[DeepSeek R1] Qwen2.5 Distillations

keras_hub/api/models/__init__.py

@@ -141,6 +141,24 @@
 from keras_hub.src.models.deeplab_v3.deeplab_v3_segmenter import (
     DeepLabV3ImageSegmenter as DeepLabV3ImageSegmenter,
 )
+from keras_hub.src.models.deepseek_r1.deepseek_r1_qwen_causal_lm import (
+    DeepSeekR1QwenCausalLM as DeepSeekR1Qwen2CausalLM,
+)
+from keras_hub.src.models.deepseek_r1.deepseek_r1_qwen_causal_lm import (
+    DeepSeekR1QwenCausalLM as DeepSeekR1QwenCausalLM,
+)
+from keras_hub.src.models.deepseek_r1.deepseek_r1_qwen_causal_lm_preprocessor import (
+    DeepSeekR1QwenCausalLMPreprocessor as DeepSeekR1Qwen2CausalLMPreprocessor,
+)
+from keras_hub.src.models.deepseek_r1.deepseek_r1_qwen_causal_lm_preprocessor import (
+    DeepSeekR1QwenCausalLMPreprocessor as DeepSeekR1QwenCausalLMPreprocessor,
+)
+from keras_hub.src.models.deepseek_r1.deepseek_r1_qwen_tokenizer import (
+    DeepSeekR1QwenTokenizer as DeepSeekR1Qwen2Tokenizer,
+)
+from keras_hub.src.models.deepseek_r1.deepseek_r1_qwen_tokenizer import (
+    DeepSeekR1QwenTokenizer as DeepSeekR1QwenTokenizer,
+)
 from keras_hub.src.models.deit.deit_backbone import DeiTBackbone as DeiTBackbone
 from keras_hub.src.models.deit.deit_image_classifier import (
     DeiTImageClassifier as DeiTImageClassifier,

keras_hub/api/tokenizers/__init__.py

@@ -22,6 +22,12 @@
 from keras_hub.src.models.deberta_v3.deberta_v3_tokenizer import (
     DebertaV3Tokenizer as DebertaV3Tokenizer,
 )
+from keras_hub.src.models.deepseek_r1.deepseek_r1_qwen_tokenizer import (
+    DeepSeekR1QwenTokenizer as DeepSeekR1Qwen2Tokenizer,
+)
+from keras_hub.src.models.deepseek_r1.deepseek_r1_qwen_tokenizer import (
+    DeepSeekR1QwenTokenizer as DeepSeekR1QwenTokenizer,
+)
 from keras_hub.src.models.distil_bert.distil_bert_tokenizer import (
     DistilBertTokenizer as DistilBertTokenizer,
 )

