diff --git a/docs/source/en/_toctree.yml b/docs/source/en/_toctree.yml
index 4b34ccf0e3e9..2c8ed28b279e 100644
--- a/docs/source/en/_toctree.yml
+++ b/docs/source/en/_toctree.yml
@@ -395,6 +395,8 @@
         title: DeBERTa
       - local: model_doc/deberta-v2
         title: DeBERTa-v2
+      - local: model_doc/deepseek_v3
+        title: DeepSeek-V3
       - local: model_doc/dialogpt
         title: DialoGPT
       - local: model_doc/diffllama
diff --git a/docs/source/en/index.md b/docs/source/en/index.md
index 16d2dd3efd27..57ea615d8215 100644
--- a/docs/source/en/index.md
+++ b/docs/source/en/index.md
@@ -119,6 +119,7 @@ Flax), PyTorch, and/or TensorFlow.
 |                       [DeBERTa](model_doc/deberta)                       |       ✅        |         ✅         |      ❌      |
 |                    [DeBERTa-v2](model_doc/deberta-v2)                    |       ✅        |         ✅         |      ❌      |
 |          [Decision Transformer](model_doc/decision_transformer)          |       ✅        |         ❌         |      ❌      |
+|                   [DeepSeek-V3](model_doc/deepseek_v3)                   |       ✅        |         ❌         |      ❌      |
 |               [Deformable DETR](model_doc/deformable_detr)               |       ✅        |         ❌         |      ❌      |
 |                          [DeiT](model_doc/deit)                          |       ✅        |         ✅         |      ❌      |
 |                        [DePlot](model_doc/deplot)                        |       ✅        |         ❌         |      ❌      |
diff --git a/docs/source/en/model_doc/deepseek_v3.md b/docs/source/en/model_doc/deepseek_v3.md
new file mode 100644
index 000000000000..e44d78739069
--- /dev/null
+++ b/docs/source/en/model_doc/deepseek_v3.md
@@ -0,0 +1,41 @@
+<!--Copyright 2025 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.
+
+-->
+
+# DeepSeek-V3
+
+## Overview
+
+The DeepSeek-V3 model was proposed in [DeepSeek-V3 Technical Report](https://arxiv.org/abs/2412.19437) by DeepSeek-AI Team.
+
+The abstract from the paper is the following:
+We present DeepSeek-V3, a strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token. To achieve efficient inference and cost-effective training, DeepSeek-V3 adopts Multi-head Latent Attention (MLA) and DeepSeekMoE architectures, which were thoroughly validated in DeepSeek-V2. Furthermore, DeepSeek-V3 pioneers an auxiliary-loss-free strategy for load balancing and sets a multi-token prediction training objective for stronger performance. We pre-train DeepSeek-V3 on 14.8 trillion diverse and high-quality tokens, followed by Supervised Fine-Tuning and Reinforcement Learning stages to fully harness its capabilities. Comprehensive evaluations reveal that DeepSeek-V3 outperforms other open-source models and achieves performance comparable to leading closed-source models. Despite its excellent performance, DeepSeek-V3 requires only 2.788M H800 GPU hours for its full training. In addition, its training process is remarkably stable. Throughout the entire training process, we did not experience any irrecoverable loss spikes or perform any rollbacks. The model checkpoints are available at https://github.com/deepseek-ai/DeepSeek-V3.
+
+### Usage tips
+The model uses Multi-head Latent Attention (MLA) and DeepSeekMoE architectures for efficient inference and cost-effective training. It employs an auxiliary-loss-free strategy for load balancing and multi-token prediction training objective. The model can be used for various language tasks after being pre-trained on 14.8 trillion tokens and going through Supervised Fine-Tuning and Reinforcement Learning stages.
