keras-hub-nightly 0.16.1.dev202409250340__py3-none-any.whl → 0.16.1.dev202409270338__py3-none-any.whl

keras_hub/src/models/bert/bert_masked_lm.py

@@ -1,17 +1,3 @@
-# Copyright 2024 The KerasHub Authors
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
 import keras
 
 from keras_hub.src.api_export import keras_hub_export

keras_hub/src/models/bert/bert_masked_lm_preprocessor.py

@@ -1,17 +1,3 @@
-# Copyright 2024 The KerasHub Authors
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
 from keras_hub.src.api_export import keras_hub_export
 from keras_hub.src.models.bert.bert_backbone import BertBackbone
 from keras_hub.src.models.bert.bert_tokenizer import BertTokenizer

keras_hub/src/models/bert/bert_presets.py

@@ -1,16 +1,3 @@
-# Copyright 2024 The KerasHub Authors
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
 """BERT model preset configurations."""
 
 backbone_presets = {

keras_hub/src/models/bert/bert_text_classifier.py

@@ -1,17 +1,3 @@
-# Copyright 2024 The KerasHub Authors
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
 import keras
 
 from keras_hub.src.api_export import keras_hub_export

keras_hub/src/models/bert/bert_text_classifier_preprocessor.py

@@ -1,17 +1,3 @@
-# Copyright 2024 The KerasHub Authors
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
 from keras_hub.src.api_export import keras_hub_export
 from keras_hub.src.models.bert.bert_backbone import BertBackbone
 from keras_hub.src.models.bert.bert_tokenizer import BertTokenizer

keras_hub/src/models/bert/bert_tokenizer.py

@@ -1,17 +1,3 @@
-# Copyright 2024 The KerasHub Authors
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
 from keras_hub.src.api_export import keras_hub_export
 from keras_hub.src.models.bert.bert_backbone import BertBackbone
 from keras_hub.src.tokenizers.word_piece_tokenizer import WordPieceTokenizer

keras_hub/src/models/bloom/__init__.py

@@ -1,17 +1,3 @@
-# Copyright 2024 The KerasHub Authors
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
 from keras_hub.src.models.bloom.bloom_backbone import BloomBackbone
 from keras_hub.src.models.bloom.bloom_presets import backbone_presets
 from keras_hub.src.utils.preset_utils import register_presets

keras_hub/src/models/bloom/bloom_attention.py

@@ -1,16 +1,3 @@
-# Copyright 2024 The KerasHub Authors
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
 import math
 
 import keras

keras_hub/src/models/bloom/bloom_backbone.py

@@ -1,17 +1,3 @@
-# Copyright 2024 The KerasHub Authors
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
 import keras
 
 from keras_hub.src.api_export import keras_hub_export

keras_hub/src/models/bloom/bloom_causal_lm.py

@@ -1,18 +1,3 @@
-# Copyright 2024 The KerasHub Authors
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-
 from keras import ops
 
 from keras_hub.src.api_export import keras_hub_export

keras_hub/src/models/bloom/bloom_causal_lm_preprocessor.py

@@ -1,18 +1,3 @@
-# Copyright 2024 The KerasHub Authors
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-
 from keras_hub.src.api_export import keras_hub_export
 from keras_hub.src.models.bloom.bloom_backbone import BloomBackbone
 from keras_hub.src.models.bloom.bloom_tokenizer import BloomTokenizer

keras_hub/src/models/bloom/bloom_decoder.py

@@ -1,16 +1,3 @@
-# Copyright 2024 The KerasHub Authors
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
 # import keras
 # from keras import ops
 import keras

keras_hub/src/models/bloom/bloom_presets.py

@@ -1,16 +1,3 @@
-# Copyright 2024 The KerasHub Authors
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
 """BLOOM model preset configurations."""
 