keras_hub/src/models/deepseek_r1/deepseek_r1_qwen_causal_lm.py

@@ -0,0 +1,300 @@
+import keras
+from keras import ops
+
+from keras_hub.src.api_export import keras_hub_export
+from keras_hub.src.models.causal_lm import CausalLM
+from keras_hub.src.models.deepseek_r1.deepseek_r1_qwen_causal_lm_preprocessor import (  # noqa: E501
+    DeepSeekR1QwenCausalLMPreprocessor,
+)
+from keras_hub.src.models.qwen.qwen_backbone import QwenBackbone
+from keras_hub.src.utils.tensor_utils import any_equal
+
+
+@keras_hub_export(
+    [
+        "keras_hub.models.DeepSeekR1QwenCausalLM",
+        "keras_hub.models.DeepSeekR1Qwen2CausalLM",
+    ]
+)
+class DeepSeekR1QwenCausalLM(CausalLM):
+    backbone_cls = QwenBackbone
+    preprocessor_cls = DeepSeekR1QwenCausalLMPreprocessor
+
+    def __init__(self, backbone, preprocessor=None, **kwargs):
+        # === Layers ===
+        self.backbone = backbone
+        self.preprocessor = preprocessor
+
+        # === Functional Model ===
+        # This must be "backbone.input" i.e. the full input structure,
+        # rather than "backbone.inputs" which is the flattened list of inputs.
+        inputs = backbone.input
+        hidden_states = backbone(inputs)
+        outputs = backbone.token_embedding(hidden_states, reverse=True)
+        super().__init__(
+            inputs=inputs,
+            outputs=outputs,
+            **kwargs,
+        )
+
+    def call_with_cache(
+        self,
+        token_ids,
+        cache,
+        cache_update_index,
+    ):
+        """Forward pass of `DeepSeekR1QwenCausalLM` with cache.
+
+        `call_with_cache` adds an additional forward pass for the model for
+        autoregressive inference. Unlike calling the model directly, this method
+        allows caching previous key/value Tensors in multi-head attention layer,
+        and avoids recomputing the outputs of seen tokens.
+
+        Args:
+            token_ids: a dense int Tensor with shape `(batch_size, max_length)`.
+            cache: a dense float Tensor, the cache of key and value.
+            cache_update_index: int, or int Tensor. The index of current inputs
+            in the whole sequence.
+
+        Returns:
+            A (logits, hidden_states, cache) tuple. Where `logits` is the
+            language model logits for the input token_ids, `hidden_states` is
+            the final hidden representation of the input tokens, and `cache` is
+            the decoding cache.
+        """
+        x = self.backbone.token_embedding(token_ids)
+        # Each decoder layer has a cache; we update them separately.
+        updated_cache = []
+        for i in range(self.backbone.num_layers):
+            current_cache = cache[:, i, ...]
+            x, next_cache = self.backbone.transformer_layers[i](
+                x,
+                self_attention_cache=current_cache,
+                self_attention_cache_update_index=cache_update_index,
+            )
+            updated_cache.append(next_cache)
+        cache = ops.stack(updated_cache, axis=1)
+        hidden_states = x = self.backbone.layer_norm(x)
+        logits = self.backbone.token_embedding(x, reverse=True)
+        return logits, hidden_states, cache
+
+    def _build_cache(self, token_ids):
+        """Build an empty cache for use with `call_with_cache()`."""
+        batch_size = ops.shape(token_ids)[0]
+        max_length = ops.shape(token_ids)[1]
+        num_layers = self.backbone.num_layers
+        num_key_value_heads = self.backbone.num_key_value_heads
+        head_dim = self.backbone.hidden_dim // self.backbone.num_query_heads
+        shape = [
+            batch_size,
+            num_layers,
+            2,
+            max_length,
+            num_key_value_heads,
+            head_dim,
+        ]
+        cache = ops.zeros(shape, dtype=self.compute_dtype)
+        # Seed the cache.
+        _, hidden_states, cache = self.call_with_cache(token_ids, cache, 0)
+        return hidden_states, cache
+
+    def generate_step(
+        self,
+        inputs,
+        stop_token_ids=None,
+    ):
+        """A compilable generation function for a single batch of inputs.
+
+        This function represents the inner, XLA-compilable, generation function
+        for a single batch of inputs. Inputs should have the same structure as
+        model inputs, a dictionary with keys `"token_ids"` and `"padding_mask"`.
+
+        Args:
+            inputs: A dictionary with two keys `"token_ids"` and
