diff --git "a/Multilingual_toxic_comment_classifier/keras_metadata.pb" "b/Multilingual_toxic_comment_classifier/keras_metadata.pb" new file mode 100644--- /dev/null +++ "b/Multilingual_toxic_comment_classifier/keras_metadata.pb" @@ -0,0 +1,462 @@ + +�Oroot"_tf_keras_network*�O{"name": "model", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": false, "class_name": "Functional", "config": {"name": "model", "trainable": true, "layers": [{"class_name": "InputLayer", "config": {"batch_input_shape": {"class_name": "__tuple__", "items": [null, 192]}, "dtype": "int32", "sparse": false, "ragged": false, "name": "input_ids"}, "name": "input_ids", "inbound_nodes": []}, {"class_name": "InputLayer", "config": {"batch_input_shape": {"class_name": "__tuple__", "items": [null, 192]}, "dtype": "int32", "sparse": false, "ragged": false, "name": "attention_mask"}, "name": "attention_mask", "inbound_nodes": []}, {"class_name": "Custom>TFXLMRobertaMainLayer", "config": {"name": "roberta", "trainable": true, "dtype": "float32", "config": {"return_dict": true, "output_hidden_states": false, "output_attentions": false, "torchscript": false, "torch_dtype": null, "use_bfloat16": false, "tf_legacy_loss": false, "pruned_heads": {}, "tie_word_embeddings": true, "is_encoder_decoder": false, "is_decoder": false, "cross_attention_hidden_size": null, "add_cross_attention": false, "tie_encoder_decoder": false, "max_length": 20, "min_length": 0, "do_sample": false, "early_stopping": false, "num_beams": 1, "num_beam_groups": 1, "diversity_penalty": 0.0, "temperature": 1.0, "top_k": 50, "top_p": 1.0, "typical_p": 1.0, "repetition_penalty": 1.0, "length_penalty": 1.0, "no_repeat_ngram_size": 0, "encoder_no_repeat_ngram_size": 0, "bad_words_ids": null, "num_return_sequences": 1, "chunk_size_feed_forward": 0, "output_scores": false, "return_dict_in_generate": false, "forced_bos_token_id": null, "forced_eos_token_id": null, "remove_invalid_values": false, "exponential_decay_length_penalty": null, "suppress_tokens": null, "begin_suppress_tokens": null, "architectures": ["XLMRobertaForMaskedLM"], "finetuning_task": null, "id2label": {"0": "LABEL_0", "1": "LABEL_1"}, "label2id": {"LABEL_0": 0, "LABEL_1": 1}, "tokenizer_class": null, "prefix": null, "bos_token_id": 0, "pad_token_id": 1, "eos_token_id": 2, "sep_token_id": null, "decoder_start_token_id": null, "task_specific_params": null, "problem_type": null, "_name_or_path": "xlm-roberta-large", "transformers_version": "4.28.1", "model_type": "xlm-roberta", "output_past": true, "vocab_size": 250002, "hidden_size": 1024, "num_hidden_layers": 24, "num_attention_heads": 16, "hidden_act": "gelu", "intermediate_size": 4096, "hidden_dropout_prob": 0.1, "attention_probs_dropout_prob": 0.1, "max_position_embeddings": 514, "type_vocab_size": 1, "initializer_range": 0.02, "layer_norm_eps": 1e-05, "position_embedding_type": "absolute", "use_cache": true, "classifier_dropout": null}}, "name": "roberta", "inbound_nodes": [[["input_ids", 0, 0, {"attention_mask": ["attention_mask", 0, 0]}]]]}, {"class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "relu", "use_bias": true, "kernel_initializer": {"class_name": "GlorotUniform", "config": {"seed": null}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "name": "dense", "inbound_nodes": [[["roberta", 0, 1, {}]]]}, {"class_name": "Dense", "config": {"name": "outputs", "trainable": true, "dtype": "float32", "units": 1, "activation": "sigmoid", "use_bias": true, "kernel_initializer": {"class_name": "GlorotUniform", "config": {"seed": null}}, "bias_initializer": {"class_name": "Zeros", "config": {}}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "name": "outputs", "inbound_nodes": [[["dense", 0, 0, {}]]]}], "input_layers": [["input_ids", 0, 0], ["attention_mask", 0, 0]], "output_layers": [["outputs", 0, 0]]}, "shared_object_id": 9, "input_spec": [{"class_name": "InputSpec", "config": {"dtype": null, "shape": {"class_name": "__tuple__", "items": [null, 192]}, "ndim": 2, "max_ndim": null, "min_ndim": null, "axes": {}}}, {"class_name": "InputSpec", "config": {"dtype": null, "shape": {"class_name": "__tuple__", "items": [null, 192]}, "ndim": 2, "max_ndim": null, "min_ndim": null, "axes": {}}}], "build_input_shape": [{"class_name": "TensorShape", "items": [null, 192]}, {"class_name": "TensorShape", "items": [null, 192]}], "is_graph_network": true, "full_save_spec": {"class_name": "__tuple__", "items": [[[{"class_name": "TypeSpec", "type_spec": "tf.TensorSpec", "serialized": [{"class_name": "TensorShape", "items": [null, 192]}, "int32", "input_ids"]}, {"class_name": "TypeSpec", "type_spec": "tf.TensorSpec", "serialized": [{"class_name": "TensorShape", "items": [null, 192]}, "int32", "attention_mask"]}]], {}]}, "save_spec": [{"class_name": "TypeSpec", "type_spec": "tf.TensorSpec", "serialized": [{"class_name": "TensorShape", "items": [null, 192]}, "int32", "input_ids"]}, {"class_name": "TypeSpec", "type_spec": "tf.TensorSpec", "serialized": [{"class_name": "TensorShape", "items": [null, 192]}, "int32", "attention_mask"]}], "keras_version": "2.12.0", "backend": "tensorflow", "model_config": {"class_name": "Functional", "config": {"name": "model", "trainable": true, "layers": [{"class_name": "InputLayer", "config": {"batch_input_shape": {"class_name": "__tuple__", "items": [null, 192]}, "dtype": "int32", "sparse": false, "ragged": false, "name": "input_ids"}, "name": "input_ids", "inbound_nodes": [], "shared_object_id": 0}, {"class_name": "InputLayer", "config": {"batch_input_shape": {"class_name": "__tuple__", "items": [null, 192]}, "dtype": "int32", "sparse": false, "ragged": false, "name": "attention_mask"}, "name": "attention_mask", "inbound_nodes": [], "shared_object_id": 1}, {"class_name": "Custom>TFXLMRobertaMainLayer", "config": {"name": "roberta", "trainable": true, "dtype": "float32", "config": {"return_dict": true, "output_hidden_states": false, "output_attentions": false, "torchscript": false, "torch_dtype": null, "use_bfloat16": false, "tf_legacy_loss": false, "pruned_heads": {}, "tie_word_embeddings": true, "is_encoder_decoder": false, "is_decoder": false, "cross_attention_hidden_size": null, "add_cross_attention": false, "tie_encoder_decoder": false, "max_length": 20, "min_length": 0, "do_sample": false, "early_stopping": false, "num_beams": 1, "num_beam_groups": 1, "diversity_penalty": 0.0, "temperature": 1.0, "top_k": 50, "top_p": 1.0, "typical_p": 1.0, "repetition_penalty": 1.0, "length_penalty": 1.0, "no_repeat_ngram_size": 0, "encoder_no_repeat_ngram_size": 0, "bad_words_ids": null, "num_return_sequences": 1, "chunk_size_feed_forward": 0, "output_scores": false, "return_dict_in_generate": false, "forced_bos_token_id": null, "forced_eos_token_id": null, "remove_invalid_values": false, "exponential_decay_length_penalty": null, "suppress_tokens": null, "begin_suppress_tokens": null, "architectures": ["XLMRobertaForMaskedLM"], "finetuning_task": null, "id2label": {"0": "LABEL_0", "1": "LABEL_1"}, "label2id": {"LABEL_0": 0, "LABEL_1": 1}, "tokenizer_class": null, "prefix": null, "bos_token_id": 0, "pad_token_id": 1, "eos_token_id": 2, "sep_token_id": null, "decoder_start_token_id": null, "task_specific_params": null, "problem_type": null, "_name_or_path": "xlm-roberta-large", "transformers_version": "4.28.1", "model_type": "xlm-roberta", "output_past": true, "vocab_size": 250002, "hidden_size": 1024, "num_hidden_layers": 24, "num_attention_heads": 16, "hidden_act": "gelu", "intermediate_size": 4096, "hidden_dropout_prob": 0.1, "attention_probs_dropout_prob": 0.1, "max_position_embeddings": 514, "type_vocab_size": 1, "initializer_range": 0.02, "layer_norm_eps": 1e-05, "position_embedding_type": "absolute", "use_cache": true, "classifier_dropout": null}}, "name": "roberta", "inbound_nodes": [[["input_ids", 0, 0, {"attention_mask": ["attention_mask", 0, 0]}]]], "shared_object_id": 2}, {"class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "relu", "use_bias": true, "kernel_initializer": {"class_name": "GlorotUniform", "config": {"seed": null}, "shared_object_id": 3}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 4}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "name": "dense", "inbound_nodes": [[["roberta", 0, 1, {}]]], "shared_object_id": 5}, {"class_name": "Dense", "config": {"name": "outputs", "trainable": true, "dtype": "float32", "units": 1, "activation": "sigmoid", "use_bias": true, "kernel_initializer": {"class_name": "GlorotUniform", "config": {"seed": null}, "shared_object_id": 6}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 7}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "name": "outputs", "inbound_nodes": [[["dense", 0, 0, {}]]], "shared_object_id": 8}], "input_layers": [["input_ids", 0, 0], ["attention_mask", 0, 0]], "output_layers": [["outputs", 0, 0]]}}, "training_config": {"loss": {"class_name": "BinaryCrossentropy", "config": {"reduction": "auto", "name": "binary_crossentropy", "from_logits": false, "label_smoothing": 0.0, "axis": -1}, "shared_object_id": 12}, "metrics": [[{"class_name": "AUC", "config": {"name": "auc", "dtype": "float32", "num_thresholds": 200, "curve": "ROC", "summation_method": "interpolation", "multi_label": false, "num_labels": null, "label_weights": null, "from_logits": false}, "shared_object_id": 13}]], "weighted_metrics": null, "loss_weights": null, "optimizer_config": {"class_name": "Custom>Adam", "config": {"name": "Adam", "weight_decay": 1e-06, "clipnorm": null, "global_clipnorm": null, "clipvalue": null, "use_ema": false, "ema_momentum": 0.99, "ema_overwrite_frequency": null, "jit_compile": false, "is_legacy_optimizer": false, "learning_rate": 9.999999747378752e-06, "beta_1": 0.9, "beta_2": 0.999, "epsilon": 1e-07, "amsgrad": false}}}}2 +� root.layer-0"_tf_keras_input_layer*�{"class_name": "InputLayer", "name": "input_ids", "dtype": "int32", "sparse": false, "ragged": false, "batch_input_shape": {"class_name": "__tuple__", "items": [null, 192]}, "config": {"batch_input_shape": {"class_name": "__tuple__", "items": [null, 192]}, "dtype": "int32", "sparse": false, "ragged": false, "name": "input_ids"}}2 +� root.layer-1"_tf_keras_input_layer*�{"class_name": "InputLayer", "name": "attention_mask", "dtype": "int32", "sparse": false, "ragged": false, "batch_input_shape": {"class_name": "__tuple__", "items": [null, 192]}, "config": {"batch_input_shape": {"class_name": "__tuple__", "items": [null, 192]}, "dtype": "int32", "sparse": false, "ragged": false, "name": "attention_mask"}}2 +�root.layer_with_weights-0"_tf_keras_layer*�{"name": "roberta", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Custom>TFXLMRobertaMainLayer", "config": {"name": "roberta", "trainable": true, "dtype": "float32", "config": {"return_dict": true, "output_hidden_states": false, "output_attentions": false, "torchscript": false, "torch_dtype": null, "use_bfloat16": false, "tf_legacy_loss": false, "pruned_heads": {}, "tie_word_embeddings": true, "is_encoder_decoder": false, "is_decoder": false, "cross_attention_hidden_size": null, "add_cross_attention": false, "tie_encoder_decoder": false, "max_length": 