+
+## DeepseekV3Config
+
+[[autodoc]] DeepseekV3Config
+
+## DeepseekV3Model
+
+[[autodoc]] DeepseekV3Model
+    - forward
+
+## DeepseekV3ForCausalLM
+
+[[autodoc]] DeepseekV3ForCausalLM
+    - forward
diff --git a/docs/source/en/perf_infer_gpu_one.md b/docs/source/en/perf_infer_gpu_one.md
index 8455c3003316..c309f3862437 100644
--- a/docs/source/en/perf_infer_gpu_one.md
+++ b/docs/source/en/perf_infer_gpu_one.md
@@ -47,6 +47,7 @@ FlashAttention-2 is currently supported for the following architectures:
 * [Cohere2](https://huggingface.co/docs/transformers/model_doc/cohere2#transformers.Cohere2Model)
 * [GLM](https://huggingface.co/docs/transformers/model_doc/glm#transformers.GLMModel)
 * [Dbrx](https://huggingface.co/docs/transformers/model_doc/dbrx#transformers.DbrxModel)
+* [DeepseekV3](https://huggingface.co/docs/transformers/model_doc/deepseek_v3#transformers.DeepseekV3Model)
 * [DiffLlama](https://huggingface.co/docs/transformers/model_doc/diffllama#transformers.DiffLlamaModel)
 * [DistilBert](https://huggingface.co/docs/transformers/model_doc/distilbert#transformers.DistilBertModel)
 * [Emu3](https://huggingface.co/docs/transformers/model_doc/emu3)
@@ -246,6 +247,7 @@ For now, Transformers supports SDPA inference and training for the following arc
 * [Dbrx](https://huggingface.co/docs/transformers/model_doc/dbrx#transformers.DbrxModel)
 * [DeiT](https://huggingface.co/docs/transformers/model_doc/deit#transformers.DeiTModel)
 * [DepthPro](https://huggingface.co/docs/transformers/model_doc/depth_pro#transformers.DepthProModel)
+* [DeepseekV3](https://huggingface.co/docs/transformers/model_doc/deepseek_v3#transformers.DeepseekV3Model)
 * [DiffLlama](https://huggingface.co/docs/transformers/model_doc/diffllama#transformers.DiffLlamaModel)
 * [Dinov2](https://huggingface.co/docs/transformers/en/model_doc/dinov2)
 * [Dinov2_with_registers](https://huggingface.co/docs/transformers/en/model_doc/dinov2)
diff --git a/src/transformers/__init__.py b/src/transformers/__init__.py
index f9dcee68c2eb..3c1cec5cb185 100755
--- a/src/transformers/__init__.py
+++ b/src/transformers/__init__.py
@@ -343,6 +343,7 @@
     ],
     "models.deberta_v2": ["DebertaV2Config"],
     "models.decision_transformer": ["DecisionTransformerConfig"],
+    "models.deepseek_v3": ["DeepseekV3Config"],
     "models.deformable_detr": ["DeformableDetrConfig"],
     "models.deit": ["DeiTConfig"],
     "models.deprecated": [],
@@ -1985,6 +1986,13 @@
             "DecisionTransformerPreTrainedModel",
         ]
     )
+    _import_structure["models.deepseek_v3"].extend(
+        [
+            "DeepseekV3ForCausalLM",
+            "DeepseekV3Model",
+            "DeepseekV3PreTrainedModel",
+        ]
+    )
     _import_structure["models.deformable_detr"].extend(
         [
             "DeformableDetrForObjectDetection",
@@ -5452,6 +5460,9 @@
     from .models.decision_transformer import (
         DecisionTransformerConfig,
     )
+    from .models.deepseek_v3 import (
+        DeepseekV3Config,
+    )
     from .models.deformable_detr import (
         DeformableDetrConfig,
     )
@@ -7044,6 +7055,11 @@
             DecisionTransformerModel,
             DecisionTransformerPreTrainedModel,
         )
+        from .models.deepseek_v3 import (
+            DeepseekV3ForCausalLM,
+            DeepseekV3Model,
+            DeepseekV3PreTrainedModel,
+        )
         from .models.deformable_detr import (
             DeformableDetrForObjectDetection,
             DeformableDetrModel,
diff --git a/src/transformers/modeling_rope_utils.py b/src/transformers/modeling_rope_utils.py
index b2d343e0237f..beb0e9e2f793 100644
--- a/src/transformers/modeling_rope_utils.py
+++ b/src/transformers/modeling_rope_utils.py
@@ -189,13 +189,31 @@ def _compute_yarn_parameters(
     partial_rotary_factor = config.partial_rotary_factor if hasattr(config, "partial_rotary_factor") else 1.0
     head_dim = getattr(config, "head_dim", config.hidden_size // config.num_attention_heads)
     dim = int(head_dim * partial_rotary_factor)
-    max_position_embeddings = config.max_position_embeddings
     factor = config.rope_scaling["factor"]
+    attention_factor = config.rope_scaling.get("attention_factor")
+    mscale = config.rope_scaling.get("mscale")
+    mscale_all_dim = config.rope_scaling.get("mscale_all_dim")
+
+    # NOTE: DeekSeek-V3 (and potentially other models) modify `max_position_embeddings` and have a
+    # `original_max_position_embeddings` field containing the pretrained value. They use the ratio between these two
+    # values to compute the default attention scaling factor, instead of using `factor`.
