296 questions
I am running Phi 1-5 and I'm receiving the same error. I fixed this by changing the adapters_name variable in the console_chat.py to reflect the checkpoint directory during my training run for finetuning. I my case, I changed as follows:
# adapters_name = "../models/qlora/qlora/gpu-cpu_model/adapter" # Ensure this path is correctly set before running
adapters_name = "../models/checkpoints/checkpoint-1875"