 backbone_presets = {

keras_hub/src/models/bloom/bloom_tokenizer.py

@@ -1,18 +1,3 @@
-# Copyright 2024 The KerasHub Authors
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
-
 from keras_hub.src.api_export import keras_hub_export
 from keras_hub.src.models.bloom.bloom_backbone import BloomBackbone
 from keras_hub.src.tokenizers.byte_pair_tokenizer import BytePairTokenizer

keras_hub/src/models/causal_lm.py

@@ -1,17 +1,3 @@
-# Copyright 2024 The KerasHub Authors
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-
 import itertools
 from functools import partial
 

keras_hub/src/models/causal_lm_preprocessor.py

@@ -1,16 +1,3 @@
-# Copyright 2024 The KerasHub Authors
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
 import keras
 
 from keras_hub.src.api_export import keras_hub_export

keras_hub/src/models/{stable_diffusion_v3 → clip}/clip_encoder_block.py

@@ -1,16 +1,4 @@
-# Copyright 2024 The KerasHub Authors
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     https://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
+from keras import dtype_policies
 from keras import layers
 from keras import ops
 
@@ -43,7 +31,7 @@ class CLIPEncoderBlock(layers.Layer):
             intermediate_activation = quick_gelu
 
         self.layer_norm_1 = layers.LayerNormalization(
-            epsilon=0.00001, dtype=self.dtype_policy, name="layer_norm_1"
+            epsilon=1e-5, dtype="float32", name="layer_norm_1"
         )
         self.attention = layers.MultiHeadAttention(
             num_heads,
@@ -52,7 +40,7 @@ class CLIPEncoderBlock(layers.Layer):
             name="attention",
         )
         self.layer_norm_2 = layers.LayerNormalization(
-            epsilon=0.00001, dtype=self.dtype_policy, name="layer_norm_2"
+            epsilon=1e-5, dtype="float32", name="layer_norm_2"
         )
         self.dense_1 = layers.Dense(
             self.intermediate_dim, dtype=self.dtype_policy, name="dense_1"
@@ -67,6 +55,11 @@ class CLIPEncoderBlock(layers.Layer):
     def build(self, input_shape):
         self.layer_norm_1.build(input_shape)
         self.attention.build(input_shape, input_shape, input_shape)
+        # Before Keras 3.2, there was no setter for `dtype_policy`. Directly
+        # assign a `DTypePolicy` instead.
+        self.attention._softmax.dtype_policy = dtype_policies.DTypePolicy(
+            "float32"
+        )
         self.layer_norm_2.build(input_shape)
         self.dense_1.build(input_shape)
         input_shape = self.dense_1.compute_output_shape(input_shape)

