Update josie_dict.txt

josie_dict.txt

@@ -1,403 +1,578 @@
-Model's state_dict:
-encoder.modality_preprocessors.vision.cls_token          torch.Size([1, 1, 768])
-encoder.modality_preprocessors.vision.rgbt_stem.proj.1.weight    torch.Size([768, 3, 2, 14, 14])
-encoder.modality_preprocessors.vision.pos_embedding_helper.pos_embed     torch.Size([1, 7681, 768])
+Qwen2-0.5B as the Reasoner for tisrt sucsessfull multimodal-realtime inference.
+
+On a MacBook Air M1 8GB with ca. 3s latency.
+On RTX 4090 with ca. 800ms.
+
+
+encoder.modality_preprocessors.vision.cls_token          torch.Size([1, 1, 1280])
+encoder.modality_preprocessors.vision.rgbt_stem.proj.1.weight    torch.Size([1280, 3, 2, 14, 14])
+encoder.modality_preprocessors.vision.pos_embedding_helper.pos_embed     torch.Size([1, 257, 1280])
 encoder.modality_preprocessors.audio.cls_token   torch.Size([1, 1, 768])
 encoder.modality_preprocessors.audio.rgbt_stem.proj.weight       torch.Size([768, 1, 16, 16])
 encoder.modality_preprocessors.audio.rgbt_stem.norm_layer.weight         torch.Size([768])
+encoder.modality_preprocessors.audio.rgbt_stem.norm_layer.bias   torch.Size([768])
 encoder.modality_preprocessors.audio.pos_embedding_helper.pos_embed      torch.Size([1, 229, 768])
-encoder.modality_preprocessors.depth.cls_token   torch.Size([1, 1, 384])
-encoder.modality_preprocessors.depth.depth_stem.proj.weight      torch.Size([384, 1, 16, 16])
-encoder.modality_preprocessors.depth.depth_stem.norm_layer.weight        torch.Size([384])
-encoder.modality_preprocessors.depth.pos_embedding_helper.pos_embed      torch.Size([1, 197, 384])
-encoder.modality_preprocessors.thermal.cls_token         torch.Size([1, 1, 768])
-encoder.modality_preprocessors.thermal.rgbt_stem.proj.weight     torch.Size([768, 1, 16, 16])
-encoder.modality_preprocessors.thermal.rgbt_stem.norm_layer.weight       torch.Size([768])
-encoder.modality_preprocessors.thermal.pos_embedding_helper.pos_embed    torch.Size([1, 197, 768])
-encoder.modality_transformers.vision.pre_transformer_layer.0.weight      torch.Size([768])
-encoder.modality_transformers.vision.blocks.0.attn.in_proj_weight        torch.Size([2304, 768])
-encoder.modality_transformers.vision.blocks.0.attn.in_proj_bias          torch.Size([2304])
-encoder.modality_transformers.vision.blocks.0.attn.out_proj.weight       torch.Size([768, 768])
-encoder.modality_transformers.vision.blocks.0.attn.out_proj.bias         torch.Size([768])
-encoder.modality_transformers.vision.blocks.0.norm1.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.0.norm2.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.0.mlp.w1.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.0.mlp.w2.weight      torch.Size([768, 512])
-encoder.modality_transformers.vision.blocks.0.mlp.w3.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.1.attn.in_proj_weight        torch.Size([2304, 768])
-encoder.modality_transformers.vision.blocks.1.attn.in_proj_bias          torch.Size([2304])
-encoder.modality_transformers.vision.blocks.1.attn.out_proj.weight       torch.Size([768, 768])
-encoder.modality_transformers.vision.blocks.1.attn.out_proj.bias         torch.Size([768])
-encoder.modality_transformers.vision.blocks.1.norm1.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.1.norm2.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.1.mlp.w1.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.1.mlp.w2.weight      torch.Size([768, 512])
-encoder.modality_transformers.vision.blocks.1.mlp.w3.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.2.attn.in_proj_weight        torch.Size([2304, 768])
-encoder.modality_transformers.vision.blocks.2.attn.in_proj_bias          torch.Size([2304])
-encoder.modality_transformers.vision.blocks.2.attn.out_proj.weight       torch.Size([768, 768])
-encoder.modality_transformers.vision.blocks.2.attn.out_proj.bias         torch.Size([768])
-encoder.modality_transformers.vision.blocks.2.norm1.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.2.norm2.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.2.mlp.w1.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.2.mlp.w2.weight      torch.Size([768, 512])
-encoder.modality_transformers.vision.blocks.2.mlp.w3.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.3.attn.in_proj_weight        torch.Size([2304, 768])
-encoder.modality_transformers.vision.blocks.3.attn.in_proj_bias          torch.Size([2304])
-encoder.modality_transformers.vision.blocks.3.attn.out_proj.weight       torch.Size([768, 768])
-encoder.modality_transformers.vision.blocks.3.attn.out_proj.bias         torch.Size([768])
-encoder.modality_transformers.vision.blocks.3.norm1.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.3.norm2.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.3.mlp.w1.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.3.mlp.w2.weight      torch.Size([768, 512])
-encoder.modality_transformers.vision.blocks.3.mlp.w3.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.4.attn.in_proj_weight        torch.Size([2304, 768])
-encoder.modality_transformers.vision.blocks.4.attn.in_proj_bias          torch.Size([2304])
-encoder.modality_transformers.vision.blocks.4.attn.out_proj.weight       torch.Size([768, 768])
-encoder.modality_transformers.vision.blocks.4.attn.out_proj.bias         torch.Size([768])
-encoder.modality_transformers.vision.blocks.4.norm1.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.4.norm2.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.4.mlp.w1.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.4.mlp.w2.weight      torch.Size([768, 512])
-encoder.modality_transformers.vision.blocks.4.mlp.w3.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.5.attn.in_proj_weight        torch.Size([2304, 768])
-encoder.modality_transformers.vision.blocks.5.attn.in_proj_bias          torch.Size([2304])
-encoder.modality_transformers.vision.blocks.5.attn.out_proj.weight       torch.Size([768, 768])
-encoder.modality_transformers.vision.blocks.5.attn.out_proj.bias         torch.Size([768])
-encoder.modality_transformers.vision.blocks.5.norm1.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.5.norm2.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.5.mlp.w1.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.5.mlp.w2.weight      torch.Size([768, 512])
-encoder.modality_transformers.vision.blocks.5.mlp.w3.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.6.attn.in_proj_weight        torch.Size([2304, 768])
-encoder.modality_transformers.vision.blocks.6.attn.in_proj_bias          torch.Size([2304])
-encoder.modality_transformers.vision.blocks.6.attn.out_proj.weight       torch.Size([768, 768])
-encoder.modality_transformers.vision.blocks.6.attn.out_proj.bias         torch.Size([768])
-encoder.modality_transformers.vision.blocks.6.norm1.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.6.norm2.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.6.mlp.w1.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.6.mlp.w2.weight      torch.Size([768, 512])
-encoder.modality_transformers.vision.blocks.6.mlp.w3.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.7.attn.in_proj_weight        torch.Size([2304, 768])
-encoder.modality_transformers.vision.blocks.7.attn.in_proj_bias          torch.Size([2304])
-encoder.modality_transformers.vision.blocks.7.attn.out_proj.weight       torch.Size([768, 768])
-encoder.modality_transformers.vision.blocks.7.attn.out_proj.bias         torch.Size([768])
-encoder.modality_transformers.vision.blocks.7.norm1.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.7.norm2.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.7.mlp.w1.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.7.mlp.w2.weight      torch.Size([768, 512])
-encoder.modality_transformers.vision.blocks.7.mlp.w3.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.8.attn.in_proj_weight        torch.Size([2304, 768])
-encoder.modality_transformers.vision.blocks.8.attn.in_proj_bias          torch.Size([2304])
-encoder.modality_transformers.vision.blocks.8.attn.out_proj.weight       torch.Size([768, 768])
-encoder.modality_transformers.vision.blocks.8.attn.out_proj.bias         torch.Size([768])
-encoder.modality_transformers.vision.blocks.8.norm1.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.8.norm2.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.8.mlp.w1.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.8.mlp.w2.weight      torch.Size([768, 512])
-encoder.modality_transformers.vision.blocks.8.mlp.w3.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.9.attn.in_proj_weight        torch.Size([2304, 768])
-encoder.modality_transformers.vision.blocks.9.attn.in_proj_bias          torch.Size([2304])
-encoder.modality_transformers.vision.blocks.9.attn.out_proj.weight       torch.Size([768, 768])
-encoder.modality_transformers.vision.blocks.9.attn.out_proj.bias         torch.Size([768])
-encoder.modality_transformers.vision.blocks.9.norm1.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.9.norm2.weight       torch.Size([768])
