About Seaco-paraformer bias_embed weights

[locasxe]

❓ Questions and Help

I found that: in function load_pretrained_model: logging.debug("Loaded src_state keys: {}".format(src_state.keys())) logging.debug("Loaded dst_state keys: {}".format(dst_state.keys())) dst_state.update(src_state) obj.load_state_dict(dst_state, strict=True) if dst_state.update(src_state) is removed: obj.load_state_dict(src_state, strict=True) ERROR====》Missing key(s) in state_dict: “bias_embed.weight”

What is your question?

So, indeed, bias_embed.weight is missing in src_state, is this normal? And why?

What have you tried?

What's your environment?

• OS (e.g., Linux):
• FunASR Version (e.g., 1.0.0):
• ModelScope Version (e.g., 1.11.0):
• PyTorch Version (e.g., 2.0.0):
• How you installed funasr (pip, source):
• Python version:
• GPU (e.g., V100M32)
• CUDA/cuDNN version (e.g., cuda11.7):
• Docker version (e.g., funasr-runtime-sdk-cpu-0.4.1)
• Any other relevant information:

[R1ckShi]

bias_embed is initialized but not used, hotwords are embedded with the asr decoder embedding