+    if "original_max_position_embeddings" in config.rope_scaling:
+        original_max_position_embeddings = config.rope_scaling["original_max_position_embeddings"]
+        factor = config.max_position_embeddings / original_max_position_embeddings
+    else:
+        original_max_position_embeddings = config.max_position_embeddings
+
+    def get_mscale(scale, mscale=1):
+        if scale <= 1:
+            return 1.0
+        return 0.1 * mscale * math.log(scale) + 1.0
 
     # Sets the attention factor as suggested in the paper
-    attention_factor = config.rope_scaling.get("attention_factor")
     if attention_factor is None:
-        attention_factor = 0.1 * math.log(factor) + 1.0
+        if mscale and mscale_all_dim:
+            attention_factor = float(get_mscale(factor, mscale) / get_mscale(factor, mscale_all_dim))
+        else:
+            attention_factor = get_mscale(factor)
 
     # Optional config options
     # beta_fast/beta_slow: as suggested in the paper, default to 32/1 (correspondingly)
@@ -227,7 +245,7 @@ def linear_ramp_factor(min, max, dim):
     inv_freq_extrapolation = 1.0 / pos_freqs
     inv_freq_interpolation = 1.0 / (factor * pos_freqs)
 
-    low, high = find_correction_range(beta_fast, beta_slow, dim, base, max_position_embeddings)
+    low, high = find_correction_range(beta_fast, beta_slow, dim, base, original_max_position_embeddings)
 
     # Get n-dimensional rotational scaling corrected for extrapolation
     inv_freq_extrapolation_factor = 1 - linear_ramp_factor(low, high, dim // 2).float().to(device)
@@ -235,7 +253,6 @@ def linear_ramp_factor(min, max, dim):
         inv_freq_interpolation * (1 - inv_freq_extrapolation_factor)
         + inv_freq_extrapolation * inv_freq_extrapolation_factor
     )
-
     return inv_freq, attention_factor
 
 
@@ -425,7 +442,14 @@ def _validate_yarn_parameters(config: PretrainedConfig, ignore_keys: Optional[se
     rope_scaling = config.rope_scaling
     rope_type = rope_scaling.get("rope_type", rope_scaling.get("type", None))  # BC: "rope_type" was originally "type"
     required_keys = {"rope_type", "factor"}
-    optional_keys = {"attention_factor", "beta_fast", "beta_slow"}
+    optional_keys = {
+        "attention_factor",
+        "beta_fast",
+        "beta_slow",
+        "original_max_position_embeddings",
+        "mscale",
+        "mscale_all_dim",
+    }
     received_keys = set(rope_scaling.keys())
     _check_received_keys(rope_type, received_keys, required_keys, optional_keys, ignore_keys=ignore_keys)
 
diff --git a/src/transformers/models/__init__.py b/src/transformers/models/__init__.py
index b84d184225c0..a11cd2e7f1e5 100644
--- a/src/transformers/models/__init__.py
+++ b/src/transformers/models/__init__.py
@@ -70,6 +70,7 @@
     deberta,
     deberta_v2,
     decision_transformer,
+    deepseek_v3,
     deformable_detr,
     deit,
     deprecated,
diff --git a/src/transformers/models/auto/configuration_auto.py b/src/transformers/models/auto/configuration_auto.py
index 08f97da89610..ed6cac1e9dc5 100644
--- a/src/transformers/models/auto/configuration_auto.py
+++ b/src/transformers/models/auto/configuration_auto.py
@@ -88,6 +88,7 @@
         ("deberta", "DebertaConfig"),
         ("deberta-v2", "DebertaV2Config"),
         ("decision_transformer", "DecisionTransformerConfig"),
+        ("deepseek_v3", "DeepseekV3Config"),
         ("deformable_detr", "DeformableDetrConfig"),
         ("deit", "DeiTConfig"),
         ("depth_anything", "DepthAnythingConfig"),
@@ -412,6 +413,7 @@
         ("deberta", "DeBERTa"),
         ("deberta-v2", "DeBERTa-v2"),
         ("decision_transformer", "Decision Transformer"),
+        ("deepseek_v3", "DeepSeek-V3"),
         ("deformable_detr", "Deformable DETR"),
         ("deit", "DeiT"),
         ("deplot", "DePlot"),
diff --git a/src/transformers/models/auto/modeling_auto.py b/src/transformers/models/auto/modeling_auto.py
index 383e90e042ca..4a9f88258b95 100644
