full image - Repost: Can Google Colab be used to fine-tune a model? (from Reddit.com, Can Google Colab be used to fine-tune a model?)
Mining:
Exchanges:
Donations:
I need to create a neural network for my degree project, I am in a company, but unfortunately the laptop that was given to me cannot in any way move an LLM locally with Ollama, Due to several issues, the company still hasn't told me anything to fix that, so today I tried to use Colab to perform the process, but I found that it runs out of RAM when trying to run any 8b model.Outside of that I don't know why it keeps running in the system RAM and doesn't take the GPU, and that my institutional account does not have any pro plan in Colab, so I only use the free plan that has 12 GB on the system, and 15 GB on the GPU (in comparison the laptop has only 8GB of RAM without a GPU of any kind) but I do have unlimited storage, so all I have to do is upload the models to drive and then import them.My idea was to use Aya 8b for the necessary retraining and then restrict it on most topics so that it only delved into common conversation topics and the medical field, which is the line of business of this company (a testing laboratory) but then I found "medicine-chat" which seems to be more focused on what I need, but none of the models run, and I'm also starting to have problems both with storage on the machine in Google colab and locally. Any advice for this?
Social Media Icons