Some weights of AdaptCLIPVisionModel were not initialized from the model checkpoint at openai/clip-vit-large-patch14-336 and are newly initialized

[mzamini92]

Hi I am gettinf this warning. is that ok?

Some weights of AdaptCLIPVisionModel were not initialized from the model checkpoint at openai/clip-vit-large-patch14-336 and are newly initialized: ['image_model.embeddings.class_embedding', 'image_model.embeddings.patch_embedding.weight', 'image_model.embeddings.position_embedding.weight', 'image_model.encoder.layers.0.layer_norm1.bias', 'image_model.encoder.layers.0.layer_norm1.weight', 'image_model.encoder.layers.0.layer_norm2.bias', 'image_model.encoder.layers.0.layer_norm2.weight', 'image_model.encoder.layers.0.mlp.fc1.bias', 'image_model.encoder.layers.0.mlp.fc1.weight', 'image_model.encoder.layers.0.mlp.fc2.bias', 'image_model.encoder.layers.0.mlp.fc2.weight', 'image_model.encoder.layers.0.self_attn.k_proj.bias', 'image_model.encoder.layers.0.self_attn.k_proj.weight', 'image_model.encoder.layers.0.self_attn.out_proj.bias', 'image_model.encoder.layers.0.self_attn.out_proj.weight', 'image_model.encoder.layers.0.self_attn.q_proj.bias', 'image_model.encoder.layers.0.self_attn.q_proj.weight', 'image_model.encoder.layers.0.self_attn.v_proj.bias', 'image_model.encoder.layers.0.self_attn.v_proj.weight', 'image_model.encoder.layers.1.layer_norm1.bias', 'image_model.encoder.layers.1.layer_norm1.weight', 'image_model.encoder.layers.1.layer_norm2.bias', 'image_model.encoder.layers.1.layer_norm2.weight', 'image_model.encoder.layers.1.mlp.fc1.bias', 'image_model.encoder.layers.1.mlp.fc1.weight', 'image_model.encoder.layers.1.mlp.fc2.bias', 'image_model.encoder.layers.1.mlp.fc2.weight', 'image_model.encoder.layers.1.self_attn.k_proj.bias', 'image_model.encoder.layers.1.self_attn.k_proj.weight', 'image_model.encoder.layers.1.self_attn.out_proj.bias', 'image_model.encoder.layers.1.self_attn.out_proj.weight', 'image_model.encoder.layers.1.self_attn.q_proj.bias', 'image_model.encoder.layers.1.self_attn.q_proj.weight', 'image_model.encoder.layers.1.self_attn.v_proj.bias', 'image_model.encoder.layers.1.self_attn.v_proj.weight', 'image_model.encoder.layers.10.layer_norm1.bias', 'image_model.encoder.layers.10.layer_norm1.weight', 'image_model.encoder.layers.10.layer_norm2.bias', 'image_model.encoder.layers.10.layer_norm2.weight', 'image_model.encoder.layers.10.mlp.fc1.bias', 'image_model.encoder.layers.10.mlp.fc1.weight', 'image_model.encoder.layers.10.mlp.fc2.bias', 'image_model.encoder.layers.10.mlp.fc2.weight', 'image_model.encoder.layers.10.self_attn.k_proj.bias', 'image_model.encoder.layers.10.self_attn.k_proj.weight', 'image_model.encoder.layers.10.self_attn.out_proj.bias', 'image_model.encoder.layers.10.self_attn.out_proj.weight', 'image_model.encoder.layers.10.self_attn.q_proj.bias', 'image_model.encoder.layers.10.self_attn.q_proj.weight', 'image_model.encoder.layers.10.self_attn.v_proj.bias', 'image_model.encoder.layers.10.self_attn.v_proj.weight', 'image_model.encoder.layers.11.layer_norm1.bias', 'image_model.encoder.layers.11.layer_norm1.weight', 