keras-hub-nightly 0.15.0.dev20240823171555__py3-none-any.whl → 0.16.0.dev2024092017__py3-none-any.whl

keras_hub/src/models/stable_diffusion_v3/mmdit.py

@@ -0,0 +1,427 @@
+# 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
+from keras import layers
+from keras import models
+from keras import ops
+
+from keras_hub.src.layers.modeling.position_embedding import PositionEmbedding
+from keras_hub.src.models.stable_diffusion_v3.mmdit_block import MMDiTBlock
+from keras_hub.src.utils.keras_utils import standardize_data_format
+
+
+class PatchEmbedding(layers.Layer):
+    def __init__(self, patch_size, hidden_dim, data_format=None, **kwargs):
+        super().__init__(**kwargs)
+        self.patch_size = int(patch_size)
+        self.hidden_dim = int(hidden_dim)
+        data_format = standardize_data_format(data_format)
+
+        self.patch_embedding = layers.Conv2D(
+            hidden_dim,
+            kernel_size=patch_size,
+            strides=patch_size,
+            data_format=data_format,
+            dtype=self.dtype_policy,
+            name="patch_embedding",
+        )
+
+    def build(self, input_shape):
+        self.patch_embedding.build(input_shape)
+
+    def call(self, inputs):
+        x = self.patch_embedding(inputs)
+        x_shape = ops.shape(x)
+        x = ops.reshape(x, (x_shape[0], x_shape[1] * x_shape[2], x_shape[3]))
+        return x
+
+    def get_config(self):
+        config = super().get_config()
+        config.update(
+            {
+                "patch_size": self.patch_size,
+                "hidden_dim": self.hidden_dim,
+            }
+        )
+        return config
+
+
+class AdjustablePositionEmbedding(PositionEmbedding):
+    def __init__(
+        self,
+        height,
+        width,
+        initializer="glorot_uniform",
+        **kwargs,
+    ):
+        height = int(height)
+        width = int(width)
+        sequence_length = height * width
+        super().__init__(sequence_length, initializer, **kwargs)
+        self.height = height
+        self.width = width
+
+    def call(self, inputs, height=None, width=None):
+        height = height or self.height
+        width = width or self.width
+        shape = ops.shape(inputs)
+        feature_length = shape[-1]
+        top = ops.floor_divide(self.height - height, 2)
+        left = ops.floor_divide(self.width - width, 2)
+        position_embedding = ops.convert_to_tensor(self.position_embeddings)
+        position_embedding = ops.reshape(
+            position_embedding, (self.height, self.width, feature_length)
+        )
+        position_embedding = ops.slice(
+            position_embedding,
+            (top, left, 0),
+            (height, width, feature_length),
+        )
+        position_embedding = ops.reshape(
+            position_embedding, (height * width, feature_length)
+        )
+        position_embedding = ops.expand_dims(position_embedding, axis=0)
+        return position_embedding
+
+    def compute_output_shape(self, input_shape):
+        return input_shape
+
+
+class TimestepEmbedding(layers.Layer):
+    def __init__(
+        self, embedding_dim, frequency_dim=256, max_period=10000, **kwargs
+    ):
+        super().__init__(**kwargs)
+        self.embedding_dim = int(embedding_dim)
+        self.frequency_dim = int(frequency_dim)
+        self.max_period = float(max_period)
+        self.half_frequency_dim = self.frequency_dim // 2
+
+        self.mlp = models.Sequential(
+            [
+                layers.Dense(
+                    embedding_dim, activation="silu", dtype=self.dtype_policy
+                ),
+                layers.Dense(
+                    embedding_dim, activation=None, dtype=self.dtype_policy
+                ),
+            ],
+            name="mlp",
+        )
+
+    def build(self, inputs_shape):
+        embedding_shape = list(inputs_shape)[1:]
+        embedding_shape.append(self.frequency_dim)
+        self.mlp.build(embedding_shape)
+
+    def _create_timestep_embedding(self, inputs):
+        compute_dtype = keras.backend.result_type(self.compute_dtype, "float32")
+        x = ops.cast(inputs, compute_dtype)
+        freqs = ops.exp(
+            ops.divide(
+                ops.multiply(
+                    -math.log(self.max_period),
+                    ops.arange(0, self.half_frequency_dim, dtype="float32"),
+                ),
+                self.half_frequency_dim,
+            )
+        )
+        freqs = ops.cast(freqs, compute_dtype)
+        x = ops.multiply(x, ops.expand_dims(freqs, axis=0))
+        embedding = ops.concatenate([ops.cos(x), ops.sin(x)], axis=-1)
+        if self.frequency_dim % 2 != 0:
+            embedding = ops.pad(embedding, [[0, 0], [0, 1]])
+        return ops.cast(embedding, self.compute_dtype)
+
+    def call(self, inputs, training=None):