20, "min_length": 0, "do_sample": false, "early_stopping": false, "num_beams": 1, "num_beam_groups": 1, "diversity_penalty": 0.0, "temperature": 1.0, "top_k": 50, "top_p": 1.0, "typical_p": 1.0, "repetition_penalty": 1.0, "length_penalty": 1.0, "no_repeat_ngram_size": 0, "encoder_no_repeat_ngram_size": 0, "bad_words_ids": null, "num_return_sequences": 1, "chunk_size_feed_forward": 0, "output_scores": false, "return_dict_in_generate": false, "forced_bos_token_id": null, "forced_eos_token_id": null, "remove_invalid_values": false, "exponential_decay_length_penalty": null, "suppress_tokens": null, "begin_suppress_tokens": null, "architectures": ["XLMRobertaForMaskedLM"], "finetuning_task": null, "id2label": {"0": "LABEL_0", "1": "LABEL_1"}, "label2id": {"LABEL_0": 0, "LABEL_1": 1}, "tokenizer_class": null, "prefix": null, "bos_token_id": 0, "pad_token_id": 1, "eos_token_id": 2, "sep_token_id": null, "decoder_start_token_id": null, "task_specific_params": null, "problem_type": null, "_name_or_path": "xlm-roberta-large", "transformers_version": "4.28.1", "model_type": "xlm-roberta", "output_past": true, "vocab_size": 250002, "hidden_size": 1024, "num_hidden_layers": 24, "num_attention_heads": 16, "hidden_act": "gelu", "intermediate_size": 4096, "hidden_dropout_prob": 0.1, "attention_probs_dropout_prob": 0.1, "max_position_embeddings": 514, "type_vocab_size": 1, "initializer_range": 0.02, "layer_norm_eps": 1e-05, "position_embedding_type": "absolute", "use_cache": true, "classifier_dropout": null}}, "inbound_nodes": [[["input_ids", 0, 0, {"attention_mask": ["attention_mask", 0, 0]}]]], "shared_object_id": 2, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5]}}2 +�root.layer_with_weights-1"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "relu", "use_bias": true, "kernel_initializer": {"class_name": "GlorotUniform", "config": {"seed": null}, "shared_object_id": 3}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 4}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "inbound_nodes": [[["roberta", 0, 1, {}]]], "shared_object_id": 5, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 14}, "build_input_shape": {"class_name": "TensorShape", "items": [null, 1024]}}2 +�root.layer_with_weights-2"_tf_keras_layer*�{"name": "outputs", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "outputs", "trainable": true, "dtype": "float32", "units": 1, "activation": "sigmoid", "use_bias": true, "kernel_initializer": {"class_name": "GlorotUniform", "config": {"seed": null}, "shared_object_id": 6}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 7}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "inbound_nodes": [[["dense", 0, 0, {}]]], "shared_object_id": 8, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 15}, "build_input_shape": {"class_name": "TensorShape", "items": [null, 1024]}}2 +�!root.layer_with_weights-0.encoder"_tf_keras_layer*�{"name": "encoder", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaEncoder", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +� root.layer_with_weights-0.pooler"_tf_keras_layer*�{"name": "pooler", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaPooler", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�$root.layer_with_weights-0.embeddings"_tf_keras_layer*�{"name": "embeddings", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaEmbeddings", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5]}}2 +��&root.layer_with_weights-0.pooler.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "tanh", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 16}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 17}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 18, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 19}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 1024]}}2 +��.root.layer_with_weights-0.embeddings.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 20}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 21}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 22, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��,root.layer_with_weights-0.embeddings.dropout"_tf_keras_layer*�{"name": "dropout_72", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_72", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 23, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��root.keras_api.metrics.0"_tf_keras_metric*�{"class_name": "Mean", "name": "loss", "dtype": "float32", "config": {"name": "loss", "dtype": "float32"}, "shared_object_id": 24}2 +��root.keras_api.metrics.1"_tf_keras_metric*�{"class_name": "AUC", "name": "auc", "dtype": "float32", "config": {"name": "auc", "dtype": "float32", "num_thresholds": 200, "curve": "ROC", "summation_method": "interpolation", "multi_label": false, "num_labels": null, "label_weights": null, "from_logits": false}, "shared_object_id": 13}2 +�� +)root.layer_with_weights-0.encoder.layer.0"_tf_keras_layer*�{"name": "layer_._0", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +)root.layer_with_weights-0.encoder.layer.1"_tf_keras_layer*�{"name": "layer_._1", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +)root.layer_with_weights-0.encoder.layer.2"_tf_keras_layer*�{"name": "layer_._2", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +)root.layer_with_weights-0.encoder.layer.3"_tf_keras_layer*�{"name": "layer_._3", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +)root.layer_with_weights-0.encoder.layer.4"_tf_keras_layer*�{"name": "layer_._4", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +)root.layer_with_weights-0.encoder.layer.5"_tf_keras_layer*�{"name": "layer_._5", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +)root.layer_with_weights-0.encoder.layer.6"_tf_keras_layer*�{"name": "layer_._6", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +)root.layer_with_weights-0.encoder.layer.7"_tf_keras_layer*�{"name": "layer_._7", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +)root.layer_with_weights-0.encoder.layer.8"_tf_keras_layer*�{"name": "layer_._8", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +)root.layer_with_weights-0.encoder.layer.9"_tf_keras_layer*�{"name": "layer_._9", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +*root.layer_with_weights-0.encoder.layer.10"_tf_keras_layer*�{"name": "layer_._10", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +*root.layer_with_weights-0.encoder.layer.11"_tf_keras_layer*�{"name": "layer_._11", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +*root.layer_with_weights-0.encoder.layer.12"_tf_keras_layer*�{"name": "layer_._12", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +*root.layer_with_weights-0.encoder.layer.13"_tf_keras_layer*�{"name": "layer_._13", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +*root.layer_with_weights-0.encoder.layer.14"_tf_keras_layer*�{"name": "layer_._14", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +*root.layer_with_weights-0.encoder.layer.15"_tf_keras_layer*�{"name": "layer_._15", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +*root.layer_with_weights-0.encoder.layer.16"_tf_keras_layer*�{"name": "layer_._16", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +*root.layer_with_weights-0.encoder.layer.17"_tf_keras_layer*�{"name": "layer_._17", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +*root.layer_with_weights-0.encoder.layer.18"_tf_keras_layer*�{"name": "layer_._18", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +*root.layer_with_weights-0.encoder.layer.19"_tf_keras_layer*�{"name": "layer_._19", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +*root.layer_with_weights-0.encoder.layer.20"_tf_keras_layer*�{"name": "layer_._20", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +*root.layer_with_weights-0.encoder.layer.21"_tf_keras_layer*�{"name": "layer_._21", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +*root.layer_with_weights-0.encoder.layer.22"_tf_keras_layer*�{"name": "layer_._22", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +*root.layer_with_weights-0.encoder.layer.23"_tf_keras_layer*�{"name": "layer_._23", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaLayer", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +3root.layer_with_weights-0.encoder.layer.0.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +6root.layer_with_weights-0.encoder.layer.0.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +5root.layer_with_weights-0.encoder.layer.0.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� +3root.layer_with_weights-0.encoder.layer.1.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +6root.layer_with_weights-0.encoder.layer.1.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +5root.layer_with_weights-0.encoder.layer.1.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� +3root.layer_with_weights-0.encoder.layer.2.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +6root.layer_with_weights-0.encoder.layer.2.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +5root.layer_with_weights-0.encoder.layer.2.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� +3root.layer_with_weights-0.encoder.layer.3.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +6root.layer_with_weights-0.encoder.layer.3.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +5root.layer_with_weights-0.encoder.layer.3.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� +3root.layer_with_weights-0.encoder.layer.4.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +6root.layer_with_weights-0.encoder.layer.4.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� +5root.layer_with_weights-0.encoder.layer.4.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 3root.layer_with_weights-0.encoder.layer.5.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.5.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 5root.layer_with_weights-0.encoder.layer.5.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 3root.layer_with_weights-0.encoder.layer.6.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.6.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 5root.layer_with_weights-0.encoder.layer.6.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 3root.layer_with_weights-0.encoder.layer.7.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.7.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 5root.layer_with_weights-0.encoder.layer.7.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 3root.layer_with_weights-0.encoder.layer.8.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.8.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 5root.layer_with_weights-0.encoder.layer.8.