'image_model.encoder.layers.11.layer_norm2.bias', 'image_model.encoder.layers.11.layer_norm2.weight', 'image_model.encoder.layers.11.mlp.fc1.bias', 'image_model.encoder.layers.11.mlp.fc1.weight', 'image_model.encoder.layers.11.mlp.fc2.bias', 'image_model.encoder.layers.11.mlp.fc2.weight', 'image_model.encoder.layers.11.self_attn.k_proj.bias', 'image_model.encoder.layers.11.self_attn.k_proj.weight', 'image_model.encoder.layers.11.self_attn.out_proj.bias', 'image_model.encoder.layers.11.self_attn.out_proj.weight', 'image_model.encoder.layers.11.self_attn.q_proj.bias', 'image_model.encoder.layers.11.self_attn.q_proj.weight', 'image_model.encoder.layers.11.self_attn.v_proj.bias', 'image_model.encoder.layers.11.self_attn.v_proj.weight', 'image_model.encoder.layers.12.layer_norm1.bias', 'image_model.encoder.layers.12.layer_norm1.weight', 'image_model.encoder.layers.12.layer_norm2.bias', 'image_model.encoder.layers.12.layer_norm2.weight', 'image_model.encoder.layers.12.mlp.fc1.bias', 'image_model.encoder.layers.12.mlp.fc1.weight', 'image_model.encoder.layers.12.mlp.fc2.bias', 'image_model.encoder.layers.12.mlp.fc2.weight', 'image_model.encoder.layers.12.self_attn.k_proj.bias', 'image_model.encoder.layers.12.self_attn.k_proj.weight', 'image_model.encoder.layers.12.self_attn.out_proj.bias', 'image_model.encoder.layers.12.self_attn.out_proj.weight', 'image_model.encoder.layers.12.self_attn.q_proj.bias', 'image_model.encoder.layers.12.self_attn.q_proj.weight', 'image_model.encoder.layers.12.self_attn.v_proj.bias', 'image_model.encoder.layers.12.self_attn.v_proj.weight', 'image_model.encoder.layers.13.layer_norm1.bias', 'image_model.encoder.layers.13.layer_norm1.weight', 'image_model.encoder.layers.13.layer_norm2.bias', 'image_model.encoder.layers.13.layer_norm2.weight', 'image_model.encoder.layers.13.mlp.fc1.bias', 'image_model.encoder.layers.13.mlp.fc1.weight', 'image_model.encoder.layers.13.mlp.fc2.bias', 'image_model.encoder.layers.13.mlp.fc2.weight', 'image_model.encoder.layers.13.self_attn.k_proj.bias', 'image_model.encoder.layers.13.self_attn.k_proj.weight', 'image_model.encoder.layers.13.self_attn.out_proj.bias', 'image_model.encoder.layers.13.self_attn.out_proj.weight', 'image_model.encoder.layers.13.self_attn.q_proj.bias', 'image_model.encoder.layers.13.self_attn.q_proj.weight', 'image_model.encoder.layers.13.self_attn.v_proj.bias', 'image_model.encoder.layers.13.self_attn.v_proj.weight', 'image_model.encoder.layers.14.layer_norm1.bias', 'image_model.encoder.layers.14.layer_norm1.weight', 'image_model.encoder.layers.14.layer_norm2.bias', 'image_model.encoder.layers.14.layer_norm2.weight', 'image_model.encoder.layers.14.mlp.fc1.bias', 'image_model.encoder.layers.14.mlp.fc1.weight', 'image_model.encoder.layers.14.mlp.fc2.bias', 'image_model.encoder.layers.14.mlp.fc2.weight', 'image_model.encoder.layers.14.self_attn.k_proj.bias', 'image_model.encoder.layers.14.self_attn.k_proj.weight', 'image_model.encoder.layers.14.self_attn.out_proj.bias', 'image_model.encoder.layers.14.self_attn.out_proj.weight', 'image_model.encoder.layers.14.self_attn.q_proj.bias', 