ccsweets
ccsweets
> Now in pip install. > > We're still working on MOE support I am using Gemma3-27b model with gptq 4bit quantisation.. also for work at my company. Of course,...
> Does MoE fine-tuning work fine with the latest release? I tried but I still get: > > ``` > packages/unsloth_zoo/training_utils.py", line 170, in prepare_model_for_training > exec(f"model.{name}.to({str(dtype)})") > > File...
I also hope for gemma3 support.