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 3root.layer_with_weights-0.encoder.layer.9.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.9.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 5root.layer_with_weights-0.encoder.layer.9.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 4root.layer_with_weights-0.encoder.layer.10.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 7root.layer_with_weights-0.encoder.layer.10.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.10.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 4root.layer_with_weights-0.encoder.layer.11.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 7root.layer_with_weights-0.encoder.layer.11.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.11.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 4root.layer_with_weights-0.encoder.layer.12.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 7root.layer_with_weights-0.encoder.layer.12.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.12.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 4root.layer_with_weights-0.encoder.layer.13.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 7root.layer_with_weights-0.encoder.layer.13.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.13.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 4root.layer_with_weights-0.encoder.layer.14.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 7root.layer_with_weights-0.encoder.layer.14.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.14.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 4root.layer_with_weights-0.encoder.layer.15.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 7root.layer_with_weights-0.encoder.layer.15.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.15.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 4root.layer_with_weights-0.encoder.layer.16.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 7root.layer_with_weights-0.encoder.layer.16.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.16.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 4root.layer_with_weights-0.encoder.layer.17.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 7root.layer_with_weights-0.encoder.layer.17.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.17.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 4root.layer_with_weights-0.encoder.layer.18.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 7root.layer_with_weights-0.encoder.layer.18.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.18.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 4root.layer_with_weights-0.encoder.layer.19.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 7root.layer_with_weights-0.encoder.layer.19.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.19.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 4root.layer_with_weights-0.encoder.layer.20.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 7root.layer_with_weights-0.encoder.layer.20.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.20.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 4root.layer_with_weights-0.encoder.layer.21.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 7root.layer_with_weights-0.encoder.layer.21.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.21.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 4root.layer_with_weights-0.encoder.layer.22.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 7root.layer_with_weights-0.encoder.layer.22.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.22.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� 4root.layer_with_weights-0.encoder.layer.23.attention"_tf_keras_layer*�{"name": "attention", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 7root.layer_with_weights-0.encoder.layer.23.intermediate"_tf_keras_layer*�{"name": "intermediate", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaIntermediate", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� 6root.layer_with_weights-0.encoder.layer.23.bert_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 4096]}}2 +�� Broot.layer_with_weights-0.encoder.layer.0.attention.self_attention"_tf_keras_layer*�{"name": "self", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� @root.layer_with_weights-0.encoder.layer.0.attention.dense_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +�� root.layer_with_weights-0.encoder.layer.10.bert_output.dropout"_tf_keras_layer*�{"name": "dropout_32", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_32", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 156, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Croot.layer_with_weights-0.encoder.layer.11.attention.self_attention"_tf_keras_layer*�{"name": "self", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Aroot.layer_with_weights-0.encoder.layer.11.attention.dense_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��=root.layer_with_weights-0.encoder.layer.11.intermediate.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 4096, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 157}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 158}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 159, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 160}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��root.layer_with_weights-0.encoder.layer.11.bert_output.dropout"_tf_keras_layer*�{"name": "dropout_35", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_35", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 168, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Croot.layer_with_weights-0.encoder.layer.12.attention.self_attention"_tf_keras_layer*�{"name": "self", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Aroot.layer_with_weights-0.encoder.layer.12.attention.dense_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��=root.layer_with_weights-0.encoder.layer.12.intermediate.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 4096, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 169}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 170}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 171, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 172}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��root.layer_with_weights-0.encoder.layer.12.bert_output.dropout"_tf_keras_layer*�{"name": "dropout_38", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_38", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 180, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Croot.layer_with_weights-0.encoder.layer.13.attention.self_attention"_tf_keras_layer*�{"name": "self", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Aroot.layer_with_weights-0.encoder.layer.13.attention.dense_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��=root.layer_with_weights-0.encoder.layer.13.intermediate.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 4096, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 181}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 182}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 183, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 184}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��root.layer_with_weights-0.encoder.layer.13.bert_output.dropout"_tf_keras_layer*�{"name": "dropout_41", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_41", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 192, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Croot.layer_with_weights-0.encoder.layer.14.attention.self_attention"_tf_keras_layer*�{"name": "self", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Aroot.layer_with_weights-0.encoder.layer.14.attention.dense_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��=root.layer_with_weights-0.encoder.layer.14.intermediate.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 4096, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 193}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 194}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 195, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 196}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��root.layer_with_weights-0.encoder.layer.14.bert_output.dropout"_tf_keras_layer*�{"name": "dropout_44", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_44", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 204, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Croot.layer_with_weights-0.encoder.layer.15.attention.self_attention"_tf_keras_layer*�{"name": "self", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Aroot.layer_with_weights-0.encoder.layer.15.attention.dense_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��=root.layer_with_weights-0.encoder.layer.15.intermediate.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 4096, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 205}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 206}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 207, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 208}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��root.layer_with_weights-0.encoder.layer.15.bert_output.dropout"_tf_keras_layer*�{"name": "dropout_47", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_47", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 216, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Croot.layer_with_weights-0.encoder.layer.16.attention.self_attention"_tf_keras_layer*�{"name": "self", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Aroot.layer_with_weights-0.encoder.layer.16.attention.dense_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��=root.layer_with_weights-0.encoder.layer.16.intermediate.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 4096, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 217}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 218}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 219, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 220}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��root.layer_with_weights-0.encoder.layer.16.bert_output.dropout"_tf_keras_layer*�{"name": "dropout_50", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_50", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 228, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Croot.layer_with_weights-0.encoder.layer.17.attention.self_attention"_tf_keras_layer*�{"name": "self", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Aroot.layer_with_weights-0.encoder.layer.17.attention.dense_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��=root.layer_with_weights-0.encoder.layer.17.intermediate.