keras-hub-nightly 0.15.0.dev20240823171555__py3-none-any.whl

keras_hub/src/models/bloom/bloom_preprocessor.py

@@ -0,0 +1,185 @@
+# 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
+from keras_hub.src.layers.preprocessing.start_end_packer import StartEndPacker
+from keras_hub.src.models.bloom.bloom_tokenizer import BloomTokenizer
+from keras_hub.src.models.preprocessor import Preprocessor
+from keras_hub.src.utils.keras_utils import (
+    convert_inputs_to_list_of_tensor_segments,
+)
+
+
+@keras_hub_export("keras_hub.models.BloomPreprocessor")
+class BloomPreprocessor(Preprocessor):
+    """BLOOM 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"`, that can
+        be passed directly to a `keras_hub.models.BloomBackbone`.
+
+    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.
+
+    Args:
+        tokenizer: A `keras_hub.models.BloomTokenizer` 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.
+
+    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.
+
+    Examples:
+
+    Directly calling the layer on data.
+    ```python
+    preprocessor = keras_hub.models.BloomPreprocessor.from_preset(
+        "bloom_560m_multi"
+    )
+    # Tokenize and pack a single sentence.
+    preprocessor("The quick brown fox jumped.")
+
+    # Tokenize a batch of single sentences.
+    preprocessor(["The quick brown fox jumped.", "Call me Ishmael."])
+
+    # Custom vocabulary.
+    features = ["a quick fox.", "a fox quick."]
+    vocab = {"<pad>": 0, "<s>":1, "</s>":2, "a": 3, "Ġquick": 4, "Ġfox": 5}
+    merges = ["Ġ q", "u i", "c k", "ui ck", "Ġq uick"]
+    merges += ["Ġ f", "o x", "Ġf ox"]
+    tokenizer = keras_hub.models.BloomTokenizer(
+        vocabulary=vocab,
+        merges=merges,
+    )
+    preprocessor = keras_hub.models.BloomPreprocessor(tokenizer=tokenizer)
+    preprocessor("The quick brown fox jumped.")
+    ```
+
+    Mapping with `tf.data.Dataset`.
+    ```python
+    preprocessor = keras_hub.models.BloomPreprocessor.from_preset(
+        "bloom_560m_multi"
+    )
+
+    text = tf.constant(["The quick brown fox jumped.", "Call me Ishmael."])
+    label = tf.constant([1, 1])
+
+    # Map labeled single sentences.
+    ds = tf.data.Dataset.from_tensor_slices((text, label))
+    ds = ds.map(preprocessor, num_parallel_calls=tf.data.AUTOTUNE)
+
+    # Map unlabeled single sentences.
+    ds = tf.data.Dataset.from_tensor_slices(text)
+    ds = ds.map(preprocessor, num_parallel_calls=tf.data.AUTOTUNE)
+    ```
+    """
+
+    tokenizer_cls = BloomTokenizer
+
+    def __init__(
+        self,
+        tokenizer,
+        sequence_length=1024,
+        add_start_token=True,
+        add_end_token=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
+
+    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
+
+    def call(
+        self,
+        x,
+        y=None,
+        sample_weight=None,
+        sequence_length=None,
+    ):
+        x = convert_inputs_to_list_of_tensor_segments(x)
+        if len(x) != 1:
+            raise ValueError(
+                "BLOOM requires each input feature to contain only "
+                f"one segment, but received {len(x)}. If you are using BLOOM "
+                "for a multi-segment classification task, please refer to "
+                "classification models like BERT or RoBERTa."
+            )
+        sequence_length = sequence_length or self.sequence_length
+        token_ids, padding_mask = self.packer(
+            self.tokenizer(x[0]),
+            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,
+            }
+        )
+        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/bloom/bloom_presets.py

