Week 1 update in the coding period for GSoC with Sugar Labs.
Hello everyone, before reading this blog you can read my previous blog for a better understanding of my project — https://medium.com/@khadarvsk/gsoc-with-sugar-labs-e616f2a5a330
project — “Make your own lesson plans”
This week I have worked on creating a session chatbot to generate lesson plans.
How session chatbot is different from a conversational chatbot?
The session chatbot remembers only that session data but the conversational chatbot can remember the information that you shared/communicated before 2 days also.
Why a conversational chatbot?
By default the chatbots that we create using LLMs are not conversational they are not capable of remembering the previous conversation.
But, in our case, we need a dialog conversation between the Bot and the user because the draft generated by the Bot may require some changes, we can give another prompt to the Bot to make these changes, and then the model will react correspondingly.
Models that we are using for generating the output:
Mistral — https://huggingface.co/TheBloke/zephyr-7B-alpha-GGUF/blob/main/zephyr-7b-alpha.Q5_K_S.gguf
llama3 — https://huggingface.co/meta-llama/Meta-Llama-3-8B
screenshots of my work —
As you can see in the screenshots the model can remember the previous conversation and react based on that.
Next week's work —
- Train the model with lesson plans.
- By using a session chatbot build a conversational chatbot.
- If time permits work on JSON data of the project.
Thank you