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 4096, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 229}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 230}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 231, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 232}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��root.layer_with_weights-0.encoder.layer.17.bert_output.dropout"_tf_keras_layer*�{"name": "dropout_53", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_53", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 240, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Croot.layer_with_weights-0.encoder.layer.18.attention.self_attention"_tf_keras_layer*�{"name": "self", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Aroot.layer_with_weights-0.encoder.layer.18.attention.dense_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��=root.layer_with_weights-0.encoder.layer.18.intermediate.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 4096, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 241}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 242}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 243, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 244}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��root.layer_with_weights-0.encoder.layer.18.bert_output.dropout"_tf_keras_layer*�{"name": "dropout_56", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_56", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 252, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Croot.layer_with_weights-0.encoder.layer.19.attention.self_attention"_tf_keras_layer*�{"name": "self", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Aroot.layer_with_weights-0.encoder.layer.19.attention.dense_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��=root.layer_with_weights-0.encoder.layer.19.intermediate.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 4096, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 253}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 254}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 255, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 256}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��root.layer_with_weights-0.encoder.layer.19.bert_output.dropout"_tf_keras_layer*�{"name": "dropout_59", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_59", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 264, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Croot.layer_with_weights-0.encoder.layer.20.attention.self_attention"_tf_keras_layer*�{"name": "self", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Aroot.layer_with_weights-0.encoder.layer.20.attention.dense_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��=root.layer_with_weights-0.encoder.layer.20.intermediate.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 4096, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 265}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 266}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 267, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 268}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��root.layer_with_weights-0.encoder.layer.20.bert_output.dropout"_tf_keras_layer*�{"name": "dropout_62", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_62", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 276, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Croot.layer_with_weights-0.encoder.layer.21.attention.self_attention"_tf_keras_layer*�{"name": "self", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Aroot.layer_with_weights-0.encoder.layer.21.attention.dense_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��=root.layer_with_weights-0.encoder.layer.21.intermediate.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 4096, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 277}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 278}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 279, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 280}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��root.layer_with_weights-0.encoder.layer.21.bert_output.dropout"_tf_keras_layer*�{"name": "dropout_65", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_65", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 288, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Croot.layer_with_weights-0.encoder.layer.22.attention.self_attention"_tf_keras_layer*�{"name": "self", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Aroot.layer_with_weights-0.encoder.layer.22.attention.dense_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��=root.layer_with_weights-0.encoder.layer.22.intermediate.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 4096, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 289}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 290}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 291, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 292}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��root.layer_with_weights-0.encoder.layer.22.bert_output.dropout"_tf_keras_layer*�{"name": "dropout_68", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_68", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 300, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Croot.layer_with_weights-0.encoder.layer.23.attention.self_attention"_tf_keras_layer*�{"name": "self", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfAttention", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Aroot.layer_with_weights-0.encoder.layer.23.attention.dense_output"_tf_keras_layer*�{"name": "output", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "TFXLMRobertaSelfOutput", "config": {"layer was saved without config": true}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��=root.layer_with_weights-0.encoder.layer.23.intermediate.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 4096, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 301}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 302}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 303, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 304}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��root.layer_with_weights-0.encoder.layer.23.bert_output.dropout"_tf_keras_layer*�{"name": "dropout_71", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_71", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 312, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.0.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 313}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 314}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 315, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 316}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Froot.layer_with_weights-0.encoder.layer.0.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 317}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 318}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 319, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 320}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.0.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 321}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 322}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 323, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 324}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.0.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 325, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Froot.layer_with_weights-0.encoder.layer.0.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 326}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 327}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 328, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 329}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.0.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 330}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 331}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 332, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.0.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_1", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_1", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 333, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.1.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 334}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 335}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 336, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 337}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Froot.layer_with_weights-0.encoder.layer.1.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 338}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 339}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 340, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 341}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.1.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 342}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 343}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 344, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 345}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.1.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_3", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_3", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 346, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Froot.layer_with_weights-0.encoder.layer.1.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 347}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 348}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 349, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 350}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.1.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 351}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 352}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 353, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.1.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_4", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_4", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 354, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.2.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 355}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 356}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 357, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 358}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Froot.layer_with_weights-0.encoder.layer.2.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 359}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 360}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 361, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 362}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.2.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 363}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 364}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 365, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 366}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.2.