+        timestep_embedding = self._create_timestep_embedding(inputs)
+        return self.mlp(timestep_embedding, training=training)
+
+    def get_config(self):
+        config = super().get_config()
+        config.update(
+            {
+                "embedding_dim": self.embedding_dim,
+                "max_period": self.max_period,
+            }
+        )
+        return config
+
+    def compute_output_shape(self, inputs_shape):
+        output_shape = list(inputs_shape)[1:]
+        output_shape.append(self.embedding_dim)
+        return output_shape
+
+
+class OutputLayer(layers.Layer):
+    def __init__(self, hidden_dim, output_dim, **kwargs):
+        super().__init__(**kwargs)
+        self.hidden_dim = hidden_dim
+        self.output_dim = output_dim
+        num_modulation = 2
+
+        self.adaptive_norm_modulation = models.Sequential(
+            [
+                layers.Activation("silu", dtype=self.dtype_policy),
+                layers.Dense(
+                    num_modulation * hidden_dim, dtype=self.dtype_policy
+                ),
+            ],
+            name="adaptive_norm_modulation",
+        )
+        self.norm = layers.LayerNormalization(
+            epsilon=1e-6,
+            center=False,
+            scale=False,
+            dtype=self.dtype_policy,
+            name="norm",
+        )
+        self.output_dense = layers.Dense(
+            output_dim,  # patch_size ** 2 * input_channels
+            use_bias=True,
+            dtype=self.dtype_policy,
+            name="output_dense",
+        )
+
+    def build(self, inputs_shape, timestep_embedding_shape):
+        self.adaptive_norm_modulation.build(timestep_embedding_shape)
+        self.norm.build(inputs_shape)
+        self.output_dense.build(inputs_shape)
+
+    def _modulate(self, inputs, shift, scale):
+        shift = ops.expand_dims(shift, axis=1)
+        scale = ops.expand_dims(scale, axis=1)
+        return ops.add(ops.multiply(inputs, ops.add(scale, 1.0)), shift)
+
+    def call(self, inputs, timestep_embedding, training=None):
+        x = inputs
+        modulation = self.adaptive_norm_modulation(
+            timestep_embedding, training=training
+        )
+        modulation = ops.reshape(modulation, (-1, 2, self.hidden_dim))
+        shift, scale = ops.unstack(modulation, 2, axis=1)
+        x = self._modulate(self.norm(x), shift, scale)
+        x = self.output_dense(x, training=training)
+        return x
+
+    def get_config(self):
+        config = super().get_config()
+        config.update(
+            {
+                "hidden_dim": self.hidden_dim,
+                "output_dim": self.output_dim,
+            }
+        )
+        return config
+
+
+class Unpatch(layers.Layer):
+    def __init__(self, patch_size, output_dim, **kwargs):
+        super().__init__(**kwargs)
+        self.patch_size = int(patch_size)
+        self.output_dim = int(output_dim)
+
+    def call(self, inputs, height, width):
+        patch_size = self.patch_size
+        output_dim = self.output_dim
+        x = ops.reshape(
+            inputs,
+            (-1, height, width, patch_size, patch_size, output_dim),
+        )
+        # (b, h, w, p1, p2, o) -> (b, h, p1, w, p2, o)
+        x = ops.transpose(x, (0, 1, 3, 2, 4, 5))
+        return ops.reshape(
+            x,
+            (-1, height * patch_size, width * patch_size, output_dim),
+        )
+
+    def get_config(self):
+        config = super().get_config()
+        config.update(
+            {
+                "patch_size": self.patch_size,
+                "output_dim": self.output_dim,
+            }
+        )
+        return config
+
+    def compute_output_shape(self, inputs_shape):
+        inputs_shape = list(inputs_shape)
+        return [inputs_shape[0], None, None, self.output_dim]
+
+
+class MMDiT(keras.Model):
+    def __init__(
+        self,
+        patch_size,
+        num_heads,
+        hidden_dim,
+        depth,
+        position_size,
+        output_dim,
+        mlp_ratio=4.0,
+        latent_shape=(64, 64, 16),
+        context_shape=(1024, 4096),
+        pooled_projection_shape=(2048,),
+        data_format=None,
+        dtype=None,
+        **kwargs,
+    ):
+        if None in latent_shape:
+            raise ValueError(
+                "`latent_shape` must be fully specified. "
+                f"Received: latent_shape={latent_shape}"
+            )
+        image_height = latent_shape[0] // patch_size
+        image_width = latent_shape[1] // patch_size
+        output_dim_in_final = patch_size**2 * output_dim
+        data_format = standardize_data_format(data_format)
+        if data_format != "channels_last":
+            raise NotImplementedError(
+                "Currently only 'channels_last' is supported."