-encoder.modality_transformers.vision.blocks.9.mlp.w1.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.9.mlp.w2.weight      torch.Size([768, 512])
-encoder.modality_transformers.vision.blocks.9.mlp.w3.weight      torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.10.attn.in_proj_weight       torch.Size([2304, 768])
-encoder.modality_transformers.vision.blocks.10.attn.in_proj_bias         torch.Size([2304])
-encoder.modality_transformers.vision.blocks.10.attn.out_proj.weight      torch.Size([768, 768])
-encoder.modality_transformers.vision.blocks.10.attn.out_proj.bias        torch.Size([768])
-encoder.modality_transformers.vision.blocks.10.norm1.weight      torch.Size([768])
-encoder.modality_transformers.vision.blocks.10.norm2.weight      torch.Size([768])
-encoder.modality_transformers.vision.blocks.10.mlp.w1.weight     torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.10.mlp.w2.weight     torch.Size([768, 512])
-encoder.modality_transformers.vision.blocks.10.mlp.w3.weight     torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.11.attn.in_proj_weight       torch.Size([2304, 768])
-encoder.modality_transformers.vision.blocks.11.attn.in_proj_bias         torch.Size([2304])
-encoder.modality_transformers.vision.blocks.11.attn.out_proj.weight      torch.Size([768, 768])
-encoder.modality_transformers.vision.blocks.11.attn.out_proj.bias        torch.Size([768])
-encoder.modality_transformers.vision.blocks.11.norm1.weight      torch.Size([768])
-encoder.modality_transformers.vision.blocks.11.norm2.weight      torch.Size([768])
-encoder.modality_transformers.vision.blocks.11.mlp.w1.weight     torch.Size([512, 768])
-encoder.modality_transformers.vision.blocks.11.mlp.w2.weight     torch.Size([768, 512])
-encoder.modality_transformers.vision.blocks.11.mlp.w3.weight     torch.Size([512, 768])
-encoder.modality_transformers.audio.pre_transformer_layer.0.weight       torch.Size([768])
-encoder.modality_transformers.audio.blocks.0.attn.in_proj_weight         torch.Size([2304, 768])
-encoder.modality_transformers.audio.blocks.0.attn.in_proj_bias   torch.Size([2304])
-encoder.modality_transformers.audio.blocks.0.attn.bias_k         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.0.attn.bias_v         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.0.attn.out_proj.weight        torch.Size([768, 768])
-encoder.modality_transformers.audio.blocks.0.attn.out_proj.bias          torch.Size([768])
-encoder.modality_transformers.audio.blocks.0.norm1.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.0.norm2.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.0.mlp.w1.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.0.mlp.w2.weight       torch.Size([768, 512])
-encoder.modality_transformers.audio.blocks.0.mlp.w3.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.1.attn.in_proj_weight         torch.Size([2304, 768])
-encoder.modality_transformers.audio.blocks.1.attn.in_proj_bias   torch.Size([2304])
-encoder.modality_transformers.audio.blocks.1.attn.bias_k         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.1.attn.bias_v         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.1.attn.out_proj.weight        torch.Size([768, 768])
-encoder.modality_transformers.audio.blocks.1.attn.out_proj.bias          torch.Size([768])
-encoder.modality_transformers.audio.blocks.1.norm1.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.1.norm2.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.1.mlp.w1.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.1.mlp.w2.weight       torch.Size([768, 512])
-encoder.modality_transformers.audio.blocks.1.mlp.w3.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.2.attn.in_proj_weight         torch.Size([2304, 768])
-encoder.modality_transformers.audio.blocks.2.attn.in_proj_bias   torch.Size([2304])
-encoder.modality_transformers.audio.blocks.2.attn.bias_k         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.2.attn.bias_v         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.2.attn.out_proj.weight        torch.Size([768, 768])
-encoder.modality_transformers.audio.blocks.2.attn.out_proj.bias          torch.Size([768])
-encoder.modality_transformers.audio.blocks.2.norm1.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.2.norm2.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.2.mlp.w1.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.2.mlp.w2.weight       torch.Size([768, 512])
-encoder.modality_transformers.audio.blocks.2.mlp.w3.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.3.attn.in_proj_weight         torch.Size([2304, 768])
-encoder.modality_transformers.audio.blocks.3.attn.in_proj_bias   torch.Size([2304])
-encoder.modality_transformers.audio.blocks.3.attn.bias_k         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.3.attn.bias_v         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.3.attn.out_proj.weight        torch.Size([768, 768])
-encoder.modality_transformers.audio.blocks.3.attn.out_proj.bias          torch.Size([768])
-encoder.modality_transformers.audio.blocks.3.norm1.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.3.norm2.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.3.mlp.w1.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.3.mlp.w2.weight       torch.Size([768, 512])
-encoder.modality_transformers.audio.blocks.3.mlp.w3.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.4.attn.in_proj_weight         torch.Size([2304, 768])
-encoder.modality_transformers.audio.blocks.4.attn.in_proj_bias   torch.Size([2304])
-encoder.modality_transformers.audio.blocks.4.attn.bias_k         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.4.attn.bias_v         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.4.attn.out_proj.weight        torch.Size([768, 768])
-encoder.modality_transformers.audio.blocks.4.attn.out_proj.bias          torch.Size([768])
-encoder.modality_transformers.audio.blocks.4.norm1.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.4.norm2.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.4.mlp.w1.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.4.mlp.w2.weight       torch.Size([768, 512])
-encoder.modality_transformers.audio.blocks.4.mlp.w3.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.5.attn.in_proj_weight         torch.Size([2304, 768])
-encoder.modality_transformers.audio.blocks.5.attn.in_proj_bias   torch.Size([2304])
-encoder.modality_transformers.audio.blocks.5.attn.bias_k         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.5.attn.bias_v         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.5.attn.out_proj.weight        torch.Size([768, 768])
-encoder.modality_transformers.audio.blocks.5.attn.out_proj.bias          torch.Size([768])
-encoder.modality_transformers.audio.blocks.5.norm1.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.5.norm2.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.5.mlp.w1.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.5.mlp.w2.weight       torch.Size([768, 512])
-encoder.modality_transformers.audio.blocks.5.mlp.w3.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.6.attn.in_proj_weight         torch.Size([2304, 768])
-encoder.modality_transformers.audio.blocks.6.attn.in_proj_bias   torch.Size([2304])
-encoder.modality_transformers.audio.blocks.6.attn.bias_k         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.6.attn.bias_v         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.6.attn.out_proj.weight        torch.Size([768, 768])
-encoder.modality_transformers.audio.blocks.6.attn.out_proj.bias          torch.Size([768])
-encoder.modality_transformers.audio.blocks.6.norm1.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.6.norm2.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.6.mlp.w1.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.6.mlp.w2.weight       torch.Size([768, 512])
-encoder.modality_transformers.audio.blocks.6.mlp.w3.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.7.attn.in_proj_weight         torch.Size([2304, 768])
-encoder.modality_transformers.audio.blocks.7.attn.in_proj_bias   torch.Size([2304])
-encoder.modality_transformers.audio.blocks.7.attn.bias_k         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.7.attn.bias_v         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.7.attn.out_proj.weight        torch.Size([768, 768])
-encoder.modality_transformers.audio.blocks.7.attn.out_proj.bias          torch.Size([768])
-encoder.modality_transformers.audio.blocks.7.norm1.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.7.norm2.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.7.mlp.w1.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.7.mlp.w2.weight       torch.Size([768, 512])
-encoder.modality_transformers.audio.blocks.7.mlp.w3.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.8.attn.in_proj_weight         torch.Size([2304, 768])
