fine-tuning for warehouse management

[MohamedKHALILRouissi]

hello , first thanks you for making this availble and nice work
i have 3 question if you could help on this wether they are possible or not

is it possible to fine tune tigerbot for warehouse management ( context , i have multiple warehouse , with multiple worker , storage are and porduct and i have a dataset of 5 year , worker attendence , woker sales , shifts , days of , sales in total , income , outcome , filling request , transfer ... ) , what i want to do is to make tigerbot specific to my day to day work for warehouse management , like:

• provide me insight
• provide information at the end of the day based on RAG docs
• provide me tips or enhancement
• provide me prediction ( if i could do this , please explain how can i make something like )

[Vivicai1005]

Hi MohamedKHALILRouissi,

In your application, you can try to fine llm with data examples might look like:

Input: A prompt describing the specific information or query, such as
"Provide a summary of today's worker information as following
worker attendence: xx
woker sales: xx,
sales in total: xx,
other info..."
Output: The model's ideal response, such as a detailed summary based on your historical data. (The output should be generated from the real world scenario analysis)

Input:
" Suggest improvements for warehouse operations Based on recent trends as following:
date: 12/01
worker attendence: xx,
woker sales: xx,
sales in total: xx,
(other info...)

date: 12/02
worker attendence: xx,
woker sales: xx,
sales in total: xx,
(other info...)

date: 12/03,
worker attendence: xx,
woker sales: xx,
sales in total: xx,"
(other info...)
Output: Actionable suggestions focusing on improving efficiency, reducing costs, or enhancing worker productivity.

[MohamedKHALILRouissi]

this is called instruction based tune ?

[Vivicai1005]

it's called Supervised fine-tuning (SFT) on instruction data

[MohamedKHALILRouissi]

can you give any example or documents to follow please