shape and size mismatch of the parameter

[chiragmathur27]

Hey,
i wanted to run directed evolution for one of my sequences, but when ran the 'run_de.sh', it gave me this error:

        size mismatch for net.esm.encoder.layer.11.output.dense.weight: copying a param with shape torch.Size([480, 1920]) from checkpoint, the shape in current model is torch.Size([1280, 5120]).
        size mismatch for net.esm.encoder.layer.11.output.dense.bias: copying a param with shape torch.Size([480]) from checkpoint, the shape in current model is torch.Size([1280]).
        size mismatch for net.esm.encoder.layer.11.LayerNorm.weight: copying a param with shape torch.Size([480]) from checkpoint, the shape in current model is torch.Size([1280]).
        size mismatch for net.esm.encoder.layer.11.LayerNorm.bias: copying a param with shape torch.Size([480]) from checkpoint, the shape in current model is torch.Size([1280]).
        size mismatch for net.esm.encoder.emb_layer_norm_after.weight: copying a param with shape torch.Size([480]) from checkpoint, the shape in current model is torch.Size([1280]).
        size mismatch for net.esm.encoder.emb_layer_norm_after.bias: copying a param with shape torch.Size([480]) from checkpoint, the shape in current model is torch.Size([1280]).
        size mismatch for net.esm.pooler.dense.weight: copying a param with shape torch.Size([480, 480]) from checkpoint, the shape in current model is torch.Size([1280, 1280]).
        size mismatch for net.esm.pooler.dense.bias: copying a param with shape torch.Size([480]) from checkpoint, the shape in current model is torch.Size([1280]).
        size mismatch for net.esm.contact_head.regression.weight: copying a param with shape torch.Size([1, 240]) from checkpoint, the shape in current model is torch.Size([1, 660]).
        size mismatch for net.decoder.dense_1.weight: copying a param with shape torch.Size([512, 480]) from checkpoint, the shape in current model is torch.Size([512, 1280]). 
  
and much more
PS: I have also force downloading the weights.
Looking for your insights and thank you in advance.

[thanhtvt]

Hi, thank you for your interest!

Did you train the model from scratch or use the previous provided checkpoints?