Remove eos_token_id and pad_token_id from text generation parameters in chatbot application to streamline configuration and maintain focus on essential settings.
Remove repetition penalty and no_repeat_ngram_size parameters from text generation settings, and simplify user message handling in respond function to enhance clarity and focus in chatbot interactions.
Enhance text generation parameters by adding eos_token_id, pad_token_id, repetition_penalty, and no_repeat_ngram_size to improve output quality and prevent repetition in chatbot application.
Refactor message handling in response function to support both string and dictionary inputs. Added validation for empty messages to improve user experience by providing a clear error message when no input is received.
Enhance text generation parameters in chatbot application by setting max_length to 100 and temperature to 0.7. Simplify response function by directly passing the message to the text generator, improving clarity and efficiency in message processing.
Comment out system message handling in response function to simplify message processing and maintain focus on history management in chatbot application.
Update system message in Vietnamese to emphasize clarity in code solutions while maintaining language preference for responses. This change enhances the instruction for the chatbot's output format.
Clarify instruction for model responses to ensure output is consistently in Vietnamese. This update reinforces the language preference in the chatbot's response generation process.
Add default system message in Vietnamese for chatbot responses and clarify instruction for model output language. This ensures consistent responses in the desired language when no system message is provided.
Refactor text generation in chatbot application to utilize max_new_tokens for improved clarity and added truncation support. Removed unnecessary print statement for generated answer length.
Refactor response handling in chatbot application to support optional system messages and improve message processing. Added checks for empty responses to enhance user experience.
Update response generation in chatbot application to set default values for max_tokens, temperature, and top_p parameters. This enhancement ensures smoother operation when these parameters are not explicitly provided.
Streamline model loading and response generation in chatbot application by utilizing a text generation pipeline. Removed legacy loading methods and improved response handling for enhanced performance and clarity.
Refactor model loading process in chatbot application to prioritize local path loading, with enhanced error handling and fallback mechanisms for HuggingFace models and PEFT adapters.
Enhance model loading logic in chatbot application to support direct loading and PEFT adapter fallback. Updated model and tokenizer initialization for improved error handling and device management.
Refactor model loading and input handling in chatbot application. Updated model and tokenizer initialization, improved device management for inputs, and removed unused sliders from the Gradio interface.