+                `"padding_mask"` and batched tensor values.
+            stop_token_ids: Tuple of id's of the end token to stop on. If all
+                sequences have produced a new stop token, generation
+                will stop.
+        """
+        token_ids, padding_mask = inputs["token_ids"], inputs["padding_mask"]
+        # Create and seed cache with a single forward pass.
+        hidden_states, cache = self._build_cache(token_ids)
+        # Compute the lengths of all user inputted tokens ids.
+        row_lengths = ops.sum(ops.cast(padding_mask, "int32"), axis=-1)
+        # Start at the first index that has no user inputted id.
+        index = ops.min(row_lengths)
+
+        def next(prompt, cache, index):
+            # The cache index is the index of our previous token.
+            cache_update_index = index - 1
+            batch_size = ops.shape(prompt)[0]
+            prompt = ops.slice(prompt, [0, cache_update_index], [batch_size, 1])
+            logits, hidden_states, cache = self.call_with_cache(
+                prompt,
+                cache,
+                cache_update_index,
+            )
+            return (
+                ops.squeeze(logits, axis=1),
+                ops.squeeze(hidden_states, axis=1),
+                cache,
+            )
+
+        token_ids = self.sampler(
+            next=next,
+            prompt=token_ids,
+            cache=cache,
+            index=index,
+            mask=padding_mask,
+            stop_token_ids=stop_token_ids,
+            hidden_states=hidden_states,
+            model=self,
+        )
+
+        # Compute an output padding mask with the token ids we updated.
+        if stop_token_ids is not None:
+            # Build a mask of stop token locations not in the original
+            # prompt (not in locations where `padding_mask` is True).
+            end_locations = any_equal(
+                token_ids, stop_token_ids, ops.logical_not(padding_mask)
+            )
+            end_locations = ops.cast(end_locations, "int32")
+            # Use cumsum to get ones in all locations after end_locations.
+            cumsum = ops.cast(ops.cumsum(end_locations, axis=-1), "int32")
+            overflow = cumsum - end_locations
+            # Our padding mask is the inverse of these overflow locations.
+            padding_mask = ops.logical_not(ops.cast(overflow, "bool"))
+        else:
+            # Without early stopping, all locations will have been updated.
+            padding_mask = ops.ones_like(token_ids, dtype="bool")
+        return {
+            "token_ids": token_ids,
+            "padding_mask": padding_mask,
+        }
+
+    def score(
+        self,
+        token_ids,
+        padding_mask=None,
+        scoring_mode="logits",
+        layer_intercept_fn=None,
+        target_ids=None,
+    ):
+        """Score a generation represented by the provided token ids.
+
+        Args:
+            token_ids: A <int>[batch_size, num_tokens] tensor containing tokens
+                to score. Typically, this tensor captures the output from a call
+                to `QwenCausalLM.generate()`, i.e., tokens for both the input
+                text and the model-generated text.
+            padding_mask: A <bool>[batch_size, num_tokens] tensor indicating the
+                tokens that should be preserved during generation. This is an
+                artifact required by the `QwenBackbone` and isn't influential
+                on the computation of this function. If omitted, this function
+                uses `keras.ops.ones()` to create a tensor of the appropriate
+                shape.
+            scoring_mode: The type of scores to return, either "logits" or
+                "loss", both will be per input token.
+            layer_intercept_fn: An optional function for augmenting activations
+                with additional computation, for example, as part of
+                interpretability research. This function will be passed the
+                activations as its first parameter and a numeric index
+                associated with that backbone layer. _This index _is not_ an
+                index into `self.backbone.layers`_. The index -1 accompanies the
+                embeddings returned by calling `self.backbone.token_embedding()`
+                on `token_ids` in the forward direction. All subsequent indexes
+                will be 0-based indices for the activations returned by each of