+            )
+
+        # === Layers ===
+        self.patch_embedding = PatchEmbedding(
+            patch_size,
+            hidden_dim,
+            data_format=data_format,
+            dtype=dtype,
+            name="patch_embedding",
+        )
+        self.position_embedding_add = layers.Add(
+            dtype=dtype, name="position_embedding_add"
+        )
+        self.position_embedding = AdjustablePositionEmbedding(
+            position_size, position_size, dtype=dtype, name="position_embedding"
+        )
+        self.context_embedding = layers.Dense(
+            hidden_dim,
+            dtype=dtype,
+            name="context_embedding",
+        )
+        self.vector_embedding = models.Sequential(
+            [
+                layers.Dense(hidden_dim, activation="silu", dtype=dtype),
+                layers.Dense(hidden_dim, activation=None, dtype=dtype),
+            ],
+            name="vector_embedding",
+        )
+        self.vector_embedding_add = layers.Add(
+            dtype=dtype, name="vector_embedding_add"
+        )
+        self.timestep_embedding = TimestepEmbedding(
+            hidden_dim, dtype=dtype, name="timestep_embedding"
+        )
+        self.joint_blocks = [
+            MMDiTBlock(
+                num_heads,
+                hidden_dim,
+                mlp_ratio,
+                use_context_projection=not (i == depth - 1),
+                dtype=dtype,
+                name=f"joint_block_{i}",
+            )
+            for i in range(depth)
+        ]
+        self.output_layer = OutputLayer(
+            hidden_dim, output_dim_in_final, dtype=dtype, name="output_layer"
+        )
+        self.unpatch = Unpatch(
+            patch_size, output_dim, dtype=dtype, name="unpatch"
+        )
+
+        # === Functional Model ===
+        latent_inputs = layers.Input(shape=latent_shape, name="latent")
+        context_inputs = layers.Input(shape=context_shape, name="context")
+        pooled_projection_inputs = layers.Input(
+            shape=pooled_projection_shape, name="pooled_projection"
+        )
+        timestep_inputs = layers.Input(shape=(1,), name="timestep")
+
+        # Embeddings.
+        x = self.patch_embedding(latent_inputs)
+        position_embedding = self.position_embedding(
+            x, height=image_height, width=image_width
+        )
+        x = self.position_embedding_add([x, position_embedding])
+        context = self.context_embedding(context_inputs)
+        pooled_projection = self.vector_embedding(pooled_projection_inputs)
+        timestep_embedding = self.timestep_embedding(timestep_inputs)
+        timestep_embedding = self.vector_embedding_add(
+            [timestep_embedding, pooled_projection]
+        )
+
+        # Blocks.
+        for block in self.joint_blocks:
+            if block.use_context_projection:
+                x, context = block(x, context, timestep_embedding)
+            else:
+                x = block(x, context, timestep_embedding)
+
+        # Output layer.
+        x = self.output_layer(x, timestep_embedding)
+        outputs = self.unpatch(x, height=image_height, width=image_width)
+
+        super().__init__(
+            inputs={
+                "latent": latent_inputs,
+                "context": context_inputs,
+                "pooled_projection": pooled_projection_inputs,
+                "timestep": timestep_inputs,
+            },
+            outputs=outputs,
+            **kwargs,
+        )
+
+        # === Config ===
+        self.patch_size = patch_size
+        self.num_heads = num_heads
+        self.hidden_dim = hidden_dim
+        self.depth = depth
+        self.position_size = position_size
+        self.output_dim = output_dim
+        self.mlp_ratio = mlp_ratio
+        self.latent_shape = latent_shape
+        self.context_shape = context_shape
+        self.pooled_projection_shape = pooled_projection_shape
+
+        if dtype is not None:
+            try:
+                self.dtype_policy = keras.dtype_policies.get(dtype)
+            # Before Keras 3.2, there is no `keras.dtype_policies.get`.
+            except AttributeError:
+                if isinstance(dtype, keras.DTypePolicy):
+                    dtype = dtype.name
+                self.dtype_policy = keras.DTypePolicy(dtype)
+
+    def get_config(self):
+        config = super().get_config()
+        config.update(
+            {
+                "patch_size": self.patch_size,
+                "num_heads": self.num_heads,
+                "hidden_dim": self.hidden_dim,
+                "depth": self.depth,
+                "position_size": self.position_size,
+                "output_dim": self.output_dim,
+                "mlp_ratio": self.mlp_ratio,
+                "latent_shape": self.latent_shape,
+                "context_shape": self.context_shape,
+                "pooled_projection_shape": self.pooled_projection_shape,
+            }
+        )
+        return config