer::EncodeBatch(std::span<const char*>) to prevent divide-by-zero when computing the output tensor shape.
• Add a C API regression test that verifies EncodeBatch(nullptr, 0) throws instead of crashing.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
src/models/model.cpp	Adds empty-input validation to prevent a divide-by-zero when shaping the encoded batch tensor.
test/c_api_tests.cpp	Adds a regression test ensuring empty-batch encode throws rather than crashing.