checkpoint error

[DDgogogo]

The checkpoints you provided mismatch the model, can you help me？

Traceback (most recent call last):
  File "DiffusionBridge/sample.py", line 206, in <module>
    main()
  File "DiffusionBridge/sample.py", line 53, in main
    model.load_state_dict(torch.load(args.model_path, map_location="cpu"))
  File "/root/anaconda3/envs/ddbm/lib/python3.12/site-packages/torch/nn/modules/module.py", line 2153, in load_state_dict
    raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(
RuntimeError: Error(s) in loading state_dict for UNetModel:
        size mismatch for input_blocks.5.1.qkv.weight: copying a param with shape torch.Size([1152, 384, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 384, 1]).
        size mismatch for input_blocks.5.1.proj_out.weight: copying a param with shape torch.Size([384, 384, 1, 1]) from checkpoint, the shape in current model is torch.Size([384, 384, 1]).
        size mismatch for input_blocks.6.1.qkv.weight: copying a param with shape torch.Size([1152, 384, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 384, 1]).
        size mismatch for input_blocks.6.1.proj_out.weight: copying a param with shape torch.Size([384, 384, 1, 1]) from checkpoint, the shape in current model is torch.Size([384, 384, 1]).
        size mismatch for input_blocks.7.1.qkv.weight: copying a param with shape torch.Size([1152, 384, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 384, 1]).
        size mismatch for input_blocks.7.1.proj_out.weight: copying a param with shape torch.Size([384, 384, 1, 1]) from checkpoint, the shape in current model is torch.Size([384, 384, 1]).
        size mismatch for input_blocks.9.1.qkv.weight: copying a param with shape torch.Size([1728, 576, 1, 1]) from checkpoint, the shape in current model is torch.Size([1728, 576, 1]).
        size mismatch for input_blocks.9.1.proj_out.weight: copying a param with shape torch.Size([576, 576, 1, 1]) from checkpoint, the shape in current model is torch.Size([576, 576, 1]).
        size mismatch for input_blocks.10.1.qkv.weight: copying a param with shape torch.Size([1728, 576, 1, 1]) from checkpoint, the shape in current model is torch.Size([1728, 576, 1]).
        size mismatch for input_blocks.10.1.proj_out.weight: copying a param with shape torch.Size([576, 576, 1, 1]) from checkpoint, the shape in current model is torch.Size([576, 576, 1]).
        size mismatch for input_blocks.11.1.qkv.weight: copying a param with shape torch.Size([1728, 576, 1, 1]) from checkpoint, the shape in current model is torch.Size([1728, 576, 1]).
        size mismatch for input_blocks.11.1.proj_out.weight: copying a param with shape torch.Size([576, 576, 1, 1]) from checkpoint, the shape in current model is torch.Size([576, 576, 1]).
        size mismatch for input_blocks.13.1.qkv.weight: copying a param with shape torch.Size([2304, 768, 1, 1]) from checkpoint, the shape in current model is torch.Size([2304, 768, 1]).
        size mismatch for input_blocks.13.1.proj_out.weight: copying a param with shape torch.Size([768, 768, 1, 1]) from checkpoint, the shape in current model is torch.Size([768, 768, 1]).
        size mismatch for input_blocks.14.1.qkv.weight: copying a param with shape torch.Size([2304, 768, 1, 1]) from checkpoint, the shape in current model is torch.Size([2304, 768, 1]).
        size mismatch for input_blocks.14.1.proj_out.weight: copying a param with shape torch.Size([768, 768, 1, 1]) from checkpoint, the shape in current model is torch.Size([768, 768, 1]).
        size mismatch for input_blocks.15.1.qkv.weight: copying a param with shape torch.Size([2304, 768, 1, 1]) from checkpoint, the shape in current model is torch.Size([2304, 768, 1]).
        size mismatch for input_blocks.15.1.proj_out.weight: copying a param with shape torch.Size([768, 768, 1, 1]) from checkpoint, the shape in current model is torch.Size([768, 768, 1]).
        size mismatch for middle_block.1.qkv.weight: copying a param with shape torch.Size([2304, 768, 1, 1]) from checkpoint, the shape in current model is torch.Size([2304, 768, 1]).
        size mismatch for middle_block.1.proj_out.weight: copying a param with shape torch.Size([768, 768, 1, 1]) from checkpoint, the shape in current model is torch.Size([768, 768, 1]).
        size mismatch for output_blocks.0.1.qkv.weight: copying a param with shape torch.Size([2304, 768, 1, 1]) from checkpoint, the shape in current model is torch.Size([2304, 768, 1]).
        size mismatch for output_blocks.0.1.proj_out.weight: copying a param with shape torch.Size([768, 768, 1, 1]) from checkpoint, the shape in current model is torch.Size([768, 768, 1]).
        size mismatch for output_blocks.1.1.qkv.weight: copying a param with shape torch.Size([2304, 768, 1, 1]) from checkpoint, the shape in current model is torch.Size([2304, 768, 1]).
