Add phi3 (#1597)

* Add phi3

* Add phi3 to init

* layer naming and some nits

* Decoder layers naming

* Remove bias from einsumdense

* nit fix for layernorm

* Add SuRotary embedding

* Romve print()

* Add conversion script

* Nit fix in script

* Add phi3_4k as default preset

* Fix Doc and nit changes

* Nit in test

* Doc fix

* Add length check for rope scaling factors

* Calculate the mean of the absolute differnce in conversion script

* Fix typo

* Add tokenizer and preprocessor

* Format fix

* Fix dtype and device in conversion script

* Batch the input

* Batch the input

* Nit

* Add notify for upload

* ADd causal_lm preprocessor

* Add causal lm

* Fix format

* small fixes

* Add phi3 to the new api

* Api gen

* Public named sublayers

* Publicc named sublayers in decoder layer

* Simplify dropout

* Fix tokenizer tests

* Fix conversion script

* use preprocessor

* use preprocessor

* Fix keras input

* Fix keras model input

* Only validate with validate_dtype

* Only validate with validate_dtype

* Change seq length

* Change text

* Set pad token id to 0

* Default stop at EOS and EOT

* Add presets

* Add presets and tests to tokenizer

* Add prepreocessor preset tests

* Add preset tests to causal_lm

* Add backbone preset tests

* Naming nits

* Clean surotaryembeddding

* Lower case file name

* Save SuScaled rope factors as python lists

* Rename orignal_max seq_length to training seq_length

* Foemat

* Remove placeholders tokens from spm

* Edit examples

* Nit in generate

* Change training_seq_length to pretraining_seq_length

* Update links

---------

Co-authored-by: Matt Watson <[email protected]>

keras_nlp/api/models/__init__.py

@@ -152,6 +152,13 @@
 )
 from keras_nlp.src.models.opt.opt_preprocessor import OPTPreprocessor
 from keras_nlp.src.models.opt.opt_tokenizer import OPTTokenizer
+from keras_nlp.src.models.phi3.phi3_backbone import Phi3Backbone
+from keras_nlp.src.models.phi3.phi3_causal_lm import Phi3CausalLM
+from keras_nlp.src.models.phi3.phi3_causal_lm_preprocessor import (
+    Phi3CausalLMPreprocessor,
+)
+from keras_nlp.src.models.phi3.phi3_preprocessor import Phi3Preprocessor
+from keras_nlp.src.models.phi3.phi3_tokenizer import Phi3Tokenizer
 from keras_nlp.src.models.preprocessor import Preprocessor
 from keras_nlp.src.models.roberta.roberta_backbone import RobertaBackbone
 from keras_nlp.src.models.roberta.roberta_classifier import RobertaClassifier

keras_nlp/src/models/__init__.py

@@ -143,6 +143,13 @@
 )
 from keras_nlp.src.models.opt.opt_preprocessor import OPTPreprocessor
 from keras_nlp.src.models.opt.opt_tokenizer import OPTTokenizer
+from keras_nlp.src.models.phi3.phi3_backbone import Phi3Backbone
+from keras_nlp.src.models.phi3.phi3_causal_lm import Phi3CausalLM
+from keras_nlp.src.models.phi3.phi3_causal_lm_preprocessor import (
+    Phi3CausalLMPreprocessor,
+)
+from keras_nlp.src.models.phi3.phi3_preprocessor import Phi3Preprocessor
+from keras_nlp.src.models.phi3.phi3_tokenizer import Phi3Tokenizer
 from keras_nlp.src.models.preprocessor import Preprocessor
 from keras_nlp.src.models.roberta.roberta_backbone import RobertaBackbone
 from keras_nlp.src.models.roberta.roberta_classifier import RobertaClassifier

keras_nlp/src/models/phi3/__init__.py

@@ -0,0 +1,20 @@
+# Copyright 2024 The KerasNLP 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_nlp.src.models.phi3.phi3_backbone import Phi3Backbone
+from keras_nlp.src.models.phi3.phi3_presets import backbone_presets
+from keras_nlp.src.models.phi3.phi3_tokenizer import Phi3Tokenizer
+from keras_nlp.src.utils.preset_utils import register_presets
+
+register_presets(backbone_presets, (Phi3Backbone, Phi3Tokenizer))