keras_hub/src/models/clip/clip_preprocessor.py

@@ -0,0 +1,134 @@
+import keras
+
+from keras_hub.src.api_export import keras_hub_export
+from keras_hub.src.layers.preprocessing.start_end_packer import StartEndPacker
+from keras_hub.src.models.clip.clip_tokenizer import CLIPTokenizer
+from keras_hub.src.models.preprocessor import Preprocessor
+from keras_hub.src.utils.tensor_utils import preprocessing_function
+
+try:
+    import tensorflow as tf
+except ImportError:
+    tf = None
+
+
+@keras_hub_export("keras_hub.models.CLIPPreprocessor")
+class CLIPPreprocessor(Preprocessor):
+    """CLIP preprocessing layer which tokenizes and packs inputs.
+
+    This preprocessing layer will do 2 things:
+
+    - Tokenize the inputs using the `tokenizer`.
+    - Construct a dictionary with keys `"token_ids"`, `"padding_mask"`.
+
+    This layer can be used directly with `tf.data.Dataset.map` to preprocess
+    string data in the `(x, y, sample_weight)` format used by
+    `keras.Model.fit`.
+
+    The call method of this layer accepts three arguments, `x`, `y`, and
+    `sample_weight`. `x` can be a python string or tensor representing a single
+    segment, a list of python strings representing a batch of single segments,
+    or a list of tensors representing multiple segments to be packed together.
+    `y` and `sample_weight` are both optional, can have any format, and will be
+    passed through unaltered.
+
+    `CLIPPreprocessor` forces the input to have only one segment, as CLIP is
+    mainly used for generation tasks. For tasks having multi-segment inputs
+    like "glue/mnli", please use a model designed for classification purposes
+    such as BERT or RoBERTa.
+
+    Args:
+        tokenizer: A `keras_hub.models.CLIPTokenizer` instance.
+        sequence_length: The length of the packed inputs.
+        add_start_token: If `True`, the preprocessor will prepend the tokenizer
+            start token to each input sequence.
+        add_end_token: If `True`, the preprocessor will append the tokenizer
+            end token to each input sequence.
+        to_lower: bool. Whether to lower the inputs.
+
+    Call arguments:
+        x: A string, `tf.Tensor` or list of python strings.
+        y: Any label data. Will be passed through unaltered.
+        sample_weight: Any label weight data. Will be passed through unaltered.
+        sequence_length: Pass to override the configured `sequence_length` of
+            the layer.
+    """
+
+    # TODO: Add example once we have a CLIP model.
+
+    tokenizer_cls = CLIPTokenizer
+
+    def __init__(
+        self,
+        tokenizer,
+        sequence_length=77,
+        add_start_token=True,
+        add_end_token=True,
+        to_lower=True,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+        self.tokenizer = tokenizer
+        self.packer = None
+        self.sequence_length = sequence_length
+        self.add_start_token = add_start_token
+        self.add_end_token = add_end_token
+        self.to_lower = to_lower
+
+    def build(self, input_shape):
+        # Defer packer creation to `build()` so that we can be sure tokenizer
+        # assets have loaded when restoring a saved model.
+        self.packer = StartEndPacker(
+            start_value=self.tokenizer.start_token_id,
+            end_value=self.tokenizer.end_token_id,
+            pad_value=self.tokenizer.pad_token_id,
+            sequence_length=self.sequence_length,
+            return_padding_mask=True,
+        )
+        self.built = True
+
+    @preprocessing_function
+    def call(
+        self,
+        x,
+        y=None,
+        sample_weight=None,
+        sequence_length=None,
+    ):
+        sequence_length = sequence_length or self.sequence_length
+        if self.to_lower:
+            x = tf.strings.lower(x)
+        token_ids, padding_mask = self.packer(
+            self.tokenizer(x),
+            sequence_length=sequence_length,
+            add_start_value=self.add_start_token,
+            add_end_value=self.add_end_token,
+        )
+        x = {
+            "token_ids": token_ids,
+            "padding_mask": padding_mask,
+        }
+        return keras.utils.pack_x_y_sample_weight(x, y, sample_weight)
+
+    def get_config(self):
+        config = super().get_config()
+        config.update(
+            {
+                "sequence_length": self.sequence_length,
+                "add_start_token": self.add_start_token,
+                "add_end_token": self.add_end_token,
+                "to_lower": self.to_lower,
+            }
+        )
+        return config
+
+    @property
+    def sequence_length(self):
+        """The padded length of model input sequences."""
+        return self._sequence_length
+
+    @sequence_length.setter
+    def sequence_length(self, value):
+        self._sequence_length = value
+        if self.packer is not None:
+            self.packer.sequence_length = value