-encoder.modality_transformers.audio.blocks.8.attn.in_proj_bias   torch.Size([2304])
-encoder.modality_transformers.audio.blocks.8.attn.bias_k         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.8.attn.bias_v         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.8.attn.out_proj.weight        torch.Size([768, 768])
-encoder.modality_transformers.audio.blocks.8.attn.out_proj.bias          torch.Size([768])
-encoder.modality_transformers.audio.blocks.8.norm1.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.8.norm2.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.8.mlp.w1.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.8.mlp.w2.weight       torch.Size([768, 512])
-encoder.modality_transformers.audio.blocks.8.mlp.w3.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.9.attn.in_proj_weight         torch.Size([2304, 768])
-encoder.modality_transformers.audio.blocks.9.attn.in_proj_bias   torch.Size([2304])
-encoder.modality_transformers.audio.blocks.9.attn.bias_k         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.9.attn.bias_v         torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.9.attn.out_proj.weight        torch.Size([768, 768])
-encoder.modality_transformers.audio.blocks.9.attn.out_proj.bias          torch.Size([768])
-encoder.modality_transformers.audio.blocks.9.norm1.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.9.norm2.weight        torch.Size([768])
-encoder.modality_transformers.audio.blocks.9.mlp.w1.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.9.mlp.w2.weight       torch.Size([768, 512])
-encoder.modality_transformers.audio.blocks.9.mlp.w3.weight       torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.10.attn.in_proj_weight        torch.Size([2304, 768])
-encoder.modality_transformers.audio.blocks.10.attn.in_proj_bias          torch.Size([2304])
-encoder.modality_transformers.audio.blocks.10.attn.bias_k        torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.10.attn.bias_v        torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.10.attn.out_proj.weight       torch.Size([768, 768])
-encoder.modality_transformers.audio.blocks.10.attn.out_proj.bias         torch.Size([768])
-encoder.modality_transformers.audio.blocks.10.norm1.weight       torch.Size([768])
-encoder.modality_transformers.audio.blocks.10.norm2.weight       torch.Size([768])
-encoder.modality_transformers.audio.blocks.10.mlp.w1.weight      torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.10.mlp.w2.weight      torch.Size([768, 512])
-encoder.modality_transformers.audio.blocks.10.mlp.w3.weight      torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.11.attn.in_proj_weight        torch.Size([2304, 768])
-encoder.modality_transformers.audio.blocks.11.attn.in_proj_bias          torch.Size([2304])
-encoder.modality_transformers.audio.blocks.11.attn.bias_k        torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.11.attn.bias_v        torch.Size([1, 1, 768])
-encoder.modality_transformers.audio.blocks.11.attn.out_proj.weight       torch.Size([768, 768])
-encoder.modality_transformers.audio.blocks.11.attn.out_proj.bias         torch.Size([768])
-encoder.modality_transformers.audio.blocks.11.norm1.weight       torch.Size([768])
-encoder.modality_transformers.audio.blocks.11.norm2.weight       torch.Size([768])
-encoder.modality_transformers.audio.blocks.11.mlp.w1.weight      torch.Size([512, 768])
-encoder.modality_transformers.audio.blocks.11.mlp.w2.weight      torch.Size([768, 512])
-encoder.modality_transformers.audio.blocks.11.mlp.w3.weight      torch.Size([512, 768])
-encoder.modality_transformers.depth.pre_transformer_layer.0.weight       torch.Size([384])
-encoder.modality_transformers.depth.blocks.0.attn.in_proj_weight         torch.Size([1152, 384])
-encoder.modality_transformers.depth.blocks.0.attn.in_proj_bias   torch.Size([1152])
-encoder.modality_transformers.depth.blocks.0.attn.bias_k         torch.Size([1, 1, 384])
-encoder.modality_transformers.depth.blocks.0.attn.bias_v         torch.Size([1, 1, 384])
-encoder.modality_transformers.depth.blocks.0.attn.out_proj.weight        torch.Size([384, 384])
-encoder.modality_transformers.depth.blocks.0.attn.out_proj.bias          torch.Size([384])
-encoder.modality_transformers.depth.blocks.0.norm1.weight        torch.Size([384])
-encoder.modality_transformers.depth.blocks.0.norm2.weight        torch.Size([384])
-encoder.modality_transformers.depth.blocks.0.mlp.w1.weight       torch.Size([256, 384])
-encoder.modality_transformers.depth.blocks.0.mlp.w2.weight       torch.Size([384, 256])
-encoder.modality_transformers.depth.blocks.0.mlp.w3.weight       torch.Size([256, 384])
-encoder.modality_transformers.depth.blocks.1.attn.in_proj_weight         torch.Size([1152, 384])
-encoder.modality_transformers.depth.blocks.1.attn.in_proj_bias   torch.Size([1152])
-encoder.modality_transformers.depth.blocks.1.attn.bias_k         torch.Size([1, 1, 384])
-encoder.modality_transformers.depth.blocks.1.attn.bias_v         torch.Size([1, 1, 384])
-encoder.modality_transformers.depth.blocks.1.attn.out_proj.weight        torch.Size([384, 384])
-encoder.modality_transformers.depth.blocks.1.attn.out_proj.bias          torch.Size([384])
-encoder.modality_transformers.depth.blocks.1.norm1.weight        torch.Size([384])
-encoder.modality_transformers.depth.blocks.1.norm2.weight        torch.Size([384])
-encoder.modality_transformers.depth.blocks.1.mlp.w1.weight       torch.Size([256, 384])
-encoder.modality_transformers.depth.blocks.1.mlp.w2.weight       torch.Size([384, 256])
-encoder.modality_transformers.depth.blocks.1.mlp.w3.weight       torch.Size([256, 384])
-encoder.modality_transformers.depth.blocks.2.attn.in_proj_weight         torch.Size([1152, 384])
-encoder.modality_transformers.depth.blocks.2.attn.in_proj_bias   torch.Size([1152])
-encoder.modality_transformers.depth.blocks.2.attn.bias_k         torch.Size([1, 1, 384])
-encoder.modality_transformers.depth.blocks.2.attn.bias_v         torch.Size([1, 1, 384])
-encoder.modality_transformers.depth.blocks.2.attn.out_proj.weight        torch.Size([384, 384])
-encoder.modality_transformers.depth.blocks.2.attn.out_proj.bias          torch.Size([384])
-encoder.modality_transformers.depth.blocks.2.norm1.weight        torch.Size([384])
-encoder.modality_transformers.depth.blocks.2.norm2.weight        torch.Size([384])
-encoder.modality_transformers.depth.blocks.2.mlp.w1.weight       torch.Size([256, 384])
-encoder.modality_transformers.depth.blocks.2.mlp.w2.weight       torch.Size([384, 256])
-encoder.modality_transformers.depth.blocks.2.mlp.w3.weight       torch.Size([256, 384])
-encoder.modality_transformers.depth.blocks.3.attn.in_proj_weight         torch.Size([1152, 384])
-encoder.modality_transformers.depth.blocks.3.attn.in_proj_bias   torch.Size([1152])
-encoder.modality_transformers.depth.blocks.3.attn.bias_k         torch.Size([1, 1, 384])
-encoder.modality_transformers.depth.blocks.3.attn.bias_v         torch.Size([1, 1, 384])
-encoder.modality_transformers.depth.blocks.3.attn.out_proj.weight        torch.Size([384, 384])
-encoder.modality_transformers.depth.blocks.3.attn.out_proj.bias          torch.Size([384])
-encoder.modality_transformers.depth.blocks.3.norm1.weight        torch.Size([384])
-encoder.modality_transformers.depth.blocks.3.norm2.weight        torch.Size([384])
-encoder.modality_transformers.depth.blocks.3.mlp.w1.weight       torch.Size([256, 384])
-encoder.modality_transformers.depth.blocks.3.mlp.w2.weight       torch.Size([384, 256])
-encoder.modality_transformers.depth.blocks.3.mlp.w3.weight       torch.Size([256, 384])
-encoder.modality_transformers.depth.blocks.4.attn.in_proj_weight         torch.Size([1152, 384])
-encoder.modality_transformers.depth.blocks.4.attn.in_proj_bias   torch.Size([1152])
-encoder.modality_transformers.depth.blocks.4.attn.bias_k         torch.Size([1, 1, 384])
-encoder.modality_transformers.depth.blocks.4.attn.bias_v         torch.Size([1, 1, 384])
-encoder.modality_transformers.depth.blocks.4.attn.out_proj.weight        torch.Size([384, 384])
-encoder.modality_transformers.depth.blocks.4.attn.out_proj.bias          torch.Size([384])
-encoder.modality_transformers.depth.blocks.4.norm1.weight        torch.Size([384])
-encoder.modality_transformers.depth.blocks.4.norm2.weight        torch.Size([384])
-encoder.modality_transformers.depth.blocks.4.mlp.w1.weight       torch.Size([256, 384])
-encoder.modality_transformers.depth.blocks.4.mlp.w2.weight       torch.Size([384, 256])
-encoder.modality_transformers.depth.blocks.4.mlp.w3.weight       torch.Size([256, 384])
-encoder.modality_transformers.depth.blocks.5.attn.in_proj_weight         torch.Size([1152, 384])
-encoder.modality_transformers.depth.blocks.5.attn.in_proj_bias   torch.Size([1152])
-encoder.modality_transformers.depth.blocks.5.attn.bias_k         torch.Size([1, 1, 384])
-encoder.modality_transformers.depth.blocks.5.attn.bias_v         torch.Size([1, 1, 384])