keras_nlp/src/models/phi3/phi3_attention.py

@@ -0,0 +1,259 @@
+# Copyright 2024 The KerasNLP 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_nlp.src.backend import keras
+from keras_nlp.src.backend import ops
+from keras_nlp.src.layers.modeling.rotary_embedding import RotaryEmbedding
+from keras_nlp.src.models.phi3.phi3_rotary_embedding import (
+    Phi3SuScaledRotaryEmbedding,
+)
+from keras_nlp.src.utils.keras_utils import clone_initializer
+
+
+class Phi3Attention(keras.layers.Layer):
+    """A cached grounded query attention layer."""
+
+    def __init__(
+        self,
+        num_query_heads,
+        num_key_value_heads,
+        kernel_initializer="glorot_uniform",
+        dropout=0,
+        max_sequence_length=4096,
+        pretraining_sequence_length=4096,
+        rope_max_wavelength=10000,
+        rope_scaling_type=None,
+        rope_scaling_short_factor=None,
+        rope_scaling_long_factor=None,
+        **kwargs,
+    ):
+        super().__init__(**kwargs)
+        self.num_query_heads = num_query_heads
+        self.num_key_value_heads = num_key_value_heads
+        self.num_key_value_groups = num_query_heads // num_key_value_heads
+        self.dropout = dropout
+
+        self.max_sequence_length = max_sequence_length
+        self.pretraining_sequence_length = pretraining_sequence_length
+        self.rope_max_wavelength = rope_max_wavelength
+        self.rope_scaling_type = rope_scaling_type
+        self.rope_scaling_short_factor = rope_scaling_short_factor
+        self.rope_scaling_long_factor = rope_scaling_long_factor
+
+        self.kernel_initializer = keras.initializers.get(
+            clone_initializer(kernel_initializer)
+        )
+
+    def build(self, inputs_shape):
+        # Einsum variables:
+        # b = batch size
+        # q = query length
+        # k = key/value length
+        # m = model dim
+        # u = num query heads
+        # v = num key/value heads
+        # h = head dim
+        hidden_dim = inputs_shape[-1]
+        head_dim = hidden_dim // self.num_query_heads
+        self._norm_factor = ops.sqrt(ops.cast(head_dim, self.compute_dtype))
+
+        self.query_dense = keras.layers.EinsumDense(
+            equation="bqm,muh->bquh",
+            output_shape=(None, self.num_query_heads, head_dim),
+            kernel_initializer=self.kernel_initializer,
+            dtype=self.dtype_policy,
+            name="query",
+        )
+        self.query_dense.build(inputs_shape)
+
+        self.key_dense = keras.layers.EinsumDense(
+            equation="bkm,mvh->bkvh",
+            output_shape=(
+                None,
+                self.num_key_value_heads,
+                head_dim,
+            ),
+            kernel_initializer=self.kernel_initializer,
+            dtype=self.dtype_policy,
+            name="key",
+        )
+        self.key_dense.build(inputs_shape)
+
+        self.value_dense = keras.layers.EinsumDense(
+            equation="bkm,mvh->bkvh",
+            output_shape=(
+                None,
+                self.num_key_value_heads,
+                head_dim,
+            ),
+            kernel_initializer=self.kernel_initializer,
+            dtype=self.dtype_policy,
+            name="value",
+        )
+        self.value_dense.build(inputs_shape)
+
+        self.softmax = keras.layers.Softmax(
+            axis=-1,
+            dtype="float32",
+            name="attention_softmax",
+        )
+
+        self.dropout_layer = keras.layers.Dropout(
+            rate=self.dropout,
+            dtype=self.dtype_policy,
+        )
+
+        self.output_dense = keras.layers.EinsumDense(
+            equation="bquh,uhm->bqm",
+            output_shape=(None, hidden_dim),
+            kernel_initializer=self.kernel_initializer,
+            dtype=self.dtype_policy,
+            name="attention_output",
+        )
+        self.output_dense.build((None, None, self.num_query_heads, head_dim))
+
+        if self.rope_scaling_type is None:
+            self.rotary_embedding_layer = RotaryEmbedding(
+                max_wavelength=self.rope_max_wavelength,
+                dtype=self.dtype_policy,
+            )
+        elif self.rope_scaling_type == "su":
+            if len(self.rope_scaling_short_factor) != head_dim // 2:
+                raise ValueError(
+                    "`rope_scaling_short_factor` must be of length "
+                    "`hidden_dim//num_query_heads//2`. "
+                    "`len(rope_scaling_short_factor)` is "
+                    f"{len(self.rope_scaling_short_factor)} "
+                    f"while it should be {head_dim // 2}."
+                )
+            if len(self.rope_scaling_long_factor) != head_dim // 2:
+                raise ValueError(
+                    "`rope_scaling_long_factor` must be of length "