@@ -0,0 +1,121 @@
+# 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 = {
+    "bloom_560m_multi": {
+        "metadata": {
+            "description": (
+                "24-layer Bloom model with hidden dimension of 1024. "
+                "trained on 45 natural languages and 12 programming languages."
+            ),
+            "params": 559214592,
+            "official_name": "BLOOM",
+            "path": "bloom",
+            "model_card": "https://huggingface.co/bigscience/bloom-560m",
+        },
+        "kaggle_handle": "kaggle://keras/bloom/keras/bloom_560m_multi/3",
+    },
+    "bloom_1.1b_multi": {
+        "metadata": {
+            "description": (
+                "24-layer Bloom model with hidden dimension of 1536. "
+                "trained on 45 natural languages and 12 programming languages."
+            ),
+            "params": 1065314304,
+            "official_name": "BLOOM",
+            "path": "bloom",
+            "model_card": "https://huggingface.co/bigscience/bloom-1b1",
+        },
+        "kaggle_handle": "kaggle://keras/bloom/keras/bloom_1.1b_multi/1",
+    },
+    "bloom_1.7b_multi": {
+        "metadata": {
+            "description": (
+                "24-layer Bloom model with hidden dimension of 2048. "
+                "trained on 45 natural languages and 12 programming languages."
+            ),
+            "params": 1722408960,
+            "official_name": "BLOOM",
+            "path": "bloom",
+            "model_card": "https://huggingface.co/bigscience/bloom-1b7",
+        },
+        "kaggle_handle": "kaggle://keras/bloom/keras/bloom_1.7b_multi/1",
+    },
+    "bloom_3b_multi": {
+        "metadata": {
+            "description": (
+                "30-layer Bloom model with hidden dimension of 2560. "
+                "trained on 45 natural languages and 12 programming languages."
+            ),
+            "params": 3002557440,
+            "official_name": "BLOOM",
+            "path": "bloom",
+            "model_card": "https://huggingface.co/bigscience/bloom-3b",
+        },
+        "kaggle_handle": "kaggle://keras/bloom/keras/bloom_3b_multi/1",
+    },
+    "bloomz_560m_multi": {
+        "metadata": {
+            "description": (
+                "24-layer Bloom model with hidden dimension of 1024. "
+                "finetuned on crosslingual task mixture (xP3) dataset."
+            ),
+            "params": 559214592,
+            "official_name": "BLOOMZ",
+            "path": "bloom",
+            "model_card": "https://huggingface.co/bigscience/bloomz-560m",
+        },
+        "kaggle_handle": "kaggle://keras/bloom/keras/bloomz_560m_multi/1",
+    },
+    "bloomz_1.1b_multi": {
+        "metadata": {
+            "description": (
+                "24-layer Bloom model with hidden dimension of 1536. "
+                "finetuned on crosslingual task mixture (xP3) dataset."
+            ),
+            "params": 1065314304,
+            "official_name": "BLOOMZ",
+            "path": "bloom",
+            "model_card": "https://huggingface.co/bigscience/bloomz-1b1",
+        },
+        "kaggle_handle": "kaggle://keras/bloom/keras/bloomz_1.1b_multi/1",
+    },
+    "bloomz_1.7b_multi": {
+        "metadata": {
+            "description": (
+                "24-layer Bloom model with hidden dimension of 2048. "
+                "finetuned on crosslingual task mixture (xP3) dataset."
+            ),
+            "params": 1722408960,
+            "official_name": "BLOOMZ",
+            "path": "bloom",
+            "model_card": "https://huggingface.co/bigscience/bloomz-1b7",
+        },
+        "kaggle_handle": "kaggle://keras/bloom/keras/bloomz_1.7b_multi/1",
+    },
+    "bloomz_3b_multi": {
+        "metadata": {
+            "description": (
+                "30-layer Bloom model with hidden dimension of 2560. "
+                "finetuned on crosslingual task mixture (xP3) dataset."
+            ),
+            "params": 3002557440,
+            "official_name": "BLOOMZ",
+            "path": "bloom",
+            "model_card": "https://huggingface.co/bigscience/bloomz-3b",
+        },
+        "kaggle_handle": "kaggle://keras/bloom/keras/bloomz_3b_multi/1",
+    },
+}

