wenyang

I found the problem is that I created the experiment in the mlflow ui after running `mlflow ui`, it's solved by creating experiment in the program with this code `mlflow.create_experiment`

Have you solved your problem? I got the same issue recently.

Hi, I trained qwen2-1.5b model (with `bfloat16` dtype) via Lora, I found the dtype of trained model (merged) becomes `float32`. Is this normal? How do I cast it back to...

Thank you for helping. I use the code base from [open-instruct](https://github.com/allenai/open-instruct/tree/main), the part of training codes are: ```python # construct the peft model model = AutoModelForCausalLM.from_pretrained( args.model_name_or_path, from_tf=bool(".ckpt" in args.model_name_or_path),...

Thanks for helping. It gets the right dtype now. I made a stupid mistake :sweat_smile: .