-encoder.modality_transformers.depth.blocks.5.attn.out_proj.weight        torch.Size([384, 384])
-encoder.modality_transformers.depth.blocks.5.attn.out_proj.bias          torch.Size([384])
-encoder.modality_transformers.depth.blocks.5.norm1.weight        torch.Size([384])
-encoder.modality_transformers.depth.blocks.5.norm2.weight        torch.Size([384])
-encoder.modality_transformers.depth.blocks.5.mlp.w1.weight       torch.Size([256, 384])
-encoder.modality_transformers.depth.blocks.5.mlp.w2.weight       torch.Size([384, 256])
-encoder.modality_transformers.depth.blocks.5.mlp.w3.weight       torch.Size([256, 384])
-encoder.modality_transformers.thermal.pre_transformer_layer.0.weight     torch.Size([768])
-encoder.modality_transformers.thermal.blocks.0.attn.in_proj_weight       torch.Size([2304, 768])
-encoder.modality_transformers.thermal.blocks.0.attn.in_proj_bias         torch.Size([2304])
-encoder.modality_transformers.thermal.blocks.0.attn.bias_k       torch.Size([1, 1, 768])
-encoder.modality_transformers.thermal.blocks.0.attn.bias_v       torch.Size([1, 1, 768])
-encoder.modality_transformers.thermal.blocks.0.attn.out_proj.weight      torch.Size([768, 768])
-encoder.modality_transformers.thermal.blocks.0.attn.out_proj.bias        torch.Size([768])
-encoder.modality_transformers.thermal.blocks.0.norm1.weight      torch.Size([768])
-encoder.modality_transformers.thermal.blocks.0.norm2.weight      torch.Size([768])
-encoder.modality_transformers.thermal.blocks.0.mlp.w1.weight     torch.Size([512, 768])
-encoder.modality_transformers.thermal.blocks.0.mlp.w2.weight     torch.Size([768, 512])
-encoder.modality_transformers.thermal.blocks.0.mlp.w3.weight     torch.Size([512, 768])
-encoder.modality_transformers.thermal.blocks.1.attn.in_proj_weight       torch.Size([2304, 768])
-encoder.modality_transformers.thermal.blocks.1.attn.in_proj_bias         torch.Size([2304])
-encoder.modality_transformers.thermal.blocks.1.attn.bias_k       torch.Size([1, 1, 768])
-encoder.modality_transformers.thermal.blocks.1.attn.bias_v       torch.Size([1, 1, 768])
-encoder.modality_transformers.thermal.blocks.1.attn.out_proj.weight      torch.Size([768, 768])
-encoder.modality_transformers.thermal.blocks.1.attn.out_proj.bias        torch.Size([768])
-encoder.modality_transformers.thermal.blocks.1.norm1.weight      torch.Size([768])
-encoder.modality_transformers.thermal.blocks.1.norm2.weight      torch.Size([768])
-encoder.modality_transformers.thermal.blocks.1.mlp.w1.weight     torch.Size([512, 768])
-encoder.modality_transformers.thermal.blocks.1.mlp.w2.weight     torch.Size([768, 512])
-encoder.modality_transformers.thermal.blocks.1.mlp.w3.weight     torch.Size([512, 768])
-encoder.modality_transformers.thermal.blocks.2.attn.in_proj_weight       torch.Size([2304, 768])
-encoder.modality_transformers.thermal.blocks.2.attn.in_proj_bias         torch.Size([2304])
-encoder.modality_transformers.thermal.blocks.2.attn.bias_k       torch.Size([1, 1, 768])
-encoder.modality_transformers.thermal.blocks.2.attn.bias_v       torch.Size([1, 1, 768])
-encoder.modality_transformers.thermal.blocks.2.attn.out_proj.weight      torch.Size([768, 768])
-encoder.modality_transformers.thermal.blocks.2.attn.out_proj.bias        torch.Size([768])
-encoder.modality_transformers.thermal.blocks.2.norm1.weight      torch.Size([768])
-encoder.modality_transformers.thermal.blocks.2.norm2.weight      torch.Size([768])
-encoder.modality_transformers.thermal.blocks.2.mlp.w1.weight     torch.Size([512, 768])
-encoder.modality_transformers.thermal.blocks.2.mlp.w2.weight     torch.Size([768, 512])
-encoder.modality_transformers.thermal.blocks.2.mlp.w3.weight     torch.Size([512, 768])
-encoder.modality_transformers.thermal.blocks.3.attn.in_proj_weight       torch.Size([2304, 768])
-encoder.modality_transformers.thermal.blocks.3.attn.in_proj_bias         torch.Size([2304])
-encoder.modality_transformers.thermal.blocks.3.attn.bias_k       torch.Size([1, 1, 768])
-encoder.modality_transformers.thermal.blocks.3.attn.bias_v       torch.Size([1, 1, 768])
-encoder.modality_transformers.thermal.blocks.3.attn.out_proj.weight      torch.Size([768, 768])
-encoder.modality_transformers.thermal.blocks.3.attn.out_proj.bias        torch.Size([768])
-encoder.modality_transformers.thermal.blocks.3.norm1.weight      torch.Size([768])
-encoder.modality_transformers.thermal.blocks.3.norm2.weight      torch.Size([768])
-encoder.modality_transformers.thermal.blocks.3.mlp.w1.weight     torch.Size([512, 768])
-encoder.modality_transformers.thermal.blocks.3.mlp.w2.weight     torch.Size([768, 512])
-encoder.modality_transformers.thermal.blocks.3.mlp.w3.weight     torch.Size([512, 768])
-encoder.modality_transformers.thermal.blocks.4.attn.in_proj_weight       torch.Size([2304, 768])
-encoder.modality_transformers.thermal.blocks.4.attn.in_proj_bias         torch.Size([2304])
-encoder.modality_transformers.thermal.blocks.4.attn.bias_k       torch.Size([1, 1, 768])
-encoder.modality_transformers.thermal.blocks.4.attn.bias_v       torch.Size([1, 1, 768])
-encoder.modality_transformers.thermal.blocks.4.attn.out_proj.weight      torch.Size([768, 768])
-encoder.modality_transformers.thermal.blocks.4.attn.out_proj.bias        torch.Size([768])
-encoder.modality_transformers.thermal.blocks.4.norm1.weight      torch.Size([768])
-encoder.modality_transformers.thermal.blocks.4.norm2.weight      torch.Size([768])
-encoder.modality_transformers.thermal.blocks.4.mlp.w1.weight     torch.Size([512, 768])
-encoder.modality_transformers.thermal.blocks.4.mlp.w2.weight     torch.Size([768, 512])
-encoder.modality_transformers.thermal.blocks.4.mlp.w3.weight     torch.Size([512, 768])
-encoder.modality_transformers.thermal.blocks.5.attn.in_proj_weight       torch.Size([2304, 768])
-encoder.modality_transformers.thermal.blocks.5.attn.in_proj_bias         torch.Size([2304])
-encoder.modality_transformers.thermal.blocks.5.attn.bias_k       torch.Size([1, 1, 768])
-encoder.modality_transformers.thermal.blocks.5.attn.bias_v       torch.Size([1, 1, 768])
-encoder.modality_transformers.thermal.blocks.5.attn.out_proj.weight      torch.Size([768, 768])
-encoder.modality_transformers.thermal.blocks.5.attn.out_proj.bias        torch.Size([768])
-encoder.modality_transformers.thermal.blocks.5.norm1.weight      torch.Size([768])
-encoder.modality_transformers.thermal.blocks.5.norm2.weight      torch.Size([768])
-encoder.modality_transformers.thermal.blocks.5.mlp.w1.weight     torch.Size([512, 768])
-encoder.modality_transformers.thermal.blocks.5.mlp.w2.weight     torch.Size([768, 512])
-encoder.modality_transformers.thermal.blocks.5.mlp.w3.weight     torch.Size([512, 768])
-encoder.modality_heads.vision.0.weight   torch.Size([768])
-encoder.modality_heads.vision.2.weight   torch.Size([1024, 768])
+encoder.modality_trunks.vision.pre_transformer_layer.0.weight    torch.Size([1280])
+encoder.modality_trunks.vision.pre_transformer_layer.0.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.0.attn.in_proj_weight      torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.0.attn.in_proj_bias        torch.Size([3840])
+encoder.modality_trunks.vision.blocks.0.attn.out_proj.weight     torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.0.attn.out_proj.bias       torch.Size([1280])
+encoder.modality_trunks.vision.blocks.0.norm_1.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.0.norm_1.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.0.mlp.fc1.weight   torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.0.mlp.fc1.bias     torch.Size([5120])
+encoder.modality_trunks.vision.blocks.0.mlp.fc2.weight   torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.0.mlp.fc2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.0.norm_2.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.0.norm_2.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.1.attn.in_proj_weight      torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.1.attn.in_proj_bias        torch.Size([3840])
+encoder.modality_trunks.vision.blocks.1.attn.out_proj.weight     torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.1.attn.out_proj.bias       torch.Size([1280])
+encoder.modality_trunks.vision.blocks.1.norm_1.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.1.norm_1.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.1.mlp.fc1.weight   torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.1.mlp.fc1.bias     torch.Size([5120])
+encoder.modality_trunks.vision.blocks.1.mlp.fc2.weight   torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.1.mlp.fc2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.1.norm_2.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.1.norm_2.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.2.attn.in_proj_weight      torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.2.attn.in_proj_bias        torch.Size([3840])
+encoder.modality_trunks.vision.blocks.2.attn.out_proj.weight     torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.2.attn.out_proj.bias       torch.Size([1280])