'image_model.encoder.layers.14.self_attn.q_proj.weight', 'image_model.encoder.layers.14.self_attn.v_proj.bias', 'image_model.encoder.layers.14.self_attn.v_proj.weight', 'image_model.encoder.layers.15.layer_norm1.bias', 'image_model.encoder.layers.15.layer_norm1.weight', 'image_model.encoder.layers.15.layer_norm2.bias', 'image_model.encoder.layers.15.layer_norm2.weight', 'image_model.encoder.layers.15.mlp.fc1.bias', 'image_model.encoder.layers.15.mlp.fc1.weight', 'image_model.encoder.layers.15.mlp.fc2.bias', 'image_model.encoder.layers.15.mlp.fc2.weight', 'image_model.encoder.layers.15.self_attn.k_proj.bias', 'image_model.encoder.layers.15.self_attn.k_proj.weight', 'image_model.encoder.layers.15.self_attn.out_proj.bias', 'image_model.encoder.layers.15.self_attn.out_proj.weight', 'image_model.encoder.layers.15.self_attn.q_proj.bias', 'image_model.encoder.layers.15.self_attn.q_proj.weight', 'image_model.encoder.layers.15.self_attn.v_proj.bias', 'image_model.encoder.layers.15.self_attn.v_proj.weight', 'image_model.encoder.layers.16.layer_norm1.bias', 'image_model.encoder.layers.16.layer_norm1.weight', 'image_model.encoder.layers.16.layer_norm2.bias', 'image_model.encoder.layers.16.layer_norm2.weight', 'image_model.encoder.layers.16.mlp.fc1.bias', 'image_model.encoder.layers.16.mlp.fc1.weight', 'image_model.encoder.layers.16.mlp.fc2.bias', 'image_model.encoder.layers.16.mlp.fc2.weight', 'image_model.encoder.layers.16.self_attn.k_proj.bias', 'image_model.encoder.layers.16.self_attn.k_proj.weight', 'image_model.encoder.layers.16.self_attn.out_proj.bias', 'image_model.encoder.layers.16.self_attn.out_proj.weight', 'image_model.encoder.layers.16.self_attn.q_proj.bias', 'image_model.encoder.layers.16.self_attn.q_proj.weight', 'image_model.encoder.layers.16.self_attn.v_proj.bias', 'image_model.encoder.layers.16.self_attn.v_proj.weight', 'image_model.encoder.layers.17.layer_norm1.bias', 'image_model.encoder.layers.17.layer_norm1.weight', 'image_model.encoder.layers.17.layer_norm2.bias', 'image_model.encoder.layers.17.layer_norm2.weight', 'image_model.encoder.layers.17.mlp.fc1.bias', 'image_model.encoder.layers.17.mlp.fc1.weight', 'image_model.encoder.layers.17.mlp.fc2.bias', 'image_model.encoder.layers.17.mlp.fc2.weight', 'image_model.encoder.layers.17.self_attn.k_proj.bias', 'image_model.encoder.layers.17.self_attn.k_proj.weight', 'image_model.encoder.layers.17.self_attn.out_proj.bias', 'image_model.encoder.layers.17.self_attn.out_proj.weight', 'image_model.encoder.layers.17.self_attn.q_proj.bias', 'image_model.encoder.layers.17.self_attn.q_proj.weight', 'image_model.encoder.layers.17.self_attn.v_proj.bias', 'image_model.encoder.layers.17.self_attn.v_proj.weight', 'image_model.encoder.layers.18.layer_norm1.bias', 'image_model.encoder.layers.18.layer_norm1.weight', 'image_model.encoder.layers.18.layer_norm2.bias', 'image_model.encoder.layers.18.layer_norm2.weight', 'image_model.encoder.layers.18.mlp.fc1.bias', 'image_model.encoder.layers.18.mlp.fc1.weight', 'image_model.encoder.layers.18.mlp.fc2.bias', 'image_model.encoder.layers.18.mlp.fc2.weight', 'image_model.encoder.layers.18.self_attn.k_proj.bias', 