When trying to merge LoRA weights and save the model in Hugging Face format using
merge_lora_weights_and_save_hf_model.py, the script fails with an IsADirectoryError because the model checkpoint is sharded across multiple .bin files instead of being a single file.
When trying to merge LoRA weights and save the model in Hugging Face format using
merge_lora_weights_and_save_hf_model.py, the script fails with an IsADirectoryError because the model checkpoint is sharded across multiple .bin files instead of being a single file.