Starcoder2 model

docs/source/en/_toctree.yml

@@ -478,6 +478,8 @@
         title: Splinter
       - local: model_doc/squeezebert
         title: SqueezeBERT
+      - local: model_doc/starcoder2
+        title: Starcoder2
       - local: model_doc/stablelm
         title: StableLm
       - local: model_doc/switch_transformers

docs/source/en/model_doc/starcoder2.md

@@ -0,0 +1,53 @@
+<!--Copyright 2024 Mistral AI and The HuggingFace Team. All rights reserved.
+
+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
+
+http://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.
+
+⚠️ Note that this file is in Markdown but contain specific syntax for our doc-builder (similar to MDX) that may not be
+rendered properly in your Markdown viewer.
+
+-->
+
+# Starcoder2
+
+## Overview
+
+The Starcoder2 model was proposed in [<INSERT PAPER NAME HERE>](<INSERT PAPER LINK HERE>) by <INSERT AUTHORS HERE>.
+<INSERT SHORT SUMMARY HERE>
+
+The abstract from the paper is the following:
+
+*<INSERT PAPER ABSTRACT HERE>*
+
+Tips:
+
+<INSERT TIPS ABOUT MODEL HERE>
+
+This model was contributed by [INSERT YOUR HF USERNAME HERE](https://huggingface.co/<INSERT YOUR HF USERNAME HERE>).
+The original code can be found [here](<INSERT LINK TO GITHUB REPO HERE>).
+
+
+## Starcoder2Config
+
+[[autodoc]] Starcoder2Config
+
+## Starcoder2Model
+
+[[autodoc]] Starcoder2Model
+    - forward
+
+## Starcoder2ForCausalLM
+
+[[autodoc]] Starcoder2ForCausalLM
+    - forward
+
+## Starcoder2ForSequenceClassification
+
+[[autodoc]] Starcoder2ForSequenceClassification
+    - forward

src/transformers/__init__.py

@@ -806,6 +806,7 @@
         "SqueezeBertConfig",
         "SqueezeBertTokenizer",
     ],
+    "models.starcoder2": ["STARCODER2_PRETRAINED_CONFIG_ARCHIVE_MAP", "Starcoder2Config"],
     "models.stablelm": ["STABLELM_PRETRAINED_CONFIG_ARCHIVE_MAP", "StableLmConfig"],
     "models.swiftformer": [
         "SWIFTFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP",
@@ -3252,6 +3253,14 @@
             "SqueezeBertPreTrainedModel",
         ]
     )
+    _import_structure["models.starcoder2"].extend(
+        [
+            "Starcoder2ForCausalLM",
+            "Starcoder2ForSequenceClassification",
+            "Starcoder2Model",
+            "Starcoder2PreTrainedModel",
+        ]
+    )
     _import_structure["models.stablelm"].extend(
         [
             "StableLmForCausalLM",
@@ -5561,6 +5570,7 @@
         SqueezeBertConfig,
         SqueezeBertTokenizer,
     )
+    from .models.starcoder2 import STARCODER2_PRETRAINED_CONFIG_ARCHIVE_MAP, Starcoder2Config
     from .models.stablelm import STABLELM_PRETRAINED_CONFIG_ARCHIVE_MAP, StableLmConfig
     from .models.swiftformer import (
         SWIFTFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP,
@@ -7673,6 +7683,12 @@
             SqueezeBertModule,
             SqueezeBertPreTrainedModel,
         )
+        from .models.starcoder2 import (
+            Starcoder2ForCausalLM,
+            Starcoder2ForSequenceClassification,
+            Starcoder2Model,
+            Starcoder2PreTrainedModel,
+        )
         from .models.stablelm import (
             StableLmForCausalLM,
             StableLmForSequenceClassification,

src/transformers/models/__init__.py

@@ -202,6 +202,7 @@
     speecht5,
     splinter,
     squeezebert,
+    starcoder2,
     stablelm,
     swiftformer,
     swin,

src/transformers/models/auto/configuration_auto.py

@@ -210,6 +210,7 @@
         ("speecht5", "SpeechT5Config"),
         ("splinter", "SplinterConfig"),
         ("squeezebert", "SqueezeBertConfig"),
+        ("starcoder2", "Starcoder2Config"),
         ("stablelm", "StableLmConfig"),
         ("swiftformer", "SwiftFormerConfig"),
         ("swin", "SwinConfig"),
@@ -433,6 +434,7 @@
         ("speecht5", "SPEECHT5_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("splinter", "SPLINTER_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("squeezebert", "SQUEEZEBERT_PRETRAINED_CONFIG_ARCHIVE_MAP"),
+        ("starcoder2", "STARCODER2_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("stablelm", "STABLELM_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("swiftformer", "SWIFTFORMER_PRETRAINED_CONFIG_ARCHIVE_MAP"),
         ("swin", "SWIN_PRETRAINED_CONFIG_ARCHIVE_MAP"),
@@ -685,6 +687,7 @@
         ("speecht5", "SpeechT5"),
         ("splinter", "Splinter"),
         ("squeezebert", "SqueezeBERT"),
+        ("starcoder2", "Starcoder2"),
         ("stablelm", "StableLm"),
         ("swiftformer", "SwiftFormer"),
         ("swin", "Swin Transformer"),