'image_model.encoder.layers.18.self_attn.k_proj.weight', 'image_model.encoder.layers.18.self_attn.out_proj.bias', 'image_model.encoder.layers.18.self_attn.out_proj.weight', 'image_model.encoder.layers.18.self_attn.q_proj.bias', 'image_model.encoder.layers.18.self_attn.q_proj.weight', 'image_model.encoder.layers.18.self_attn.v_proj.bias', 'image_model.encoder.layers.18.self_attn.v_proj.weight', 'image_model.encoder.layers.19.layer_norm1.bias', 'image_model.encoder.layers.19.layer_norm1.weight', 'image_model.encoder.layers.19.layer_norm2.bias', 'image_model.encoder.layers.19.layer_norm2.weight', 'image_model.encoder.layers.19.mlp.fc1.bias', 'image_model.encoder.layers.19.mlp.fc1.weight', 'image_model.encoder.layers.19.mlp.fc2.bias', 'image_model.encoder.layers.19.mlp.fc2.weight', 'image_model.encoder.layers.19.self_attn.k_proj.bias', 'image_model.encoder.layers.19.self_attn.k_proj.weight', 'image_model.encoder.layers.19.self_attn.out_proj.bias', 'image_model.encoder.layers.19.self_attn.out_proj.weight', 'image_model.encoder.layers.19.self_attn.q_proj.bias', 'image_model.encoder.layers.19.self_attn.q_proj.weight', 'image_model.encoder.layers.19.self_attn.v_proj.bias', 'image_model.encoder.layers.19.self_attn.v_proj.weight', 'image_model.encoder.layers.2.layer_norm1.bias', 'image_model.encoder.layers.2.layer_norm1.weight', 'image_model.encoder.layers.2.layer_norm2.bias', 'image_model.encoder.layers.2.layer_norm2.weight', 'image_model.encoder.layers.2.mlp.fc1.bias', 'image_model.encoder.layers.2.mlp.fc1.weight', 'image_model.encoder.layers.2.mlp.fc2.bias', 'image_model.encoder.layers.2.mlp.fc2.weight', 'image_model.encoder.layers.2.self_attn.k_proj.bias', 'image_model.encoder.layers.2.self_attn.k_proj.weight', 'image_model.encoder.layers.2.self_attn.out_proj.bias', 'image_model.encoder.layers.2.self_attn.out_proj.weight', 'image_model.encoder.layers.2.self_attn.q_proj.bias', 'image_model.encoder.layers.2.self_attn.q_proj.weight', 'image_model.encoder.layers.2.self_attn.v_proj.bias', 'image_model.encoder.layers.2.self_attn.v_proj.weight', 'image_model.encoder.layers.20.layer_norm1.bias', 'image_model.encoder.layers.20.layer_norm1.weight', 'image_model.encoder.layers.20.layer_norm2.bias', 'image_model.encoder.layers.20.layer_norm2.weight', 'image_model.encoder.layers.20.mlp.fc1.bias', 'image_model.encoder.layers.20.mlp.fc1.weight', 'image_model.encoder.layers.20.mlp.fc2.bias', 'image_model.encoder.layers.20.mlp.fc2.weight', 'image_model.encoder.layers.20.self_attn.k_proj.bias', 'image_model.encoder.layers.20.self_attn.k_proj.weight', 'image_model.encoder.layers.20.self_attn.out_proj.bias', 'image_model.encoder.layers.20.self_attn.out_proj.weight', 'image_model.encoder.layers.20.self_attn.q_proj.bias', 'image_model.encoder.layers.20.self_attn.q_proj.weight', 'image_model.encoder.layers.20.self_attn.v_proj.bias', 'image_model.encoder.layers.20.self_attn.v_proj.weight', 'image_model.encoder.layers.21.layer_norm1.bias', 'image_model.encoder.layers.21.layer_norm1.weight', 'image_model.encoder.layers.21.layer_norm2.bias', 'image_model.encoder.layers.21.layer_norm2.weight', 