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_6", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_6", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 367, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Froot.layer_with_weights-0.encoder.layer.2.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 368}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 369}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 370, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 371}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.2.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 372}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 373}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 374, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.2.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_7", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_7", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 375, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.3.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 376}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 377}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 378, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 379}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Froot.layer_with_weights-0.encoder.layer.3.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 380}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 381}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 382, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 383}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.3.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 384}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 385}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 386, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 387}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.3.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_9", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_9", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 388, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Froot.layer_with_weights-0.encoder.layer.3.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 389}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 390}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 391, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 392}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.3.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 393}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 394}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 395, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.3.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_10", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_10", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 396, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.4.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 397}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 398}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 399, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 400}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Froot.layer_with_weights-0.encoder.layer.4.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 401}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 402}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 403, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 404}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.4.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 405}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 406}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 407, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 408}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.4.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_12", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_12", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 409, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Froot.layer_with_weights-0.encoder.layer.4.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 410}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 411}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 412, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 413}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.4.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 414}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 415}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 416, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.4.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_13", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_13", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 417, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.5.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 418}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 419}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 420, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 421}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Froot.layer_with_weights-0.encoder.layer.5.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 422}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 423}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 424, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 425}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.5.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 426}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 427}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 428, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 429}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.5.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_15", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_15", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 430, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Froot.layer_with_weights-0.encoder.layer.5.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 431}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 432}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 433, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 434}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.5.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 435}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 436}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 437, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.5.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_16", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_16", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 438, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.6.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 439}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 440}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 441, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 442}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Froot.layer_with_weights-0.encoder.layer.6.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 443}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 444}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 445, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 446}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.6.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 447}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 448}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 449, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 450}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.6.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_18", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_18", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 451, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Froot.layer_with_weights-0.encoder.layer.6.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 452}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 453}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 454, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 455}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.6.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 456}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 457}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 458, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.6.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_19", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_19", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 459, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.7.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 460}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 461}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 462, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 463}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Froot.layer_with_weights-0.encoder.layer.7.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 464}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 465}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 466, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 467}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.7.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 468}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 469}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 470, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 471}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.7.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_21", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_21", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 472, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Froot.layer_with_weights-0.encoder.layer.7.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 473}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 474}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 475, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 476}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.7.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 477}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 478}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 479, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.7.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_22", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_22", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 480, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.8.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 481}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 482}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 483, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 484}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Froot.layer_with_weights-0.encoder.layer.8.