santa hat
deerdeer nosedeer glow
Sign In

LoRas on Mac soon?

My understanding is that the reason LoRas could not be created on the Mac is because the Mac did not support tensorflow. given that it does now, does anyone know if LoRa's are coming soon to the Mac?

2 Answers

Macs don't support tensorflow yes, but tha'ts only LOCALLY, because NVIDA cards dont' exist for Macs anymore (...*sighs* stupid tech wars) - BUT, if you use Colabs, or Runpod, vast etc (i realize i'm semi literally repeating here but i'm a mac user, can confirm everything above)_ - If you need someting beefier than a COLAB instance, then use AWS, or even Rundpod, VastAI or something similar.

Yo ucan get Dedicated GPU hosts for cheaper than most of the on demand ones from what i've seen abotu 90 USD maybe onwards of about 200 USD a month? Out of most people's budgets yes, but its all depending on your needs.

I run both a macbook pro 2017 model and imac 2019 model - SD CAN run locally, but i would NEVER dare try and even TRAIN a model or lora on it locally even WITH the "LOW VRAM OPTION" - this is before and after tensorflow lol.

You run the risk of it not even recognizing your GPU, even on a desktop machine (unless you're on mac mini M1/M2) - it really depends on your hardware - if you're running m1/M2 Apple Silicon it may work? But it may not.

I'm sorry but your understanding is not entirely accurate. The reason why LoRas couldn't be created on the Mac is not just because of the lack of support for TensorFlow, but mainly because of the limited GPU resources on most Mac machines, which makes it difficult to train large deep learning models like LoRas.

While it is true that TensorFlow is now available on the Mac, the hardware limitation still exists. However, there are ways to train LoRas on the Mac using cloud-based solutions such as Google Colab or AWS EC2 instances with GPU support. Alternatively, you can also consider using pre-trained LoRa models available on Hugging Face or other model repositories and fine-tune them on your own data using transfer learning techniques.

In summary, while it is technically possible to train LoRas on the Mac with the right hardware and software setup, it is still challenging due to limited GPU resources.

Your answer