⚠️ 🔴 Add ministral model

docs/source/en/_toctree.yml

@@ -581,6 +581,8 @@
         title: MegatronGPT2
       - local: model_doc/minimax
         title: MiniMax
+      - local: model_doc/ministral
+        title: Ministral
       - local: model_doc/mistral
         title: Mistral
       - local: model_doc/mixtral

docs/source/en/model_doc/ministral.md

@@ -0,0 +1,86 @@
+<!--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.
+
+-->
+
+<div style="float: right;">
+    <div class="flex flex-wrap space-x-1">
+        <img alt="PyTorch" src="https://img.shields.io/badge/PyTorch-DE3412?style=flat&logo=pytorch&logoColor=white">
+        <img alt="FlashAttention" src="https://img.shields.io/badge/%E2%9A%A1%EF%B8%8E%20FlashAttention-eae0c8?style=flat">
+        <img alt="SDPA" src="https://img.shields.io/badge/SDPA-DE3412?style=flat&logo=pytorch&logoColor=white">
+        <img alt="Tensor parallelism" src="https://img.shields.io/badge/Tensor%20parallelism-06b6d4?style=flat&logoColor=white">
+    </div>
+</div>
+
+# Ministral
+
+[Ministral](https://huggingface.co/mistralai/Ministral-8B-Instruct-2410) is a 8B parameter language model that extends the Mistral architecture with alternating attention pattern. Unlike Mistral, that uses either full attention or sliding window attention consistently, Ministral alternates between full attention and sliding window attention layers, in a pattern of 1 full attention layer followed by 3 sliding window attention layers. This allows for a 128K context length support.
+
+This architecture turns out to coincide with Qwen2, with the main difference being the presence of biases in attention projections in Ministral.
+
+
+You can find the Ministral checkpoints under the [Mistral AI](https://huggingface.co/mistralai) organization.
+
+## Usage
+
+The example below demonstrates how to use Ministral for text generation:
+
+```python
+>>> import torch
+>>> from transformers import AutoModelForCausalLM, AutoTokenizer
+
+>>> model = AutoModelForCausalLM.from_pretrained("mistralai/Ministral-8B-Instruct-2410", torch_dtype=torch.bfloat16, attn_implementation="sdpa", device_map="auto")
+>>> tokenizer = AutoTokenizer.from_pretrained("mistralai/Ministral-8B-Instruct-2410")
+
+>>> messages = [
+...     {"role": "user", "content": "What is your favourite condiment?"},
+...     {"role": "assistant", "content": "Well, I'm quite partial to a good squeeze of fresh lemon juice. It adds just the right amount of zesty flavour to whatever I'm cooking up in the kitchen!"},
+...     {"role": "user", "content": "Do you have mayonnaise recipes?"}
+... ]
+
+>>> model_inputs = tokenizer.apply_chat_template(messages, return_tensors="pt").to("cuda")
+
+>>> generated_ids = model.generate(model_inputs, max_new_tokens=100, do_sample=True)
+>>> tokenizer.batch_decode(generated_ids)[0]
+"Mayonnaise can be made as follows: (...)"
+```
+
+## MinistralConfig
+
+[[autodoc]] MinistralConfig
+
+## MinistralModel
+
+[[autodoc]] MinistralModel
+    - forward
+
+## MinistralForCausalLM
+
+[[autodoc]] MinistralForCausalLM
+    - forward
+
+## MinistralForSequenceClassification
+
+[[autodoc]] MinistralForSequenceClassification
+    - forward
+
+## MinistralForTokenClassification
+
+[[autodoc]] MinistralForTokenClassification
+    - forward
+
+## MinistralForQuestionAnswering
+
+[[autodoc]] MinistralForQuestionAnswering
+- forward

src/transformers/models/__init__.py

@@ -208,6 +208,7 @@
     from .mgp_str import *
     from .mimi import *
     from .minimax import *
+    from .ministral import *
     from .mistral import *
     from .mistral3 import *
     from .mixtral import *