'image_model.encoder.layers.21.mlp.fc1.bias', 'image_model.encoder.layers.21.mlp.fc1.weight', 'image_model.encoder.layers.21.mlp.fc2.bias', 'image_model.encoder.layers.21.mlp.fc2.weight', 'image_model.encoder.layers.21.self_attn.k_proj.bias', 'image_model.encoder.layers.21.self_attn.k_proj.weight', 'image_model.encoder.layers.21.self_attn.out_proj.bias', 'image_model.encoder.layers.21.self_attn.out_proj.weight', 'image_model.encoder.layers.21.self_attn.q_proj.bias', 'image_model.encoder.layers.21.self_attn.q_proj.weight', 'image_model.encoder.layers.21.self_attn.v_proj.bias', 'image_model.encoder.layers.21.self_attn.v_proj.weight', 'image_model.encoder.layers.22.layer_norm1.bias', 'image_model.encoder.layers.22.layer_norm1.weight', 'image_model.encoder.layers.22.layer_norm2.bias', 'image_model.encoder.layers.22.layer_norm2.weight', 'image_model.encoder.layers.22.mlp.fc1.bias', 'image_model.encoder.layers.22.mlp.fc1.weight', 'image_model.encoder.layers.22.mlp.fc2.bias', 'image_model.encoder.layers.22.mlp.fc2.weight', 'image_model.encoder.layers.22.self_attn.k_proj.bias', 'image_model.encoder.layers.22.self_attn.k_proj.weight', 'image_model.encoder.layers.22.self_attn.out_proj.bias', 'image_model.encoder.layers.22.self_attn.out_proj.weight', 'image_model.encoder.layers.22.self_attn.q_proj.bias', 'image_model.encoder.layers.22.self_attn.q_proj.weight', 'image_model.encoder.layers.22.self_attn.v_proj.bias', 'image_model.encoder.layers.22.self_attn.v_proj.weight', 'image_model.encoder.layers.23.layer_norm1.bias', 'image_model.encoder.layers.23.layer_norm1.weight', 'image_model.encoder.layers.23.layer_norm2.bias', 'image_model.encoder.layers.23.layer_norm2.weight', 'image_model.encoder.layers.23.mlp.fc1.bias', 'image_model.encoder.layers.23.mlp.fc1.weight', 'image_model.encoder.layers.23.mlp.fc2.bias', 'image_model.encoder.layers.23.mlp.fc2.weight', 'image_model.encoder.layers.23.self_attn.k_proj.bias', 'image_model.encoder.layers.23.self_attn.k_proj.weight', 'image_model.encoder.layers.23.self_attn.out_proj.bias', 'image_model.encoder.layers.23.self_attn.out_proj.weight', 'image_model.encoder.layers.23.self_attn.q_proj.bias', 'image_model.encoder.layers.23.self_attn.q_proj.weight', 'image_model.encoder.layers.23.self_attn.v_proj.bias', 'image_model.encoder.layers.23.self_attn.v_proj.weight', 'image_model.encoder.layers.3.layer_norm1.bias', 'image_model.encoder.layers.3.layer_norm1.weight', 'image_model.encoder.layers.3.layer_norm2.bias', 'image_model.encoder.layers.3.layer_norm2.weight', 'image_model.encoder.layers.3.mlp.fc1.bias', 'image_model.encoder.layers.3.mlp.fc1.weight', 'image_model.encoder.layers.3.mlp.fc2.bias', 'image_model.encoder.layers.3.mlp.fc2.weight', 'image_model.encoder.layers.3.self_attn.k_proj.bias', 'image_model.encoder.layers.3.self_attn.k_proj.weight', 'image_model.encoder.layers.3.self_attn.out_proj.bias', 'image_model.encoder.layers.3.self_attn.out_proj.weight', 'image_model.encoder.layers.3.self_attn.q_proj.bias', 'image_model.encoder.layers.3.self_attn.q_proj.weight', 'image_model.encoder.layers.3.self_attn.v_proj.bias', 