src/transformers/models/auto/modeling_auto.py

@@ -200,6 +200,7 @@
         ("speecht5", "SpeechT5Model"),
         ("splinter", "SplinterModel"),
         ("squeezebert", "SqueezeBertModel"),
+        ("starcoder2", "Starcoder2Model"),
         ("stablelm", "StableLmModel"),
         ("swiftformer", "SwiftFormerModel"),
         ("swin", "SwinModel"),
@@ -461,6 +462,7 @@
         ("roformer", "RoFormerForCausalLM"),
         ("rwkv", "RwkvForCausalLM"),
         ("speech_to_text_2", "Speech2Text2ForCausalLM"),
+        ("starcoder2", "Starcoder2ForCausalLM"),
         ("stablelm", "StableLmForCausalLM"),
         ("transfo-xl", "TransfoXLLMHeadModel"),
         ("trocr", "TrOCRForCausalLM"),
@@ -808,6 +810,7 @@
         ("roc_bert", "RoCBertForSequenceClassification"),
         ("roformer", "RoFormerForSequenceClassification"),
         ("squeezebert", "SqueezeBertForSequenceClassification"),
+        ("starcoder2", "Starcoder2ForSequenceClassification"),
         ("stablelm", "StableLmForSequenceClassification"),
         ("t5", "T5ForSequenceClassification"),
         ("tapas", "TapasForSequenceClassification"),

src/transformers/models/auto/tokenization_auto.py

@@ -391,6 +391,7 @@
                 "squeezebert",
                 ("SqueezeBertTokenizer", "SqueezeBertTokenizerFast" if is_tokenizers_available() else None),
             ),
+            ("starcoder2", ("GPT2Tokenizer", "GPT2TokenizerFast" if is_tokenizers_available() else None)),
             ("stablelm", (None, "GPTNeoXTokenizerFast" if is_tokenizers_available() else None)),
             (
                 "switch_transformers",

src/transformers/models/starcoder2/__init__.py

@@ -0,0 +1,62 @@
+# Copyright 2024 Starcoder2 AI and The HuggingFace Inc. team. All rights reserved.
+#
+# 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
+#
+#     http://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 typing import TYPE_CHECKING
+
+from ...utils import (
+    OptionalDependencyNotAvailable,
+    _LazyModule,
+    is_torch_available,
+)
+
+
+_import_structure = {
+    "configuration_starcoder2": ["STARCODER2_PRETRAINED_CONFIG_ARCHIVE_MAP", "Starcoder2Config"],
+}
+
+
+try:
+    if not is_torch_available():
+        raise OptionalDependencyNotAvailable()
+except OptionalDependencyNotAvailable:
+    pass
+else:
+    _import_structure["modeling_starcoder2"] = [
+        "Starcoder2ForCausalLM",
+        "Starcoder2Model",
+        "Starcoder2PreTrainedModel",
+        "Starcoder2ForSequenceClassification",
+    ]
+
+
+if TYPE_CHECKING:
+    from .configuration_starcoder2 import STARCODER2_PRETRAINED_CONFIG_ARCHIVE_MAP, Starcoder2Config
+
+    try:
+        if not is_torch_available():
+            raise OptionalDependencyNotAvailable()
+    except OptionalDependencyNotAvailable:
+        pass
+    else:
+        from .modeling_starcoder2 import (
+            Starcoder2ForCausalLM,
+            Starcoder2ForSequenceClassification,
+            Starcoder2Model,
+            Starcoder2PreTrainedModel,
+        )
+
+
+else:
+    import sys
+
+    sys.modules[__name__] = _LazyModule(__name__, globals()["__file__"], _import_structure, module_spec=__spec__)