src/transformers/models/auto/configuration_auto.py

@@ -249,6 +249,7 @@
         ("mgp-str", "MgpstrConfig"),
         ("mimi", "MimiConfig"),
         ("minimax", "MiniMaxConfig"),
+        ("ministral", "MinistralConfig"),
         ("mistral", "MistralConfig"),
         ("mistral3", "Mistral3Config"),
         ("mixtral", "MixtralConfig"),
@@ -679,6 +680,7 @@
         ("mgp-str", "MGP-STR"),
         ("mimi", "Mimi"),
         ("minimax", "MiniMax"),
+        ("ministral", "Ministral"),
         ("mistral", "Mistral"),
         ("mistral3", "Mistral3"),
         ("mixtral", "Mixtral"),

src/transformers/models/auto/modeling_auto.py

@@ -249,6 +249,7 @@ class _BaseModelWithGenerate(PreTrainedModel, GenerationMixin):
         ("mgp-str", "MgpstrForSceneTextRecognition"),
         ("mimi", "MimiModel"),
         ("minimax", "MiniMaxModel"),
+        ("ministral", "MinistralModel"),
         ("mistral", "MistralModel"),
         ("mistral3", "Mistral3Model"),
         ("mixtral", "MixtralModel"),
@@ -682,6 +683,7 @@ class _BaseModelWithGenerate(PreTrainedModel, GenerationMixin):
         ("mega", "MegaForCausalLM"),
         ("megatron-bert", "MegatronBertForCausalLM"),
         ("minimax", "MiniMaxForCausalLM"),
+        ("ministral", "MinistralForCausalLM"),
         ("mistral", "MistralForCausalLM"),
         ("mixtral", "MixtralForCausalLM"),
         ("mllama", "MllamaForCausalLM"),
@@ -1233,6 +1235,7 @@ class _BaseModelWithGenerate(PreTrainedModel, GenerationMixin):
         ("mega", "MegaForSequenceClassification"),
         ("megatron-bert", "MegatronBertForSequenceClassification"),
         ("minimax", "MiniMaxForSequenceClassification"),
+        ("ministral", "MinistralForSequenceClassification"),
         ("mistral", "MistralForSequenceClassification"),
         ("mixtral", "MixtralForSequenceClassification"),
         ("mobilebert", "MobileBertForSequenceClassification"),
@@ -1331,6 +1334,7 @@ class _BaseModelWithGenerate(PreTrainedModel, GenerationMixin):
         ("mega", "MegaForQuestionAnswering"),
         ("megatron-bert", "MegatronBertForQuestionAnswering"),
         ("minimax", "MiniMaxForQuestionAnswering"),
+        ("ministral", "MinistralForQuestionAnswering"),
         ("mistral", "MistralForQuestionAnswering"),
         ("mixtral", "MixtralForQuestionAnswering"),
         ("mobilebert", "MobileBertForQuestionAnswering"),
@@ -1443,6 +1447,7 @@ class _BaseModelWithGenerate(PreTrainedModel, GenerationMixin):
         ("mega", "MegaForTokenClassification"),
         ("megatron-bert", "MegatronBertForTokenClassification"),
         ("minimax", "MiniMaxForTokenClassification"),
+        ("ministral", "MinistralForTokenClassification"),
         ("mistral", "MistralForTokenClassification"),
         ("mixtral", "MixtralForTokenClassification"),
         ("mobilebert", "MobileBertForTokenClassification"),

src/transformers/models/auto/tokenization_auto.py

@@ -780,7 +780,7 @@ def tokenizer_class_from_name(class_name: str) -> Union[type[Any], None]:
     for module_name, tokenizers in TOKENIZER_MAPPING_NAMES.items():
         if class_name in tokenizers:
             module_name = model_type_to_module_name(module_name)
-            if module_name in ["mistral", "mixtral"] and class_name == "MistralCommonTokenizer":
+            if module_name in ["mistral", "mixtral", "ministral"] and class_name == "MistralCommonTokenizer":
                 module = importlib.import_module(".tokenization_mistral_common", "transformers")
             else:
                 module = importlib.import_module(f".{module_name}", "transformers.models")

src/transformers/models/ministral/__init__.py

@@ -0,0 +1,27 @@
+# Copyright 2023 Mistral 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 _LazyModule
+from ...utils.import_utils import define_import_structure
+
+
+if TYPE_CHECKING:
+    from .configuration_ministral import *
+    from .modeling_ministral import *
+else:
+    import sys
+
+    _file = globals()["__file__"]
+    sys.modules[__name__] = _LazyModule(__name__, _file, define_import_structure(_file), module_spec=__spec__)

