Fix tokenizer added_tokens format and numpy compatibility

Fix HF Space tokenizer startup crash

Remove added_tokens file before tokenizer init

Use slow Roberta tokenizer for CodeT5

Sanitize added_tokens before tokenizer load

Fix tokenizer init fallback for HF Spaces startup

Add use_fast and trust_remote_code for tokenizer compatibility

Switch to CodeT5+ (codet5p-220m) model

Switch to CodeGen-350M for better tokenizer compatibility

Switch to CodeT5+ for better transformers compatibility

Fix tokenizer compatibility issue with use_fast=False

Add CodeT5 model integration

Add application file