| { |
| "metadata": { |
| "total_size": 14524788228.0 |
| }, |
| "weight_map": { |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_vision_embedding/vars": "model_00000.weights.h5", |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_vision_embedding/patch_embedding/vars": "model_00000.weights.h5", |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_vision_embedding/position_embedding/vars": "model_00000.weights.h5", |
| "/layers/meta_clip2_vision_encoder/layers/layer_normalization/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_1/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_1/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_1/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_1/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_1/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_1/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_1/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_1/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_2/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_2/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_2/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_2/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_2/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_2/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_2/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_2/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_3/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_3/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_3/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_3/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_3/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_3/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_3/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_3/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_4/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_4/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_4/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_4/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_4/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_4/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_4/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_4/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_5/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_5/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_5/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_5/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_5/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_5/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_5/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_5/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_6/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_6/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_6/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_6/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_6/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_6/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_6/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_6/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_7/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_7/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_7/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_7/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_7/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_7/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_7/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_7/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_8/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_8/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_8/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_8/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_8/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_8/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_8/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_8/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_9/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_9/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_9/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_9/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_9/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_9/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_9/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_9/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_10/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_10/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_10/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_10/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_10/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_10/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_10/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_10/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_11/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_11/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_11/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_11/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_11/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_11/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_11/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_11/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_12/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_12/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_12/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_12/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_12/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_12/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_12/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_12/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_13/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_13/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_13/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_13/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_13/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_13/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_13/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_13/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_14/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_14/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_14/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_14/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_14/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_14/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_14/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_14/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_15/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_15/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_15/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_15/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_15/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_15/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_15/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_15/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_16/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_16/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_16/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_16/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_16/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_16/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_16/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_16/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_17/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_17/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_17/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_17/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_17/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_17/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_17/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_17/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_18/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_18/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_18/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_18/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_18/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_18/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_18/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_18/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_19/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_19/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_19/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_19/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_19/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_19/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_19/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_19/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_20/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_20/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_20/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_20/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_20/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_20/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_20/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_20/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_21/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_21/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_21/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_21/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_21/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_21/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_21/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_21/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_22/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_22/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_22/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_22/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_22/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_22/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_22/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_22/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_23/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_23/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_23/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_23/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_23/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_23/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_23/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_23/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_24/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_24/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_24/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_24/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_24/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_24/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_24/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_24/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_25/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_25/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_25/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_25/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_25/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_25/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_25/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_25/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_26/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_26/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_26/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_26/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_26/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_26/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_26/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_26/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_27/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_27/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_27/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_27/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_27/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_27/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_27/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_27/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_28/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_28/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_28/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_28/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_28/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_28/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_28/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_28/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_29/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_29/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_29/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_29/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_29/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_29/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_29/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_29/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_30/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_30/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_30/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_30/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_30/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_30/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_30/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_30/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_31/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_31/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_31/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_31/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_31/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_31/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_31/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_31/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_32/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_32/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_32/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_32/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_32/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_32/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_32/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_32/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_33/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_33/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_33/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_33/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_33/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_33/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_33/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_33/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_34/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_34/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_34/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_34/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_34/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_34/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_34/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_34/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_35/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_35/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_35/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_35/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_35/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_35/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_35/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_35/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_36/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_36/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_36/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_36/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_36/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_36/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_36/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_36/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_37/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_37/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_37/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_37/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_37/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_37/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_37/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_37/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_38/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_38/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_38/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_38/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_38/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_38/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_38/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_38/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_39/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_39/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_39/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_39/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_39/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_39/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_39/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_39/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_40/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_40/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_40/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_40/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_40/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_40/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_40/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_40/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_41/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_41/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_41/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_41/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_41/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_41/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_41/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_41/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_42/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_42/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_42/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_42/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_42/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_42/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_42/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_42/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_43/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_43/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_43/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_43/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_43/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_43/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_43/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_43/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_44/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_44/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_44/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_44/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_44/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_44/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_44/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_44/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_45/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_45/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_45/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_45/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_45/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_45/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_45/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_45/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_46/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_46/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_46/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_46/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_46/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_46/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_46/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_46/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_47/attention/key_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_47/attention/output_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_47/attention/query_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_47/attention/value_dense/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_47/dense_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_47/dense_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_47/layer_norm_1/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_vision_encoder/layers/meta_clip2_encoder_layer_47/layer_norm_2/vars": [ |
| "model_00000.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/token_and_position_embedding/position_embedding/vars": "model_00000.weights.h5", |
| "/layers/meta_clip2_text_encoder/layers/token_and_position_embedding/token_embedding/vars": "model_00001.weights.h5", |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_1/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_1/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_1/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_1/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_1/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_1/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_1/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_1/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_2/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_2/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_2/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_2/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_2/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_2/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_2/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_2/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_3/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_3/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_3/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_3/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_3/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_3/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_3/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_3/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_4/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_4/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_4/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_4/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_4/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_4/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_4/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_4/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_5/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_5/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_5/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_5/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_5/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_5/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_5/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_5/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_6/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_6/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_6/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_6/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_6/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_6/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_6/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_6/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_7/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_7/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_7/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_7/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_7/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_7/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_7/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_7/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_8/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_8/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_8/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_8/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_8/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_8/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_8/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_8/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_9/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_9/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_9/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_9/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_9/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_9/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_9/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_9/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_10/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_10/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_10/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_10/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_10/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_10/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_10/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_10/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_11/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_11/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_11/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_11/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_11/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_11/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_11/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_11/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_12/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_12/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_12/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_12/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_12/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_12/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_12/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_12/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_13/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_13/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_13/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_13/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_13/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_13/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_13/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_13/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_14/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_14/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_14/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_14/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_14/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_14/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_14/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_14/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_15/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_15/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_15/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_15/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_15/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_15/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_15/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_15/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_16/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_16/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_16/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_16/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_16/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_16/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_16/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_16/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_17/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_17/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_17/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_17/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_17/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_17/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_17/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_17/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_18/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_18/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_18/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_18/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_18/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_18/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_18/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_18/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_19/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_19/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_19/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_19/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_19/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_19/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_19/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_19/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_20/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_20/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_20/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_20/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_20/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_20/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_20/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_20/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_21/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_21/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_21/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_21/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_21/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_21/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_21/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_21/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_22/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_22/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_22/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_22/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_22/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_22/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_22/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_22/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_23/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_23/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_23/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_23/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_23/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_23/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_23/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_23/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_24/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_24/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_24/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_24/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_24/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_24/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_24/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_24/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_25/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_25/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_25/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_25/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_25/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_25/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_25/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_25/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_26/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_26/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_26/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_26/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_26/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_26/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_26/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_26/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_27/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_27/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_27/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_27/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_27/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_27/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_27/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_27/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_28/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_28/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_28/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_28/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_28/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_28/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_28/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_28/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_29/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_29/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_29/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_29/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_29/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_29/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_29/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_29/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_30/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_30/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_30/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_30/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_30/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_30/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_30/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_30/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_31/attention/key_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_31/attention/output_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_31/attention/query_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_31/attention/value_dense/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_31/dense_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_31/dense_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_31/layer_norm_1/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/meta_clip2_encoder_layer_31/layer_norm_2/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/meta_clip2_text_encoder/layers/layer_normalization/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/layer_normalization/vars": [ |
| "model_00001.weights.h5" |
| ], |
| "/layers/dense/vars": "model_00001.weights.h5", |
| "/layers/dense_1/vars": "model_00001.weights.h5", |
| "/layers/meta_clip2_head/vars": "model_00001.weights.h5" |
| } |
| } |