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 485}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 486}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 487, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 488}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.8.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 489}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 490}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 491, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 492}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.8.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_24", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_24", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 493, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Froot.layer_with_weights-0.encoder.layer.8.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 494}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 495}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 496, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 497}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.8.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 498}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 499}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 500, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.8.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_25", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_25", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 501, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.9.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 502}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 503}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 504, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 505}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Froot.layer_with_weights-0.encoder.layer.9.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 506}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 507}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 508, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 509}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.9.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 510}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 511}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 512, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 513}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.9.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_27", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_27", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 514, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Froot.layer_with_weights-0.encoder.layer.9.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 515}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 516}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 517, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 518}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Jroot.layer_with_weights-0.encoder.layer.9.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 519}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 520}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 521, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Hroot.layer_with_weights-0.encoder.layer.9.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_28", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_28", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 522, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.10.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 523}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 524}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 525, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 526}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Groot.layer_with_weights-0.encoder.layer.10.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 527}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 528}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 529, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 530}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.10.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 531}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 532}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 533, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 534}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.10.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_30", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_30", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 535, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Groot.layer_with_weights-0.encoder.layer.10.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 536}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 537}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 538, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 539}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.10.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 540}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 541}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 542, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.10.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_31", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_31", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 543, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.11.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 544}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 545}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 546, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 547}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Groot.layer_with_weights-0.encoder.layer.11.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 548}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 549}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 550, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 551}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.11.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 552}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 553}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 554, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 555}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.11.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_33", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_33", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 556, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Groot.layer_with_weights-0.encoder.layer.11.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 557}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 558}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 559, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 560}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.11.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 561}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 562}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 563, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.11.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_34", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_34", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 564, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.12.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 565}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 566}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 567, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 568}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Groot.layer_with_weights-0.encoder.layer.12.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 569}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 570}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 571, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 572}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.12.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 573}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 574}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 575, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 576}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.12.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_36", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_36", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 577, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Groot.layer_with_weights-0.encoder.layer.12.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 578}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 579}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 580, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 581}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.12.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 582}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 583}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 584, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.12.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_37", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_37", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 585, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.13.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 586}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 587}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 588, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 589}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Groot.layer_with_weights-0.encoder.layer.13.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 590}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 591}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 592, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 593}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.13.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 594}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 595}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 596, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 597}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.13.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_39", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_39", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 598, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Groot.layer_with_weights-0.encoder.layer.13.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 599}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 600}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 601, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 602}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.13.