+                    "`hidden_dim//num_query_heads//2`. "
+                    "`len(rope_scaling_long_factor)` is "
+                    f"{len(self.rope_scaling_long_factor)} "
+                    f"while it should be {head_dim // 2}."
+                )
+            self.rotary_embedding_layer = Phi3SuScaledRotaryEmbedding(
+                inverese_freq_short_factor=self.rope_scaling_short_factor,
+                inverese_freq_long_factor=self.rope_scaling_long_factor,
+                max_sequence_length=self.max_sequence_length,
+                pretraining_sequence_length=self.pretraining_sequence_length,
+                max_wavelength=self.rope_max_wavelength,
+                dtype=self.dtype_policy,
+            )
+        else:
+            raise ValueError(
+                '`rope_scaling_type` must be `None` or `"su"`.'
+                "if `None` is choosed, `RotaryEmbedding` will be used."
+                'if `"su"` is choosed, `Phi3SuScaledRotaryEmbedding` will be '
+                "used."
+            )
+
+        self.built = True
+
+    def call(
+        self,
+        hidden_states,
+        attention_mask=None,
+        cache=None,
+        cache_update_index=None,
+        training=None,
+    ):
+        start_index = (
+            cache_update_index if cache_update_index is not None else 0
+        )
+
+        query = self.query_dense(hidden_states)
+        key = self.key_dense(hidden_states)
+        value = self.value_dense(hidden_states)
+
+        # Compute RoPE for queries
+        query = self.rotary_embedding_layer(query, start_index=start_index)
+        key = self.rotary_embedding_layer(key, start_index=start_index)
+
+        if cache is not None:
+            key_cache = cache[:, 0, ...]
+            value_cache = cache[:, 1, ...]
+            if cache_update_index is None:
+                key = key_cache
+                value = value_cache
+            else:
+                start = [0, cache_update_index, 0, 0]
+                key = ops.slice_update(key_cache, start, key)
+                value = ops.slice_update(value_cache, start, value)
+                cache = ops.stack((key, value), axis=1)
+        else:
+            if cache_update_index is not None:
+                raise ValueError(
+                    "`cache_update_index` should not be set if `cache` is "
+                    f"`None`. Received: cache={cache}, "
+                    f"cache_update_index={cache_update_index}"
+                )
+
+        # [batch_shape, seq_len, num_key_value_heads, head_dim]
+        # -> [batch_shape, seq_len, num_heads, head_dim]
+        key = ops.repeat(key, repeats=self.num_key_value_groups, axis=2)
+        value = ops.repeat(value, repeats=self.num_key_value_groups, axis=2)
+
+        attention_output = self._compute_attention(
+            query, key, value, attention_mask
+        )
+
+        attention_output = self.dropout_layer(
+            attention_output, training=training
+        )
+
+        attention_output = self.output_dense(attention_output)
+
+        if cache is not None:
+            return attention_output, cache
+        return attention_output
+
+    def _masked_softmax(self, attention_scores, attention_mask=None):
+        if attention_mask is not None:
+            return self.softmax(attention_scores, attention_mask[:, None, :, :])
+        return self.softmax(attention_scores)
+
+    def _compute_attention(self, query, key, value, attention_mask=None):
+        attention_scores = ops.einsum("bquh,bkuh->buqk", query, key)
+        attention_scores = attention_scores / self._norm_factor
+        attention_scores = self._masked_softmax(
+            attention_scores, attention_mask
+        )
+        attention_scores = ops.cast(attention_scores, self.compute_dtype)
+        attention_output = ops.einsum(
+            "buqk,bkuh->bquh", attention_scores, value
+        )
+
+        return attention_output
+
+    def get_config(self):
+        config = super().get_config()
+        config.update(
+            {
+                "num_query_heads": self.num_query_heads,
+                "num_key_value_heads": self.num_key_value_heads,
+                "kernel_initializer": keras.initializers.serialize(
+                    self.kernel_initializer
+                ),
+                "dropout": self.dropout,
+                "max_sequence_length": self.max_sequence_length,
+                "pretraining_sequence_length": self.pretraining_sequence_length,
+                "rope_max_wavelength": self.rope_max_wavelength,
+                "rope_scaling_type": self.rope_scaling_type,
+                "rope_scaling_short_factor": self.rope_scaling_short_factor,
+                "rope_scaling_long_factor": self.rope_scaling_long_factor,
+            }
+        )
+        return config