Increase to 600 tokens and add truncation note

Increase max_new_tokens to 400 for complete function generation

Add ignore_mismatched_sizes=True to handle config mismatch

Fix max_new_tokens=0 bug, simplify interface

Reduce max tokens for CPU timeout fix

Add accelerate, remove device_map for CPU inference

Use Textbox for gradio 5 compatibility

Use transformers AutoModelForCausalLM — no OLMo-Core dependency

Load model from HuggingFace hub