{ "metadata": { "total_size": 4227180 }, "weight_map": { "logit_scale": "pytorch_model-00001-of-00009.bin", "text_model.embeddings.position_embedding.weight": "pytorch_model-00001-of-00009.bin", "text_model.embeddings.token_embedding.weight": "pytorch_model-00001-of-00009.bin", "text_model.encoder.layers.0.layer_norm1.bias": "pytorch_model-00009-of-00009.bin", "text_model.encoder.layers.0.layer_norm1.weight": "pytorch_model-00009-of-00009.bin", "text_model.encoder.layers.0.layer_norm2.bias": "pytorch_model-00009-of-00009.bin", "text_model.encoder.layers.0.layer_norm2.weight": "pytorch_model-00009-of-00009.bin", "text_model.encoder.layers.0.mlp.fc1.bias": "pytorch_model-00009-of-00009.bin", "text_model.encoder.layers.0.mlp.fc1.weight": "pytorch_model-00009-of-00009.bin", "text_model.encoder.layers.0.mlp.fc2.bias": "pytorch_model-00009-of-00009.bin", "text_model.encoder.layers.0.mlp.fc2.weight": "pytorch_model-00009-of-00009.bin", "text_model.encoder.layers.0.self_attn.k_proj.bias": "pytorch_model-00003-of-00009.bin", "text_model.encoder.layers.0.self_attn.k_proj.weight": "pytorch_model-00002-of-00009.bin", "text_model.encoder.layers.0.self_attn.out_proj.bias": "pytorch_model-00009-of-00009.bin", "text_model.encoder.layers.0.self_attn.out_proj.weight": "pytorch_model-00008-of-00009.bin", "text_model.encoder.layers.0.self_attn.q_proj.bias": "pytorch_model-00007-of-00009.bin", "text_model.encoder.layers.0.self_attn.q_proj.weight": "pytorch_model-00006-of-00009.bin", "text_model.encoder.layers.0.self_attn.v_proj.bias": "pytorch_model-00005-of-00009.bin", "text_model.encoder.layers.0.self_attn.v_proj.weight": "pytorch_model-00004-of-00009.bin", "text_model.final_layer_norm.bias": "pytorch_model-00009-of-00009.bin", "text_model.final_layer_norm.weight": "pytorch_model-00009-of-00009.bin", "text_projection.weight": "pytorch_model-00009-of-00009.bin", "vision_model.embeddings.class_embedding": "pytorch_model-00009-of-00009.bin", "vision_model.embeddings.patch_embedding.weight": "pytorch_model-00009-of-00009.bin", "vision_model.embeddings.position_embedding.weight": "pytorch_model-00009-of-00009.bin", "vision_model.encoder.layers.0.layer_norm1.bias": "pytorch_model-00009-of-00009.bin", "vision_model.encoder.layers.0.layer_norm1.weight": "pytorch_model-00009-of-00009.bin", "vision_model.encoder.layers.0.layer_norm2.bias": "pytorch_model-00009-of-00009.bin", "vision_model.encoder.layers.0.layer_norm2.weight": "pytorch_model-00009-of-00009.bin", "vision_model.encoder.layers.0.mlp.fc1.bias": "pytorch_model-00009-of-00009.bin", "vision_model.encoder.layers.0.mlp.fc1.weight": "pytorch_model-00009-of-00009.bin", "vision_model.encoder.layers.0.mlp.fc2.bias": "pytorch_model-00009-of-00009.bin", "vision_model.encoder.layers.0.mlp.fc2.weight": "pytorch_model-00009-of-00009.bin", "vision_model.encoder.layers.0.self_attn.k_proj.bias": "pytorch_model-00009-of-00009.bin", "vision_model.encoder.layers.0.self_attn.k_proj.weight": "pytorch_model-00009-of-00009.bin", "vision_model.encoder.layers.0.self_attn.out_proj.bias": "pytorch_model-00009-of-00009.bin", "vision_model.encoder.layers.0.self_attn.out_proj.weight": "pytorch_model-00009-of-00009.bin", "vision_model.encoder.layers.0.self_attn.q_proj.bias": "pytorch_model-00009-of-00009.bin", "vision_model.encoder.layers.0.self_attn.q_proj.weight": "pytorch_model-00009-of-00009.bin", "vision_model.encoder.layers.0.self_attn.v_proj.bias": "pytorch_model-00009-of-00009.bin", "vision_model.encoder.layers.0.self_attn.v_proj.weight": "pytorch_model-00009-of-00009.bin", "vision_model.post_layernorm.bias": "pytorch_model-00009-of-00009.bin", "vision_model.post_layernorm.weight": "pytorch_model-00009-of-00009.bin", "vision_model.pre_layrnorm.bias": "pytorch_model-00009-of-00009.bin", "vision_model.pre_layrnorm.weight": "pytorch_model-00009-of-00009.bin", "visual_projection.weight": "pytorch_model-00009-of-00009.bin" } }