+encoder.modality_trunks.vision.blocks.2.norm_1.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.2.norm_1.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.2.mlp.fc1.weight   torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.2.mlp.fc1.bias     torch.Size([5120])
+encoder.modality_trunks.vision.blocks.2.mlp.fc2.weight   torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.2.mlp.fc2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.2.norm_2.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.2.norm_2.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.3.attn.in_proj_weight      torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.3.attn.in_proj_bias        torch.Size([3840])
+encoder.modality_trunks.vision.blocks.3.attn.out_proj.weight     torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.3.attn.out_proj.bias       torch.Size([1280])
+encoder.modality_trunks.vision.blocks.3.norm_1.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.3.norm_1.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.3.mlp.fc1.weight   torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.3.mlp.fc1.bias     torch.Size([5120])
+encoder.modality_trunks.vision.blocks.3.mlp.fc2.weight   torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.3.mlp.fc2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.3.norm_2.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.3.norm_2.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.4.attn.in_proj_weight      torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.4.attn.in_proj_bias        torch.Size([3840])
+encoder.modality_trunks.vision.blocks.4.attn.out_proj.weight     torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.4.attn.out_proj.bias       torch.Size([1280])
+encoder.modality_trunks.vision.blocks.4.norm_1.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.4.norm_1.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.4.mlp.fc1.weight   torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.4.mlp.fc1.bias     torch.Size([5120])
+encoder.modality_trunks.vision.blocks.4.mlp.fc2.weight   torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.4.mlp.fc2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.4.norm_2.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.4.norm_2.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.5.attn.in_proj_weight      torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.5.attn.in_proj_bias        torch.Size([3840])
+encoder.modality_trunks.vision.blocks.5.attn.out_proj.weight     torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.5.attn.out_proj.bias       torch.Size([1280])
+encoder.modality_trunks.vision.blocks.5.norm_1.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.5.norm_1.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.5.mlp.fc1.weight   torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.5.mlp.fc1.bias     torch.Size([5120])
+encoder.modality_trunks.vision.blocks.5.mlp.fc2.weight   torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.5.mlp.fc2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.5.norm_2.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.5.norm_2.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.6.attn.in_proj_weight      torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.6.attn.in_proj_bias        torch.Size([3840])
+encoder.modality_trunks.vision.blocks.6.attn.out_proj.weight     torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.6.attn.out_proj.bias       torch.Size([1280])
+encoder.modality_trunks.vision.blocks.6.norm_1.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.6.norm_1.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.6.mlp.fc1.weight   torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.6.mlp.fc1.bias     torch.Size([5120])
+encoder.modality_trunks.vision.blocks.6.mlp.fc2.weight   torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.6.mlp.fc2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.6.norm_2.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.6.norm_2.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.7.attn.in_proj_weight      torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.7.attn.in_proj_bias        torch.Size([3840])
+encoder.modality_trunks.vision.blocks.7.attn.out_proj.weight     torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.7.attn.out_proj.bias       torch.Size([1280])
+encoder.modality_trunks.vision.blocks.7.norm_1.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.7.norm_1.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.7.mlp.fc1.weight   torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.7.mlp.fc1.bias     torch.Size([5120])
+encoder.modality_trunks.vision.blocks.7.mlp.fc2.weight   torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.7.mlp.fc2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.7.norm_2.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.7.norm_2.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.8.attn.in_proj_weight      torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.8.attn.in_proj_bias        torch.Size([3840])
+encoder.modality_trunks.vision.blocks.8.attn.out_proj.weight     torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.8.attn.out_proj.bias       torch.Size([1280])
+encoder.modality_trunks.vision.blocks.8.norm_1.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.8.norm_1.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.8.mlp.fc1.weight   torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.8.mlp.fc1.bias     torch.Size([5120])
+encoder.modality_trunks.vision.blocks.8.mlp.fc2.weight   torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.8.mlp.fc2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.8.norm_2.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.8.norm_2.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.9.attn.in_proj_weight      torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.9.attn.in_proj_bias        torch.Size([3840])
+encoder.modality_trunks.vision.blocks.9.attn.out_proj.weight     torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.9.attn.out_proj.bias       torch.Size([1280])
+encoder.modality_trunks.vision.blocks.9.norm_1.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.9.norm_1.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.9.mlp.fc1.weight   torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.9.mlp.fc1.bias     torch.Size([5120])
+encoder.modality_trunks.vision.blocks.9.mlp.fc2.weight   torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.9.mlp.fc2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.9.norm_2.weight    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.9.norm_2.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.10.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.10.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.10.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.10.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.10.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.10.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.10.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.10.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.10.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.10.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.10.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.10.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.11.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.11.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.11.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.11.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.11.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.11.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.11.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.11.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.11.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.11.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.11.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.11.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.12.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.12.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.12.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.12.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.12.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.12.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.12.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.12.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.12.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.12.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.12.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.12.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.13.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.13.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.13.