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 603}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 604}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 605, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.13.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_40", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_40", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 606, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.14.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 607}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 608}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 609, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 610}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Groot.layer_with_weights-0.encoder.layer.14.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 611}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 612}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 613, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 614}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.14.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 615}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 616}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 617, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 618}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.14.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_42", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_42", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 619, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Groot.layer_with_weights-0.encoder.layer.14.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 620}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 621}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 622, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 623}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.14.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 624}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 625}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 626, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.14.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_43", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_43", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 627, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.15.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 628}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 629}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 630, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 631}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Groot.layer_with_weights-0.encoder.layer.15.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 632}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 633}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 634, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 635}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.15.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 636}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 637}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 638, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 639}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.15.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_45", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_45", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 640, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Groot.layer_with_weights-0.encoder.layer.15.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 641}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 642}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 643, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 644}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.15.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 645}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 646}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 647, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.15.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_46", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_46", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 648, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.16.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 649}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 650}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 651, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 652}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Groot.layer_with_weights-0.encoder.layer.16.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 653}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 654}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 655, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 656}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.16.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 657}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 658}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 659, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 660}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.16.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_48", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_48", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 661, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Groot.layer_with_weights-0.encoder.layer.16.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 662}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 663}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 664, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 665}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.16.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 666}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 667}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 668, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.16.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_49", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_49", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 669, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.17.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 670}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 671}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 672, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 673}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Groot.layer_with_weights-0.encoder.layer.17.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 674}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 675}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 676, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 677}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.17.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 678}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 679}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 680, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 681}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.17.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_51", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_51", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 682, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Groot.layer_with_weights-0.encoder.layer.17.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 683}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 684}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 685, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 686}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.17.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 687}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 688}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 689, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.17.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_52", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_52", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 690, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.18.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 691}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 692}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 693, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 694}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Groot.layer_with_weights-0.encoder.layer.18.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 695}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 696}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 697, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 698}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.18.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 699}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 700}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 701, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 702}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.18.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_54", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_54", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 703, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Groot.layer_with_weights-0.encoder.layer.18.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 704}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 705}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 706, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 707}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.18.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 708}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 709}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 710, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.18.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_55", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_55", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 711, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.19.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 712}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 713}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 714, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 715}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Groot.layer_with_weights-0.encoder.layer.19.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 716}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 717}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 718, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 719}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.19.