| from transformers import AutoModelForCausalLM | |
| import torch | |
| import torch.utils.dlpack | |
| # Load the original model | |
| model_name = "./mixed_llm" | |
| model = AutoModelForCausalLM.from_pretrained(model_name) | |
| # Convert the model to a different precision | |
| model = model.half() | |
| # Save the model as a safetensor | |
| model.save_pretrained(f"./mixed_llm_half", safetensors=True) |