fix: update LAST_METRICS type annotation and remove unused demo.load call

refactor: remove redundant comment in build_ui function

feat: add smoke_test and generation metrics (latency/token counts)

Refactor build_ui function to register model load event inside context for improved clarity

Fix demo.load call in predict function to ensure proper loading of model interface

Remove unsupported 'every' parameter from demo.load in build_ui function

Refactor predict function to streamline prompt building and enhance output formatting; update UI components for better user experience and status reporting

Enhance ECHO_MODE functionality and logging for improved testing and environment configuration

Add echo/useless mode to improve testing efficiency without loading models

Update app.py

Update app.py

Add tokenizer config sanitization to ensure valid file references in model snapshots

Refactor tokenizer loading in get_generator function to prioritize fast tokenizer and improve error handling

Add huggingface_hub to requirements.txt for improved model support

Enhance tokenizer loading in get_generator function with local model snapshot support and improved error handling

Improve tokenizer loading in get_generator function with enhanced error handling and logging

Fix logging format in health server to ensure proper message output

Remove return_full_text parameter from get_generator function

Refactor app.py to enhance configuration management, add health server, and improve prompt generation

Update app.py

Refactor response handling in predict function to simplify text extraction

Refactor prompt generation to use English instructions for better clarity and context

Update app.py

Update app.py

Update app.py

Update app.py

latest updates

Refactor get_generator and predict functions to simplify device handling and improve performance

Update README and app.py for improved demo instructions and UI enhancements

Refactor Gradio demo to support zero-GPU inference with dynamic settings

Add zero-GPU Gradio demo