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 720}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 721}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 722, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 723}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.19.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_57", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_57", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 724, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Groot.layer_with_weights-0.encoder.layer.19.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 725}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 726}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 727, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 728}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.19.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 729}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 730}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 731, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.19.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_58", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_58", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 732, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.20.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 733}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 734}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 735, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 736}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Groot.layer_with_weights-0.encoder.layer.20.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 737}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 738}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 739, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 740}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.20.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 741}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 742}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 743, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 744}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.20.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_60", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_60", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 745, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Groot.layer_with_weights-0.encoder.layer.20.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 746}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 747}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 748, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 749}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.20.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 750}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 751}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 752, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.20.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_61", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_61", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 753, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.21.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 754}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 755}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 756, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 757}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Groot.layer_with_weights-0.encoder.layer.21.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 758}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 759}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 760, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 761}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.21.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 762}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 763}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 764, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 765}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.21.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_63", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_63", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 766, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Groot.layer_with_weights-0.encoder.layer.21.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 767}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 768}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 769, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 770}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.21.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 771}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 772}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 773, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.21.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_64", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_64", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 774, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.22.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 775}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 776}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 777, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 778}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Groot.layer_with_weights-0.encoder.layer.22.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 779}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 780}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 781, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 782}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.22.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 783}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 784}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 785, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 786}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.22.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_66", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_66", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 787, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Groot.layer_with_weights-0.encoder.layer.22.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 788}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 789}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 790, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 791}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.22.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 792}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 793}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 794, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.22.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_67", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_67", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 795, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.23.attention.self_attention.query"_tf_keras_layer*�{"name": "query", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "query", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 796}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 797}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 798, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 799}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Groot.layer_with_weights-0.encoder.layer.23.attention.self_attention.key"_tf_keras_layer*�{"name": "key", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "key", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 800}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 801}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 802, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 803}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.23.attention.self_attention.value"_tf_keras_layer*�{"name": "value", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "value", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 804}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 805}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 806, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 807}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.23.attention.self_attention.dropout"_tf_keras_layer*�{"name": "dropout_69", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_69", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 808, "build_input_shape": {"class_name": "TensorShape", "items": [3, 16, 5, 5]}}2 +��Groot.layer_with_weights-0.encoder.layer.23.attention.dense_output.dense"_tf_keras_layer*�{"name": "dense", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dense", "config": {"name": "dense", "trainable": true, "dtype": "float32", "units": 1024, "activation": "linear", "use_bias": true, "kernel_initializer": {"class_name": "TruncatedNormal", "config": {"mean": 0.0, "stddev": 0.02, "seed": null}, "shared_object_id": 809}, "bias_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 810}, "kernel_regularizer": null, "bias_regularizer": null, "activity_regularizer": null, "kernel_constraint": null, "bias_constraint": null}, "shared_object_id": 811, "input_spec": {"class_name": "InputSpec", "config": {"dtype": null, "shape": null, "ndim": null, "max_ndim": null, "min_ndim": 2, "axes": {"-1": 1024}}, "shared_object_id": 812}, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Kroot.layer_with_weights-0.encoder.layer.23.attention.dense_output.LayerNorm"_tf_keras_layer*�{"name": "LayerNorm", "trainable": true, "expects_training_arg": false, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "LayerNormalization", "config": {"name": "LayerNorm", "trainable": true, "dtype": "float32", "axis": [2], "epsilon": 1e-05, "center": true, "scale": true, "beta_initializer": {"class_name": "Zeros", "config": {}, "shared_object_id": 813}, "gamma_initializer": {"class_name": "Ones", "config": {}, "shared_object_id": 814}, "beta_regularizer": null, "gamma_regularizer": null, "beta_constraint": null, "gamma_constraint": null}, "shared_object_id": 815, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 +��Iroot.layer_with_weights-0.encoder.layer.23.attention.dense_output.dropout"_tf_keras_layer*�{"name": "dropout_70", "trainable": true, "expects_training_arg": true, "dtype": "float32", "batch_input_shape": null, "stateful": false, "must_restore_from_config": false, "preserve_input_structure_in_config": false, "autocast": true, "class_name": "Dropout", "config": {"name": "dropout_70", "trainable": true, "dtype": "float32", "rate": 0.1, "noise_shape": null, "seed": null}, "shared_object_id": 816, "build_input_shape": {"class_name": "TensorShape", "items": [3, 5, 1024]}}2 \ No newline at end of file