keras_hub/src/models/clip/clip_text_encoder.py

@@ -0,0 +1,139 @@
+from keras import layers
+
+from keras_hub.src.layers.modeling.token_and_position_embedding import (
+    TokenAndPositionEmbedding,
+)
+from keras_hub.src.models.backbone import Backbone
+from keras_hub.src.models.clip.clip_encoder_block import CLIPEncoderBlock
+
+
+class CLIPTextEncoder(Backbone):
+    """CLIP text core network with hyperparameters.
+
+    Args:
+        vocabulary_size: int. The size of the token vocabulary.
+        embedding_dim: int. The output dimension of the embedding layer.
+        hidden_dim: int. The size of the transformer hidden state at the end
+            of each transformer layer.
+        num_layers: int. The number of transformer layers.
+        num_heads: int. The number of attention heads for each transformer.
+        intermediate_dim: int. The output dimension of the first Dense layer in
+            a two-layer feedforward network for each transformer.
+        intermediate_dim: int. The output dimension of the first Dense layer in
+            a two-layer feedforward network for each transformer.
+        intermediate_activation: activation function. The activation that
+            is used for the first Dense layer in a two-layer feedforward network
+            for each transformer.
+        intermediate_output_index: optional int. The index of the intermediate
+            output. If specified, the output will become a dictionary with two
+            keys `"sequence_output"` and `"intermediate_output"`.
+        max_sequence_length: int. The maximum sequence length that this encoder
+            can consume.
+        dtype: string or `keras.mixed_precision.DTypePolicy`. The dtype to use
+            for the models computations and weights. Note that some
+            computations, such as softmax and layer normalization will always
+            be done a float32 precision regardless of dtype.
+    """
+
+    def __init__(
+        self,
+        vocabulary_size,
+        embedding_dim,
+        hidden_dim,
+        num_layers,
+        num_heads,
+        intermediate_dim,
+        intermediate_activation="quick_gelu",
+        intermediate_output_index=None,
+        max_sequence_length=77,
+        dtype=None,
+        name=None,
+        **kwargs,
+    ):
+        if (
+            intermediate_output_index is not None
+            and intermediate_output_index < 0
+        ):
+            intermediate_output_index += num_layers
+
+        # `prefix` is used to prevent duplicate name when utilizing multiple
+        # CLIP models within a single model, such as in StableDiffusion3.
+        prefix = str(name) + "_" if name is not None else ""
+
+        # === Layers ===
+        self.embedding = TokenAndPositionEmbedding(
+            vocabulary_size=vocabulary_size,
+            sequence_length=max_sequence_length,
+            embedding_dim=embedding_dim,
+            dtype=dtype,
+            name=f"{prefix}embedding",
+        )
+        self.encoder_layers = [
+            CLIPEncoderBlock(
+                hidden_dim,
+                num_heads,
+                intermediate_dim,
+                intermediate_activation,
+                dtype=dtype,
+                name=f"{prefix}encoder_block_{i}",
+            )
+            for i in range(num_layers)
+        ]
+        self.layer_norm = layers.LayerNormalization(
+            epsilon=1e-6, dtype="float32", name=f"{prefix}layer_norm"
+        )
+
+        # === Functional Model ===
+        token_id_input = layers.Input(
+            shape=(None,), dtype="int32", name="token_ids"
+        )
+        x = self.embedding(token_id_input)
+        intermediate_output = None
+        for i, block in enumerate(self.encoder_layers):
+            x = block(x)
+            if i == intermediate_output_index:
+                intermediate_output = x
+        x = self.layer_norm(x)
+        sequence_output = x
+
+        if intermediate_output_index is not None:
+            outputs = {
+                "sequence_output": sequence_output,
+                "intermediate_output": intermediate_output,
+            }
+        else:
+            outputs = sequence_output
+        super().__init__(
+            inputs={"token_ids": token_id_input},
+            outputs=outputs,
+            name=name,
+            **kwargs,
+        )
+
+        # === Config ===
+        self.vocabulary_size = vocabulary_size
+        self.max_sequence_length = max_sequence_length
+        self.embedding_dim = embedding_dim
+        self.hidden_dim = hidden_dim
+        self.num_layers = num_layers
+        self.num_heads = num_heads
+        self.intermediate_dim = intermediate_dim
+        self.intermediate_activation = intermediate_activation
+        self.intermediate_output_index = intermediate_output_index
+
+    def get_config(self):
+        config = super().get_config()
+        config.update(
+            {
+                "vocabulary_size": self.vocabulary_size,
+                "embedding_dim": self.embedding_dim,
+                "hidden_dim": self.hidden_dim,
+                "num_layers": self.num_layers,
+                "num_heads": self.num_heads,
+                "intermediate_dim": self.intermediate_dim,
+                "intermediate_activation": self.intermediate_activation,
+                "intermediate_output_index": self.intermediate_output_index,
+                "max_sequence_length": self.max_sequence_length,
+            }
+        )
+        return config