'image_model.encoder.layers.3.self_attn.v_proj.weight', 'image_model.encoder.layers.4.layer_norm1.bias', 'image_model.encoder.layers.4.layer_norm1.weight', 'image_model.encoder.layers.4.layer_norm2.bias', 'image_model.encoder.layers.4.layer_norm2.weight', 'image_model.encoder.layers.4.mlp.fc1.bias', 'image_model.encoder.layers.4.mlp.fc1.weight', 'image_model.encoder.layers.4.mlp.fc2.bias', 'image_model.encoder.layers.4.mlp.fc2.weight', 'image_model.encoder.layers.4.self_attn.k_proj.bias', 'image_model.encoder.layers.4.self_attn.k_proj.weight', 'image_model.encoder.layers.4.self_attn.out_proj.bias', 'image_model.encoder.layers.4.self_attn.out_proj.weight', 'image_model.encoder.layers.4.self_attn.q_proj.bias', 'image_model.encoder.layers.4.self_attn.q_proj.weight', 'image_model.encoder.layers.4.self_attn.v_proj.bias', 'image_model.encoder.layers.4.self_attn.v_proj.weight', 'image_model.encoder.layers.5.layer_norm1.bias', 'image_model.encoder.layers.5.layer_norm1.weight', 'image_model.encoder.layers.5.layer_norm2.bias', 'image_model.encoder.layers.5.layer_norm2.weight', 'image_model.encoder.layers.5.mlp.fc1.bias', 'image_model.encoder.layers.5.mlp.fc1.weight', 'image_model.encoder.layers.5.mlp.fc2.bias', 'image_model.encoder.layers.5.mlp.fc2.weight', 'image_model.encoder.layers.5.self_attn.k_proj.bias', 'image_model.encoder.layers.5.self_attn.k_proj.weight', 'image_model.encoder.layers.5.self_attn.out_proj.bias', 'image_model.encoder.layers.5.self_attn.out_proj.weight', 'image_model.encoder.layers.5.self_attn.q_proj.bias', 'image_model.encoder.layers.5.self_attn.q_proj.weight', 'image_model.encoder.layers.5.self_attn.v_proj.bias', 'image_model.encoder.layers.5.self_attn.v_proj.weight', 'image_model.encoder.layers.6.layer_norm1.bias', 'image_model.encoder.layers.6.layer_norm1.weight', 'image_model.encoder.layers.6.layer_norm2.bias', 'image_model.encoder.layers.6.layer_norm2.weight', 'image_model.encoder.layers.6.mlp.fc1.bias', 'image_model.encoder.layers.6.mlp.fc1.weight', 'image_model.encoder.layers.6.mlp.fc2.bias', 'image_model.encoder.layers.6.mlp.fc2.weight', 'image_model.encoder.layers.6.self_attn.k_proj.bias', 'image_model.encoder.layers.6.self_attn.k_proj.weight', 'image_model.encoder.layers.6.self_attn.out_proj.bias', 'image_model.encoder.layers.6.self_attn.out_proj.weight', 'image_model.encoder.layers.6.self_attn.q_proj.bias', 'image_model.encoder.layers.6.self_attn.q_proj.weight', 'image_model.encoder.layers.6.self_attn.v_proj.bias', 'image_model.encoder.layers.6.self_attn.v_proj.weight', 'image_model.encoder.layers.7.layer_norm1.bias', 'image_model.encoder.layers.7.layer_norm1.weight', 'image_model.encoder.layers.7.layer_norm2.bias', 'image_model.encoder.layers.7.layer_norm2.weight', 'image_model.encoder.layers.7.mlp.fc1.bias', 'image_model.encoder.layers.7.mlp.fc1.weight', 'image_model.encoder.layers.7.mlp.fc2.bias', 'image_model.encoder.layers.7.mlp.fc2.weight', 'image_model.encoder.layers.7.self_attn.k_proj.bias', 'image_model.encoder.layers.7.self_attn.k_proj.weight', 'image_model.encoder.layers.7.self_attn.out_proj.bias', 'image_model.encoder.layers.7.self_attn.out_proj.weight', 