[BUG] Shouldnt the LoRA be the same float as the base model?

[jackswl]

Prerequisites

• I have read the documentation.
• I have checked other issues for similar problems.

Backend

Local

Interface Used

CLI

CLI Command

No response

UI Screenshots & Parameters

No response

Error Logs

    logger.info("Loading adapter...")
    base_model = AutoModelForCausalLM.from_pretrained(
        base_model_path,
        torch_dtype=torch.float16,
        low_cpu_mem_usage=True,
        trust_remote_code=ALLOW_REMOTE_CODE,
        token=token,
    )

This is the code for tools for merge_adapters

if my base model is in bfloat16, the merging should be bfloat16 too. but here, i cannot specify the base model torch_dtype. Wouldnt this be an issue? @abhishekkrthakur

Additional Information

No response

[github-actions]

This issue is stale because it has been open for 30 days with no activity.