To mitigate potential security risks and ensure the integrity of the UP-PARAM.BIN file, it is essential to follow best practices, such as:
In Hugging Face PEFT libraries, when you save a LoRA adapter (using peft_model.save_pretrained() ), the output directory might contain adapter_model.bin . However, some legacy or custom training scripts (or merged LoRA extraction tools) split the adapter into lora_A.weight and lora_B.weight . When saved manually via torch.save() , lora_B.weight is often stored as up-param.bin . up-param.bin
Are you experiencing a specific in Odin, or are you looking to customize your boot logo using this file? To mitigate potential security risks and ensure the
| Field | Possible Type | Notes | |-------|--------------|-------| | Header | uint32 magic | e.g., 0x5550424D ("UPM") | | Version | uint32 | 1, 2, etc. | | Data size | uint32 | Bytes of parameter data | | Checksum | uint32 | CRC32 or custom | | Payload | Raw floats/ints | Could be half-precision (FP16) or single-precision (FP32) | Are you experiencing a specific in Odin, or
Then, silence.