keras_hub/src/models/bloom/bloom_tokenizer.py

@@ -0,0 +1,116 @@
+# 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.tokenizers.byte_pair_tokenizer import BytePairTokenizer
+
+
+@keras_hub_export("keras_hub.models.BloomTokenizer")
+class BloomTokenizer(BytePairTokenizer):
+    """A BLOOM tokenizer using Byte-Pair Encoding subword segmentation.
+
+    This tokenizer class will tokenize raw strings into integer sequences and
+    is based on `keras_hub.tokenizers.BytePairTokenizer`. Unlike the
+    underlying tokenizer, it will check for all special tokens needed by BLOOM
+    models and provides a `from_preset()` method to automatically download
+    a matching vocabulary for a BLOOM preset.
+
+    This tokenizer does not provide truncation or padding of inputs.
+
+    If input is a batch of strings (rank > 0), the layer will output a
+    `tf.RaggedTensor` where the last dimension of the output is ragged.
+
+    If input is a scalar string (rank == 0), the layer will output a dense
+    `tf.Tensor` with static shape `[None]`.
+
+    Args:
+        vocabulary: string or dict, maps token to integer ids. If it is a
+            string, it should be the file path to a json file.
+        merges: string or list, contains the merge rule. If it is a string,
+            it should be the file path to merge rules. The merge rule file
+            should have one merge rule per line. Every merge rule contains
+            merge entities separated by a space.
+
+    Examples:
+
+    ```python
+    # Unbatched input.
+    tokenizer = keras_hub.models.BloomTokenizer.from_preset("bloom_560m_multi")
+    tokenizer("The quick brown fox jumped.")
+
+    # Batched input.
+    tokenizer(["The quick brown fox jumped.", "The fox slept."])
+
+    # Detokenization.
+    tokenizer.detokenize(tokenizer("The quick brown fox jumped."))
+
+    # Custom vocabulary.
+    vocab = {"<s>": 0, "</s>": 1, "<pad>": 2, "a": 3, "Ġquick": 4, "Ġfox": 5}
+    merges = ["Ġ q", "u i", "c k", "ui ck", "Ġq uick"]
+    merges += ["Ġ f", "o x", "Ġf ox"]
+    tokenizer = keras_hub.models.BloomTokenizer(vocabulary=vocab, merges=merges)
+    tokenizer("a quick fox.")
+    ```
+    """
+
+    def __init__(
+        self,
+        vocabulary=None,
+        merges=None,
+        **kwargs,
+    ):
+        self.start_token = "<s>"
+        self.end_token = "</s>"
+        self.pad_token = "<pad>"
+
+        super().__init__(
+            vocabulary=vocabulary,
+            merges=merges,
+            unsplittable_tokens=[
+                self.start_token,
+                self.end_token,
+                self.pad_token,
+            ],
+            **kwargs,
+        )
+
+    def set_vocabulary_and_merges(self, vocabulary, merges):
+        super().set_vocabulary_and_merges(vocabulary, merges)
+
+        if vocabulary is not None:
+            # Check for necessary special tokens.
+            for token in [self.start_token, self.end_token, self.pad_token]:
+                if token not in self.get_vocabulary():
+                    raise ValueError(
+                        f"Cannot find token `'{token}'` in the provided "
+                        f"`vocabulary`. Please provide `'{token}'` in "
+                        "your `vocabulary` or use a pretrained `vocabulary` name."
+                    )
+
+            self.start_token_id = self.token_to_id(self.start_token)
+            self.end_token_id = self.token_to_id(self.end_token)
+            self.pad_token_id = self.token_to_id(self.pad_token)
+        else:
+            self.start_token_id = None
+            self.end_token_id = None
+            self.pad_token_id = None
+
+    def get_config(self):
+        config = super().get_config()
+        # In the constructor, we pass the list of special tokens to the
+        # `unsplittable_tokens` arg of the superclass' constructor. Hence, we
+        # delete it from the config here.
+        del config["unsplittable_tokens"]
+        return config