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.13.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.13.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.13.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.13.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.13.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.13.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.13.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.13.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.13.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.14.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.14.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.14.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.14.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.14.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.14.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.14.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.14.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.14.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.14.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.14.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.14.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.15.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.15.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.15.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.15.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.15.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.15.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.15.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.15.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.15.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.15.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.15.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.15.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.16.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.16.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.16.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.16.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.16.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.16.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.16.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.16.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.16.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.16.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.16.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.16.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.17.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.17.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.17.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.17.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.17.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.17.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.17.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.17.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.17.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.17.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.17.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.17.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.18.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.18.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.18.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.18.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.18.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.18.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.18.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.18.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.18.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.18.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.18.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.18.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.19.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.19.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.19.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.19.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.19.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.19.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.19.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.19.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.19.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.19.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.19.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.19.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.20.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.20.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.20.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.20.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.20.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.20.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.20.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.20.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.20.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.20.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.20.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.20.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.21.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.21.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.21.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.21.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.21.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.21.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.21.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.21.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.21.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.21.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.21.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.21.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.22.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.22.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.22.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.22.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.22.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.22.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.22.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.22.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.22.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.22.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.22.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.22.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.23.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.23.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.23.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.23.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.23.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.23.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.23.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.23.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.23.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.23.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.23.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.23.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.24.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.24.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.24.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.24.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.24.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.24.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.24.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.24.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.24.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.24.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.24.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.24.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.25.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.25.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.25.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.25.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.25.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.25.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.25.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.25.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.25.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.25.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.25.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.25.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.26.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.26.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.26.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.26.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.26.