        size mismatch for output_blocks.1.1.proj_out.weight: copying a param with shape torch.Size([768, 768, 1, 1]) from checkpoint, the shape in current model is torch.Size([768, 768, 1]).
        size mismatch for output_blocks.2.1.qkv.weight: copying a param with shape torch.Size([2304, 768, 1, 1]) from checkpoint, the shape in current model is torch.Size([2304, 768, 1]).
        size mismatch for output_blocks.2.1.proj_out.weight: copying a param with shape torch.Size([768, 768, 1, 1]) from checkpoint, the shape in current model is torch.Size([768, 768, 1]).
        size mismatch for output_blocks.3.1.qkv.weight: copying a param with shape torch.Size([2304, 768, 1, 1]) from checkpoint, the shape in current model is torch.Size([2304, 768, 1]).
        size mismatch for output_blocks.3.1.proj_out.weight: copying a param with shape torch.Size([768, 768, 1, 1]) from checkpoint, the shape in current model is torch.Size([768, 768, 1]).
        size mismatch for output_blocks.4.1.qkv.weight: copying a param with shape torch.Size([1728, 576, 1, 1]) from checkpoint, the shape in current model is torch.Size([1728, 576, 1]).
        size mismatch for output_blocks.4.1.proj_out.weight: copying a param with shape torch.Size([576, 576, 1, 1]) from checkpoint, the shape in current model is torch.Size([576, 576, 1]).
        size mismatch for output_blocks.5.1.qkv.weight: copying a param with shape torch.Size([1728, 576, 1, 1]) from checkpoint, the shape in current model is torch.Size([1728, 576, 1]).
        size mismatch for output_blocks.5.1.proj_out.weight: copying a param with shape torch.Size([576, 576, 1, 1]) from checkpoint, the shape in current model is torch.Size([576, 576, 1]).
        size mismatch for output_blocks.6.1.qkv.weight: copying a param with shape torch.Size([1728, 576, 1, 1]) from checkpoint, the shape in current model is torch.Size([1728, 576, 1]).
        size mismatch for output_blocks.6.1.proj_out.weight: copying a param with shape torch.Size([576, 576, 1, 1]) from checkpoint, the shape in current model is torch.Size([576, 576, 1]).
        size mismatch for output_blocks.7.1.qkv.weight: copying a param with shape torch.Size([1728, 576, 1, 1]) from checkpoint, the shape in current model is torch.Size([1728, 576, 1]).
        size mismatch for output_blocks.7.1.proj_out.weight: copying a param with shape torch.Size([576, 576, 1, 1]) from checkpoint, the shape in current model is torch.Size([576, 576, 1]).
        size mismatch for output_blocks.8.1.qkv.weight: copying a param with shape torch.Size([1152, 384, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 384, 1]).
        size mismatch for output_blocks.8.1.proj_out.weight: copying a param with shape torch.Size([384, 384, 1, 1]) from checkpoint, the shape in current model is torch.Size([384, 384, 1]).
        size mismatch for output_blocks.9.1.qkv.weight: copying a param with shape torch.Size([1152, 384, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 384, 1]).
        size mismatch for output_blocks.9.1.proj_out.weight: copying a param with shape torch.Size([384, 384, 1, 1]) from checkpoint, the shape in current model is torch.Size([384, 384, 1]).
        size mismatch for output_blocks.10.1.qkv.weight: copying a param with shape torch.Size([1152, 384, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 384, 1]).
        size mismatch for output_blocks.10.1.proj_out.weight: copying a param with shape torch.Size([384, 384, 1, 1]) from checkpoint, the shape in current model is torch.Size([384, 384, 1]).
        size mismatch for output_blocks.11.1.qkv.weight: copying a param with shape torch.Size([1152, 384, 1, 1]) from checkpoint, the shape in current model is torch.Size([1152, 384, 1]).
        size mismatch for output_blocks.11.1.proj_out.weight: copying a param with shape torch.Size([384, 384, 1, 1]) from checkpoint, the shape in current model is torch.Size([384, 384, 1]).

[zhengkw18]

What command do you use?

[DDgogogo]

What command do you use?
bash scripts/sample.sh e2h 50 dbim 0.8

[zhengkw18]

If you download the checkpoints from DDBM, please run python preprocess_ckpt.py to complete the conversion.

[DDgogogo]

If you download the checkpoints from DDBM, please run python preprocess_ckpt.py to complete the conversion.

Could you tell which paper this function is came from and why it is calculated this way?

    def get_abc(self, t):

        alpha_t, alpha_bar_t, rho_t, rho_bar_t = self.get_alpha_rho(t)
        a_t, b_t, c_t = (
            (alpha_bar_t * rho_t**2) / self.rho_T**2,
            (alpha_t * rho_bar_t**2) / self.rho_T**2,
            (alpha_t * rho_bar_t * rho_t) / self.rho_T,
        )
        return a_t, b_t, c_t

[zhengkw18]

These notations are introduced in the paper "Schrodinger Bridges Beat Diffusion Models on Text-to-Speech Synthesis". It is an equivalent way of calculating a_t, b_t, c_t in DDBM.