'image_model.encoder.layers.7.self_attn.q_proj.bias', 'image_model.encoder.layers.7.self_attn.q_proj.weight', 'image_model.encoder.layers.7.self_attn.v_proj.bias', 'image_model.encoder.layers.7.self_attn.v_proj.weight', 'image_model.encoder.layers.8.layer_norm1.bias', 'image_model.encoder.layers.8.layer_norm1.weight', 'image_model.encoder.layers.8.layer_norm2.bias', 'image_model.encoder.layers.8.layer_norm2.weight', 'image_model.encoder.layers.8.mlp.fc1.bias', 'image_model.encoder.layers.8.mlp.fc1.weight', 'image_model.encoder.layers.8.mlp.fc2.bias', 'image_model.encoder.layers.8.mlp.fc2.weight', 'image_model.encoder.layers.8.self_attn.k_proj.bias', 'image_model.encoder.layers.8.self_attn.k_proj.weight', 'image_model.encoder.layers.8.self_attn.out_proj.bias', 'image_model.encoder.layers.8.self_attn.out_proj.weight', 'image_model.encoder.layers.8.self_attn.q_proj.bias', 'image_model.encoder.layers.8.self_attn.q_proj.weight', 'image_model.encoder.layers.8.self_attn.v_proj.bias', 'image_model.encoder.layers.8.self_attn.v_proj.weight', 'image_model.encoder.layers.9.layer_norm1.bias', 'image_model.encoder.layers.9.layer_norm1.weight', 'image_model.encoder.layers.9.layer_norm2.bias', 'image_model.encoder.layers.9.layer_norm2.weight', 'image_model.encoder.layers.9.mlp.fc1.bias', 'image_model.encoder.layers.9.mlp.fc1.weight', 'image_model.encoder.layers.9.mlp.fc2.bias', 'image_model.encoder.layers.9.mlp.fc2.weight', 'image_model.encoder.layers.9.self_attn.k_proj.bias', 'image_model.encoder.layers.9.self_attn.k_proj.weight', 'image_model.encoder.layers.9.self_attn.out_proj.bias', 'image_model.encoder.layers.9.self_attn.out_proj.weight', 'image_model.encoder.layers.9.self_attn.q_proj.bias', 'image_model.encoder.layers.9.self_attn.q_proj.weight', 'image_model.encoder.layers.9.self_attn.v_proj.bias', 'image_model.encoder.layers.9.self_attn.v_proj.weight', 'image_model.post_layernorm.bias', 'image_model.post_layernorm.weight', 'image_model.pre_layrnorm.bias', 'image_model.pre_layrnorm.weight']
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.

[guozonghao96]

If you encounter this warning in the eval phrase, it is ok, because the weights of CLIP-ViT will be loaded again after transformers.from_pretrain().

[mzamini92]

Thanks. I also had another question. in the paper you mentioned:
To deal with images with varied aspect ratios, LLaVA-1.5 pads the input images into squares before feeding them into the visual encoder. This encoding method results in a waste of computation for non-square images. For example, a 1:4 image has only 25% effective computation after padding into squares. To quantify the influence, we train an unpadded version of LLaVA-1.5, by fitting the ViT position embedding into the aspect ratio of input images using 2D interpolation.
based on your train.sh in your pretraining you didn't use image_aspect_ratio and in the FT you used image_aspect_ratio pad. Are you referring to that in your paper?

[guozonghao96]

We don't use image_aspect_ratio to control the preprocess of input image. We keep the naive resolution of the input image.