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.26.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.26.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.26.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.26.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.26.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.26.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.26.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.27.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.27.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.27.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.27.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.27.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.27.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.27.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.27.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.27.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.27.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.27.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.27.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.28.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.28.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.28.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.28.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.28.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.28.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.28.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.28.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.28.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.28.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.28.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.28.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.29.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.29.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.29.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.29.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.29.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.29.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.29.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.29.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.29.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.29.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.29.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.29.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.30.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.30.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.30.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.30.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.30.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.30.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.30.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.30.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.30.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.30.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.30.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.30.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.31.attn.in_proj_weight     torch.Size([3840, 1280])
+encoder.modality_trunks.vision.blocks.31.attn.in_proj_bias       torch.Size([3840])
+encoder.modality_trunks.vision.blocks.31.attn.out_proj.weight    torch.Size([1280, 1280])
+encoder.modality_trunks.vision.blocks.31.attn.out_proj.bias      torch.Size([1280])
+encoder.modality_trunks.vision.blocks.31.norm_1.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.31.norm_1.bias     torch.Size([1280])
+encoder.modality_trunks.vision.blocks.31.mlp.fc1.weight          torch.Size([5120, 1280])
+encoder.modality_trunks.vision.blocks.31.mlp.fc1.bias    torch.Size([5120])
+encoder.modality_trunks.vision.blocks.31.mlp.fc2.weight          torch.Size([1280, 5120])
+encoder.modality_trunks.vision.blocks.31.mlp.fc2.bias    torch.Size([1280])
+encoder.modality_trunks.vision.blocks.31.norm_2.weight   torch.Size([1280])
+encoder.modality_trunks.vision.blocks.31.norm_2.bias     torch.Size([1280])
+encoder.modality_trunks.audio.blocks.0.attn.in_proj_weight       torch.Size([2304, 768])
+encoder.modality_trunks.audio.blocks.0.attn.in_proj_bias         torch.Size([2304])
+encoder.modality_trunks.audio.blocks.0.attn.bias_k       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.0.attn.bias_v       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.0.attn.out_proj.weight      torch.Size([768, 768])
+encoder.modality_trunks.audio.blocks.0.attn.out_proj.bias        torch.Size([768])
+encoder.modality_trunks.audio.blocks.0.norm_1.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.0.norm_1.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.0.mlp.fc1.weight    torch.Size([3072, 768])
+encoder.modality_trunks.audio.blocks.0.mlp.fc1.bias      torch.Size([3072])
+encoder.modality_trunks.audio.blocks.0.mlp.fc2.weight    torch.Size([768, 3072])
+encoder.modality_trunks.audio.blocks.0.mlp.fc2.bias      torch.Size([768])
+encoder.modality_trunks.audio.blocks.0.norm_2.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.0.norm_2.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.1.attn.in_proj_weight       torch.Size([2304, 768])
+encoder.modality_trunks.audio.blocks.1.attn.in_proj_bias         torch.Size([2304])
+encoder.modality_trunks.audio.blocks.1.attn.bias_k       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.1.attn.bias_v       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.1.attn.out_proj.weight      torch.Size([768, 768])
+encoder.modality_trunks.audio.blocks.1.attn.out_proj.bias        torch.Size([768])
+encoder.modality_trunks.audio.blocks.1.norm_1.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.1.norm_1.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.1.mlp.fc1.weight    torch.Size([3072, 768])
+encoder.modality_trunks.audio.blocks.1.mlp.fc1.bias      torch.Size([3072])
+encoder.modality_trunks.audio.blocks.1.mlp.fc2.weight    torch.Size([768, 3072])
+encoder.modality_trunks.audio.blocks.1.mlp.fc2.bias      torch.Size([768])
+encoder.modality_trunks.audio.blocks.1.norm_2.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.1.norm_2.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.2.attn.in_proj_weight       torch.Size([2304, 768])
+encoder.modality_trunks.audio.blocks.2.attn.in_proj_bias         torch.Size([2304])
+encoder.modality_trunks.audio.blocks.2.attn.bias_k       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.2.attn.bias_v       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.2.attn.out_proj.weight      torch.Size([768, 768])
+encoder.modality_trunks.audio.blocks.2.attn.out_proj.bias        torch.Size([768])
+encoder.modality_trunks.audio.blocks.2.norm_1.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.2.norm_1.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.2.mlp.fc1.weight    torch.Size([3072, 768])
+encoder.modality_trunks.audio.blocks.2.mlp.fc1.bias      torch.Size([3072])
+encoder.modality_trunks.audio.blocks.2.mlp.fc2.weight    torch.Size([768, 3072])
+encoder.modality_trunks.audio.blocks.2.mlp.fc2.bias      torch.Size([768])
+encoder.modality_trunks.audio.blocks.2.norm_2.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.2.norm_2.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.3.attn.in_proj_weight       torch.Size([2304, 768])
+encoder.modality_trunks.audio.blocks.3.attn.in_proj_bias         torch.Size([2304])
+encoder.modality_trunks.audio.blocks.3.attn.bias_k       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.3.attn.bias_v       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.3.attn.out_proj.weight      torch.Size([768, 768])
+encoder.modality_trunks.audio.blocks.3.attn.out_proj.bias        torch.Size([768])
+encoder.modality_trunks.audio.blocks.3.norm_1.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.3.norm_1.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.3.mlp.fc1.weight    torch.Size([3072, 768])
+encoder.modality_trunks.audio.blocks.3.mlp.fc1.bias      torch.Size([3072])
+encoder.modality_trunks.audio.blocks.3.mlp.fc2.weight    torch.Size([768, 3072])
+encoder.modality_trunks.audio.blocks.3.mlp.fc2.bias      torch.Size([768])
+encoder.modality_trunks.audio.blocks.3.norm_2.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.3.norm_2.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.4.attn.in_proj_weight       torch.Size([2304, 768])
+encoder.modality_trunks.audio.blocks.4.attn.in_proj_bias         torch.Size([2304])
+encoder.modality_trunks.audio.blocks.4.attn.bias_k       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.4.attn.bias_v       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.4.attn.out_proj.weight      torch.Size([768, 768])
+encoder.modality_trunks.audio.blocks.4.attn.out_proj.bias        torch.Size([768])