src/transformers/models/ministral/configuration_ministral.py

@@ -0,0 +1,164 @@
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+#           This file was automatically generated from src/transformers/models/ministral/modular_ministral.py.
+#               Do NOT edit this file manually as any edits will be overwritten by the generation of
+#             the file from the modular. If any change should be done, please apply the change to the
+#                          modular_ministral.py file directly. One of our CI enforces this.
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+
+from ...configuration_utils import PretrainedConfig
+
+
+class MinistralConfig(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`MinistralModel`]. It is used to instantiate an
+    Ministral 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 Ministral-8B-Instruct-2410.
+
+    [mistralai/Ministral-8B-Instruct-2410](https://huggingface.co/mistralai/Ministral-8B-Instruct-2410)
+    [mistralai/Ministral-8B-Instruct-2410](https://huggingface.co/mistralai/Ministral-8B-Instruct-2410)
+
+    Configuration objects inherit from [`PretrainedConfig`] and can be used to control the model outputs. Read the
+    documentation from [`PretrainedConfig`] for more information.
+
+
+    Args:
+        vocab_size (`int`, *optional*, defaults to 32000):
+            Vocabulary size of the Ministral model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`MinistralModel`]
+        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, check out [this
+            paper](https://huggingface.co/papers/2305.13245). If it is not specified, will default to `8`.
+        head_dim (`int`, *optional*, defaults to `hidden_size // num_attention_heads`):
+            The attention head dimension.
+        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. Ministral'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 4096):
+            Sliding window attention window size. If not specified, will default to `4096`.
+        attention_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio for the attention probabilities.
+        layer_types (`list`, *optional*):
+            Attention pattern for each layer.
+
+    ```python
+    >>> from transformers import MinistralModel, MinistralConfig
+
+    >>> # Initializing a Ministral 8B style configuration
+    >>> configuration = MinistralConfig()
+
+    >>> # Initializing a model from the Ministral 8B style configuration
+    >>> model = MinistralModel(configuration)
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "ministral"
+    keys_to_ignore_at_inference = ["past_key_values"]
+    # Default tensor parallel plan for base model `MinistralModel`
+    base_model_tp_plan = {
+        "layers.*.self_attn.q_proj": "colwise",
+        "layers.*.self_attn.k_proj": "colwise",
+        "layers.*.self_attn.v_proj": "colwise",
+        "layers.*.self_attn.o_proj": "rowwise",
+        "layers.*.mlp.gate_proj": "colwise",
+        "layers.*.mlp.up_proj": "colwise",
+        "layers.*.mlp.down_proj": "rowwise",
+    }
+    base_model_pp_plan = {
+        "embed_tokens": (["input_ids"], ["inputs_embeds"]),
+        "layers": (["hidden_states", "attention_mask"], ["hidden_states"]),
+        "norm": (["hidden_states"], ["hidden_states"]),
+    }
+
+    def __init__(
+        self,
+        vocab_size=32000,
+        hidden_size=4096,
+        intermediate_size=14336,
+        num_hidden_layers=32,
+        num_attention_heads=32,
+        num_key_value_heads=8,
+        head_dim=None,
+        hidden_act="silu",
+        max_position_embeddings=4096 * 32,
+        initializer_range=0.02,
+        rms_norm_eps=1e-6,
+        use_cache=True,
+        pad_token_id=None,
+        bos_token_id=1,
+        eos_token_id=2,
+        tie_word_embeddings=False,
+        rope_theta=10000.0,
+        sliding_window=4096,
+        attention_dropout=0.0,
+        layer_types=None,
+        **kwargs,
+    ):
+        super().__init__(
+            pad_token_id=pad_token_id,
+            bos_token_id=bos_token_id,
+            eos_token_id=eos_token_id,
+            tie_word_embeddings=tie_word_embeddings,
+            **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.head_dim = head_dim
+
+        # 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.hidden_act = hidden_act
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.attention_dropout = attention_dropout
+        self.layer_types = layer_types
+
+        if self.layer_types is None:
+            self.layer_types = [
+                "sliding_attention" if self.sliding_window is not None else "full_attention"
+            ] * num_hidden_layers
+
+
+__all__ = ["MinistralConfig"]