+                the Transformers layers in the backbone. This function must
+                return a <float>[batch_size, num_tokens, hidden_dims] tensor
+                that can be passed as an input to the next layer in the model.
+            target_ids: An <bool>[batch_size, num_tokens] tensor containing the
+                predicted tokens against which the loss should be computed. If a
+                span of tokens is provided (sequential truthy values along
+                axis=1 in the tensor), the loss will be computed as the
+                aggregate across those tokens.
+
+        Raises:
+            ValueError: If an unsupported scoring_mode is provided, or if the
+                target_ids are not provided when using ScoringMode.LOSS.
+
+        Returns:
+            The per-token scores as a tensor of size
+            <float>[batch_size, num_tokens, vocab_size] in "logits" mode, or
+            <float>[batch_size, num_tokens] in "loss" mode.
+
+        Example:
+
+        Compute gradients between embeddings and loss scores with TensorFlow:
+        ```python
+        qwen_lm = keras_hub.models.QwenCausalLM.from_preset("qwen2.5_0.5b_en")
+        generations = qwen_lm.generate(
+            ["This is a", "Where are you"],
+            max_length=30
+        )
+        preprocessed = qwen_lm.preprocessor.generate_preprocess(generations)
+        generation_ids = preprocessed["token_ids"]
+        padding_mask = preprocessed["padding_mask"]
+        target_ids = keras.ops.roll(generation_ids, shift=-1, axis=1)
+
+        embeddings = None
+        with tf.GradientTape(watch_accessed_variables=True) as tape:
+            def layer_intercept_fn(x, i):
+                if i == -1:
+                    nonlocal embeddings, tape
+                    embeddings = x
+                    tape.watch(embeddings)
+                return x
+
+            losses = qwen_lm.score(
+                token_ids=generation_ids,
+                padding_mask=padding_mask,
+                scoring_mode="loss",
+                layer_intercept_fn=layer_intercept_fn,
+                target_ids=target_ids,
+            )
+
+        grads = tape.gradient(losses, embeddings)
+        ```
+        """
+        if scoring_mode not in ("logits", "loss"):
+            raise ValueError(
+                "Unsupported scoring_mode. Must be one of 'logits' or 'loss'."
+            )
+
+        if scoring_mode == "loss" and target_ids is None:
+            raise ValueError(
+                "Cannot compute loss without targets. Please provide target "
+                "token ids via the target_ids parameter."
+            )
+
+        batch_shape = ops.shape(token_ids)[:2]
+        assert len(batch_shape) == 2
+
+        if padding_mask is None:
+            padding_mask = ops.ones(shape=batch_shape)
+
+        if layer_intercept_fn is None:
+
+            def default_layer_intercept_fn(x, unused_i):
+                return x
+
+            layer_intercept_fn = default_layer_intercept_fn
+
+        token_embeddings = self.backbone.token_embedding(token_ids)
+        x = layer_intercept_fn(token_embeddings, -1)
+
+        for i, transformer_layer in enumerate(self.backbone.transformer_layers):
+            x = transformer_layer(x, decoder_padding_mask=padding_mask)
+            x = layer_intercept_fn(x, i)
+
+        x = self.backbone.layer_norm(x)
+        logits = self.backbone.token_embedding(x, reverse=True)
+
+        if scoring_mode == "logits":
+            return logits
+
+        per_token_loss_fn = keras.losses.SparseCategoricalCrossentropy(
+            from_logits=True, reduction="none"
+        )
+        per_token_loss = per_token_loss_fn(target_ids, logits)
+        return per_token_loss

keras_hub/src/models/deepseek_r1/deepseek_r1_qwen_causal_lm_preprocessor.py

@@ -0,0 +1,20 @@
+from keras_hub.src.api_export import keras_hub_export
+from keras_hub.src.models.causal_lm_preprocessor import CausalLMPreprocessor
+from keras_hub.src.models.deepseek_r1.deepseek_r1_qwen_tokenizer import (
+    DeepSeekR1QwenTokenizer,
+)
+from keras_hub.src.models.qwen.qwen_backbone import QwenBackbone
+
+
+@keras_hub_export(
+    [
+        "keras_hub.models.DeepSeekR1QwenCausalLMPreprocessor",
+        "keras_hub.models.DeepSeekR1Qwen2CausalLMPreprocessor",
+    ]
+)
+class DeepSeekR1QwenCausalLMPreprocessor(CausalLMPreprocessor):
+    backbone_cls = QwenBackbone
+    tokenizer_cls = DeepSeekR1QwenTokenizer
+
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, **kwargs)