+encoder.modality_trunks.audio.blocks.4.norm_1.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.4.norm_1.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.4.mlp.fc1.weight    torch.Size([3072, 768])
+encoder.modality_trunks.audio.blocks.4.mlp.fc1.bias      torch.Size([3072])
+encoder.modality_trunks.audio.blocks.4.mlp.fc2.weight    torch.Size([768, 3072])
+encoder.modality_trunks.audio.blocks.4.mlp.fc2.bias      torch.Size([768])
+encoder.modality_trunks.audio.blocks.4.norm_2.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.4.norm_2.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.5.attn.in_proj_weight       torch.Size([2304, 768])
+encoder.modality_trunks.audio.blocks.5.attn.in_proj_bias         torch.Size([2304])
+encoder.modality_trunks.audio.blocks.5.attn.bias_k       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.5.attn.bias_v       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.5.attn.out_proj.weight      torch.Size([768, 768])
+encoder.modality_trunks.audio.blocks.5.attn.out_proj.bias        torch.Size([768])
+encoder.modality_trunks.audio.blocks.5.norm_1.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.5.norm_1.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.5.mlp.fc1.weight    torch.Size([3072, 768])
+encoder.modality_trunks.audio.blocks.5.mlp.fc1.bias      torch.Size([3072])
+encoder.modality_trunks.audio.blocks.5.mlp.fc2.weight    torch.Size([768, 3072])
+encoder.modality_trunks.audio.blocks.5.mlp.fc2.bias      torch.Size([768])
+encoder.modality_trunks.audio.blocks.5.norm_2.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.5.norm_2.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.6.attn.in_proj_weight       torch.Size([2304, 768])
+encoder.modality_trunks.audio.blocks.6.attn.in_proj_bias         torch.Size([2304])
+encoder.modality_trunks.audio.blocks.6.attn.bias_k       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.6.attn.bias_v       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.6.attn.out_proj.weight      torch.Size([768, 768])
+encoder.modality_trunks.audio.blocks.6.attn.out_proj.bias        torch.Size([768])
+encoder.modality_trunks.audio.blocks.6.norm_1.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.6.norm_1.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.6.mlp.fc1.weight    torch.Size([3072, 768])
+encoder.modality_trunks.audio.blocks.6.mlp.fc1.bias      torch.Size([3072])
+encoder.modality_trunks.audio.blocks.6.mlp.fc2.weight    torch.Size([768, 3072])
+encoder.modality_trunks.audio.blocks.6.mlp.fc2.bias      torch.Size([768])
+encoder.modality_trunks.audio.blocks.6.norm_2.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.6.norm_2.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.7.attn.in_proj_weight       torch.Size([2304, 768])
+encoder.modality_trunks.audio.blocks.7.attn.in_proj_bias         torch.Size([2304])
+encoder.modality_trunks.audio.blocks.7.attn.bias_k       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.7.attn.bias_v       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.7.attn.out_proj.weight      torch.Size([768, 768])
+encoder.modality_trunks.audio.blocks.7.attn.out_proj.bias        torch.Size([768])
+encoder.modality_trunks.audio.blocks.7.norm_1.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.7.norm_1.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.7.mlp.fc1.weight    torch.Size([3072, 768])
+encoder.modality_trunks.audio.blocks.7.mlp.fc1.bias      torch.Size([3072])
+encoder.modality_trunks.audio.blocks.7.mlp.fc2.weight    torch.Size([768, 3072])
+encoder.modality_trunks.audio.blocks.7.mlp.fc2.bias      torch.Size([768])
+encoder.modality_trunks.audio.blocks.7.norm_2.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.7.norm_2.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.8.attn.in_proj_weight       torch.Size([2304, 768])
+encoder.modality_trunks.audio.blocks.8.attn.in_proj_bias         torch.Size([2304])
+encoder.modality_trunks.audio.blocks.8.attn.bias_k       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.8.attn.bias_v       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.8.attn.out_proj.weight      torch.Size([768, 768])
+encoder.modality_trunks.audio.blocks.8.attn.out_proj.bias        torch.Size([768])
+encoder.modality_trunks.audio.blocks.8.norm_1.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.8.norm_1.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.8.mlp.fc1.weight    torch.Size([3072, 768])
+encoder.modality_trunks.audio.blocks.8.mlp.fc1.bias      torch.Size([3072])
+encoder.modality_trunks.audio.blocks.8.mlp.fc2.weight    torch.Size([768, 3072])
+encoder.modality_trunks.audio.blocks.8.mlp.fc2.bias      torch.Size([768])
+encoder.modality_trunks.audio.blocks.8.norm_2.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.8.norm_2.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.9.attn.in_proj_weight       torch.Size([2304, 768])
+encoder.modality_trunks.audio.blocks.9.attn.in_proj_bias         torch.Size([2304])
+encoder.modality_trunks.audio.blocks.9.attn.bias_k       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.9.attn.bias_v       torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.9.attn.out_proj.weight      torch.Size([768, 768])
+encoder.modality_trunks.audio.blocks.9.attn.out_proj.bias        torch.Size([768])
+encoder.modality_trunks.audio.blocks.9.norm_1.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.9.norm_1.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.9.mlp.fc1.weight    torch.Size([3072, 768])
+encoder.modality_trunks.audio.blocks.9.mlp.fc1.bias      torch.Size([3072])
+encoder.modality_trunks.audio.blocks.9.mlp.fc2.weight    torch.Size([768, 3072])
+encoder.modality_trunks.audio.blocks.9.mlp.fc2.bias      torch.Size([768])
+encoder.modality_trunks.audio.blocks.9.norm_2.weight     torch.Size([768])
+encoder.modality_trunks.audio.blocks.9.norm_2.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.10.attn.in_proj_weight      torch.Size([2304, 768])
+encoder.modality_trunks.audio.blocks.10.attn.in_proj_bias        torch.Size([2304])
+encoder.modality_trunks.audio.blocks.10.attn.bias_k      torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.10.attn.bias_v      torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.10.attn.out_proj.weight     torch.Size([768, 768])
+encoder.modality_trunks.audio.blocks.10.attn.out_proj.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.10.norm_1.weight    torch.Size([768])
+encoder.modality_trunks.audio.blocks.10.norm_1.bias      torch.Size([768])
+encoder.modality_trunks.audio.blocks.10.mlp.fc1.weight   torch.Size([3072, 768])
+encoder.modality_trunks.audio.blocks.10.mlp.fc1.bias     torch.Size([3072])
+encoder.modality_trunks.audio.blocks.10.mlp.fc2.weight   torch.Size([768, 3072])
+encoder.modality_trunks.audio.blocks.10.mlp.fc2.bias     torch.Size([768])
+encoder.modality_trunks.audio.blocks.10.norm_2.weight    torch.Size([768])
+encoder.modality_trunks.audio.blocks.10.norm_2.bias      torch.Size([768])
+encoder.modality_trunks.audio.blocks.11.attn.in_proj_weight      torch.Size([2304, 768])
+encoder.modality_trunks.audio.blocks.11.attn.in_proj_bias        torch.Size([2304])
+encoder.modality_trunks.audio.blocks.11.attn.bias_k      torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.11.attn.bias_v      torch.Size([1, 1, 768])
+encoder.modality_trunks.audio.blocks.11.attn.out_proj.weight     torch.Size([768, 768])
+encoder.modality_trunks.audio.blocks.11.attn.out_proj.bias       torch.Size([768])
+encoder.modality_trunks.audio.blocks.11.norm_1.weight    torch.Size([768])
+encoder.modality_trunks.audio.blocks.11.norm_1.bias      torch.Size([768])
+encoder.modality_trunks.audio.blocks.11.mlp.fc1.weight   torch.Size([3072, 768])
+encoder.modality_trunks.audio.blocks.11.mlp.fc1.bias     torch.Size([3072])
+encoder.modality_trunks.audio.blocks.11.mlp.fc2.weight   torch.Size([768, 3072])
+encoder.modality_trunks.audio.blocks.11.mlp.fc2.bias     torch.Size([768])
+encoder.modality_trunks.audio.blocks.11.norm_2.weight    torch.Size([768])
+encoder.modality_trunks.audio.blocks.11.norm_2.bias      torch.Size([768])
+encoder.modality_heads.vision.0.weight   torch.Size([1280])
+encoder.modality_heads.vision.0.bias     torch.Size([1280])
+encoder.modality_heads.vision.2.weight   torch.Size([1024, 1280])
 encoder.modality_heads.audio.0.weight    torch.Size([768])
+encoder.modality_heads.audio.0.bias      torch.Size([768])
 encoder.modality_heads.audio.2.weight    torch.Size([1024, 768])
-encoder.modality_heads.depth.0.weight    torch.Size([384])
-encoder.modality_heads.depth.2.weight    torch.Size([1024, 384])
-encoder.modality_heads.thermal.0.weight          torch.Size([768])
-encoder.modality_heads.thermal.2.weight          torch.Size([1024, 768])
+encoder.modality_postprocessors.audio.1.log_logit_scale          torch.Size([])
 reasoner.model.embed_tokens.weight       torch.Size([151936, 896])
 reasoner.model.layers.0.self_attn.q_proj.weight          torch.Size([896, 896])
 reasoner.model.layers.0.self_attn.q_proj.bias    torch.Size([896])