Having a model locally on my pc + fine tuning it + using forked bolt to make it better

bolt.diy
1

**Hi everyone I hope you read my question **
Hey , i am trying to use this free unlimited usage power but with a mql5 language model , it seems that qwen isnt the best with it as this language is not for common software uses. I am facing the problem that qwen on the forked bolt is making mistakes due to lack of some basic knowledge of constants and functions signatures . The biggest problem is that im not a programmer at all. Although after long debates and searching for errors fixes , i had a final compilation with no errors . But as expected the strategy doest work . It seems that i fixed problems with (logical problems) that dont cause errors but rather make the strategy criteria wrong .

I am definitely new to this AI stuff,very limited on knowledge and my sole goal is to code mql5 without the need to learn it ( needs at least 6 months to do so )

So been thinking if you can help me , and if this goal is actually achievable and what kind of effort i need to put it , for someone without knowledge , how difficult will it be for me to do it ?

Can i download anAI model (specially qwen coder 2.5) then fine tune it for mql5 then incorporate it with forked bolt (or maybe openhands also?)

There is paid sites thats dedicated to create a neat mql5 code as they keep these AIs up to date with the syntax of the language .

But still i want the free use , so i saw an opportunity with your work .

The Question (finally) : how can i fine tune your a qwen coder 2.5 coder model to be dedicated for mql5 language , as giving it the right text book information to process . Train it to create right strategies and have a good logical and mathematical understanding to create robots , i am asking is that even possible to fine tune a local qwen coder model then enhance it with ottodev and have the perfect almost coding model for mql5 ?

I m wondering in the web but seeing no clues if this is possible

Alright thank you

2

This certainly is possible, but you would likely have to have some coding knowledge to fine tune! But any local LLM, you are able to download it, put together a dataset (based on mql5 textbooks in your case, for example), and then run it within Ollama just like any other model.

I would do some research on fine tuning local LLMs for Ollama, then you can take that and use it for Qwen 2.5 or any other local LLM!

1 Like
3

Thanks for clarification

1 Like
4

But putting a dataset and letting model check it first isnt applicable , right?
As i need to fine tune the model first with the textbook
So (on pc only fine tuned model ) can be linked to fork ? Very nice

1 Like
5

Yep a fine tuned model put into Ollama can be used in oTToDev!

1 Like