AttributeError: 'dict' object has no attribute 'llm_config'

[fengxinbo1]

[rank0]: Traceback (most recent call last): [rank0]: File "/home/wangjing/fxb/workspace/mamba/mmMamba-main/distill_mmMamba.py", line 565, in [rank0]: main() [rank0]: File "/home/wangjing/fxb/workspace/mamba/mmMamba-main/distill_mmMamba.py", line 282, in main [rank0]: trainer = OurTrainer(accelerator=accelerator, [rank0]: File "/home/wangjing/fxb/workspace/mamba/mmMamba-main/src/trainer/distill_every_layer.py", line 25, in init [rank0]: super().init(model=model, [rank0]: File "/home/wangjing/fxb/workspace/mamba/mmMamba-main/src/trainer/default_lm.py", line 118, in init [rank0]: self.loss_fn = EndtoEnd(self.model.config.llm_config, temperature=1.0, alpha=1.0) # 蒸馏 [rank0]: AttributeError: 'dict' object has no attribute 'llm_config'

为什么按照步骤来，会有提示没有找到'llm_config'参数

[bollossom]

I also found this problem. Just assign a value to vocab_size in EndtoEnd.

#  self.loss_fn = EndtoEnd(self.model.config["llm_config"], temperature=1.0, alpha=1.0)
self.loss_fn = EndtoEnd(temperature=1.0, alpha=1.0)

class EndtoEnd(nn.Module):
    def __init__(self, temperature=2.0, alpha=1.0):
        super(EndtoEnd, self).__init__()
        self.temperature = temperature
        self.alpha = alpha
        # self.config = config
        self.vocab_size = 92553

[Hongyuan-Tao]

Sorry for the late reply. I will fix this bug.