src/transformers/models/starcoder2/configuration_starcoder2.py

@@ -0,0 +1,165 @@
+# coding=utf-8
+# Copyright 2024 Starcoder2 AI and the HuggingFace Inc. team. All rights reserved.
+#
+# 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
+#
+#     http://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.
+""" Starcoder2 model configuration"""
+
+from ...configuration_utils import PretrainedConfig
+from ...utils import logging
+
+
+logger = logging.get_logger(__name__)
+
+STARCODER2_PRETRAINED_CONFIG_ARCHIVE_MAP = {
+    "mistralai/Mistral-7B-v0.1": "https://huggingface.co/mistralai/Mistral-7B-v0.1/resolve/main/config.json",
+}
+
+
+class Starcoder2Config(PretrainedConfig):
+    r"""
+    # TODO: Update ckpt name
+    This is the configuration class to store the configuration of a [`Starcoder2Model`]. It is used to instantiate a
+    Starcoder2 model according to the specified arguments, defining the model architecture. Instantiating a configuration
+    with the defaults will yield a similar configuration to that of the Starcoder2-7B-v0.1 or Starcoder2-7B-Instruct-v0.1.
+
+    [mistralai/Mistral-7B-v0.1](https://huggingface.co/mistralai/Mistral-7B-v0.1)
+    [starcoder2ai/Starcoder2-7B-Instruct-v0.1](https://huggingface.co/starcoder2ai/Starcoder2-7B-Instruct-v0.1)
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+
+    # TODO: Update
+    Args:
+        vocab_size (`int`, *optional*, defaults to 32000):
+            Vocabulary size of the Starcoder2 model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`Starcoder2Model`]
+        hidden_size (`int`, *optional*, defaults to 4096):
+            Dimension of the hidden representations.
+        intermediate_size (`int`, *optional*, defaults to 14336):
+            Dimension of the MLP representations.
+        num_hidden_layers (`int`, *optional*, defaults to 32):
+            Number of hidden layers in the Transformer encoder.
+        num_attention_heads (`int`, *optional*, defaults to 32):
+            Number of attention heads for each attention layer in the Transformer encoder.
+        num_key_value_heads (`int`, *optional*, defaults to 8):
+            This is the number of key_value heads that should be used to implement Grouped Query Attention. If
+            `num_key_value_heads=num_attention_heads`, the model will use Multi Head Attention (MHA), if
+            `num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When
+            converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed
+            by meanpooling all the original heads within that group. For more details checkout [this
+            paper](https://arxiv.org/pdf/2305.13245.pdf). If it is not specified, will default to `8`.
+        hidden_act (`str` or `function`, *optional*, defaults to `"silu"`):
+            The non-linear activation function (function or string) in the decoder.
+        max_position_embeddings (`int`, *optional*, defaults to `4096*32`):
+            The maximum sequence length that this model might ever be used with. Starcoder2's sliding window attention
+            allows sequence of up to 4096*32 tokens.
+        initializer_range (`float`, *optional*, defaults to 0.02):
+            The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
+        rms_norm_eps (`float`, *optional*, defaults to 1e-06):
+            The epsilon used by the rms normalization layers.
+        use_cache (`bool`, *optional*, defaults to `True`):
+            Whether or not the model should return the last key/values attentions (not used by all models). Only
+            relevant if `config.is_decoder=True`.
+        pad_token_id (`int`, *optional*):
+            The id of the padding token.
+        bos_token_id (`int`, *optional*, defaults to 1):
+            The id of the "beginning-of-sequence" token.
+        eos_token_id (`int`, *optional*, defaults to 2):
+            The id of the "end-of-sequence" token.
+        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
+            Whether the model's input and output word embeddings should be tied.
+        rope_theta (`float`, *optional*, defaults to 10000.0):
+            The base period of the RoPE embeddings.
+        sliding_window (`int`, *optional*, defaults to None):
+            Sliding window attention window size. If not specified, will default to `None` (no sliding window).
+        attention_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio for the attention probabilities.
+
+    ```python
+    >>> from transformers import Starcoder2Model, Starcoder2Config
+
+    >>> # Initializing a Starcoder2 7B style configuration
+    >>> configuration = Starcoder2Config()
+
+    >>> # Initializing a model from the Starcoder2 7B style configuration
+    >>> model = Starcoder2Model(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "starcoder2"
+    keys_to_ignore_at_inference = ["past_key_values"]
+
+    def __init__(
+        self,
+        vocab_size=49152,
+        hidden_size=3072,
+        intermediate_size=12288,
+        num_hidden_layers=30,
+        num_attention_heads=24,
+        num_key_value_heads=2,
+        # TODO: Infer from hidden_act?
+        mlp_type="default",
+        hidden_act="gelu_pytorch_tanh",
+        max_position_embeddings=4096,
+        # TODO: Check init method.
+        initializer_range=0.018042,
+        norm_type="layer_norm",
+        norm_epsilon=1e-5,
+        use_cache=True,
+        # TODO: Wrong defaults?
+        bos_token_id=50256,
+        eos_token_id=50256,
+        rope_theta=10000.0,
+        sliding_window=None,
+        attention_dropout=0.0,
+        residual_dropout=0.0,
+        embedding_dropout=0.0,
+        # TODO: Implement
+        use_bias: bool = True,
+        # TODO: Other dropouts?
+        # TODO: Allow normal pos embeddings for santacoder/starcoder1?
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.sliding_window = sliding_window
+        self.use_bias = use_bias
+
+        # for backward compatibility
+        if num_key_value_heads is None:
+            num_key_value_heads = num_attention_heads
+
+        self.num_key_value_heads = num_key_value_heads
+        self.mlp_type = mlp_type
+        self.hidden_act = hidden_act
+        self.initializer_range = initializer_range
+        self.norm_type = norm_type
+        self.norm_epsilon = norm_epsilon
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.attention_dropout = attention_dropout
+        self.residual_dropout = residual_dropout
+        self.embedding_dropout = embedding_dropout
+
+        super().__init__(
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            **kwargs,
+        )