--- a/src/transformers/models/auto/modeling_auto.py
+++ b/src/transformers/models/auto/modeling_auto.py
@@ -87,6 +87,7 @@
         ("deberta", "DebertaModel"),
         ("deberta-v2", "DebertaV2Model"),
         ("decision_transformer", "DecisionTransformerModel"),
+        ("deepseek_v3", "DeepseekV3Model"),
         ("deformable_detr", "DeformableDetrModel"),
         ("deit", "DeiTModel"),
         ("depth_pro", "DepthProModel"),
@@ -506,6 +507,7 @@
         ("ctrl", "CTRLLMHeadModel"),
         ("data2vec-text", "Data2VecTextForCausalLM"),
         ("dbrx", "DbrxForCausalLM"),
+        ("deepseek_v3", "DeepseekV3ForCausalLM"),
         ("diffllama", "DiffLlamaForCausalLM"),
         ("electra", "ElectraForCausalLM"),
         ("emu3", "Emu3ForCausalLM"),
diff --git a/src/transformers/models/auto/tokenization_auto.py b/src/transformers/models/auto/tokenization_auto.py
index 5ee4f612285f..c01de1986a53 100644
--- a/src/transformers/models/auto/tokenization_auto.py
+++ b/src/transformers/models/auto/tokenization_auto.py
@@ -170,6 +170,13 @@
                     "DebertaV2TokenizerFast" if is_tokenizers_available() else None,
                 ),
             ),
+            (
+                "deepseek_v3",
+                (
+                    "LlamaTokenizer" if is_sentencepiece_available() else None,
+                    "LlamaTokenizerFast" if is_tokenizers_available() else None,
+                ),
+            ),
             (
                 "diffllama",
                 (
diff --git a/src/transformers/models/deepseek_v3/__init__.py b/src/transformers/models/deepseek_v3/__init__.py
new file mode 100644
index 000000000000..298f4c968375
--- /dev/null
+++ b/src/transformers/models/deepseek_v3/__init__.py
@@ -0,0 +1,27 @@
+# Copyright 2025 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_deepseek_v3 import *
+    from .modeling_deepseek_v3 import *
+else:
+    import sys
+
+    _file = globals()["__file__"]
+    sys.modules[__name__] = _LazyModule(__name__, _file, define_import_structure(_file), module_spec=__spec__)
diff --git a/src/transformers/models/deepseek_v3/configuration_deepseek_v3.py b/src/transformers/models/deepseek_v3/configuration_deepseek_v3.py
new file mode 100644
index 000000000000..ccfc97832c1b
--- /dev/null
+++ b/src/transformers/models/deepseek_v3/configuration_deepseek_v3.py
@@ -0,0 +1,242 @@
+# coding=utf-8
+# Copyright 2025 bzantium and the HuggingFace Inc. team. All rights reserved.
+#
+# This code is based on the DeepSeekV3 implementations from the DeepSeek AI team. (https://huggingface.co/deepseek-ai/DeepSeek-V3)
+
+# 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.
+"""DeepSeekV3 model configuration"""
+
+from ...configuration_utils import PretrainedConfig
+from ...modeling_rope_utils import rope_config_validation
+
+
+DEEPSEEK_PRETRAINED_CONFIG_ARCHIVE_MAP = {}
+
+
+class DeepseekV3Config(PretrainedConfig):
+    r"""
+    This is the configuration class to store the configuration of a [`DeepseekV3Model`]. It is used to instantiate an DeepSeek
+    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 DeepSeek-V3.
+    e.g. [bzantium/tiny-deepseek-v3](https://huggingface.co/bzantium/tiny-deepseek-v3)
+    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 129280):
+            Vocabulary size of the Deep model. Defines the number of different tokens that can be represented by the
+            `inputs_ids` passed when calling [`DeepseekV3Model`]
+        hidden_size (`int`, *optional*, defaults to 7168):
+            Dimension of the hidden representations.
+        intermediate_size (`int`, *optional*, defaults to 18432):
+            Dimension of the MLP representations.
+        moe_intermediate_size (`int`, *optional*, defaults to 2048):
+            Dimension of the MoE representations.
+        num_hidden_layers (`int`, *optional*, defaults to 61):
+            Number of hidden layers in the Transformer decoder.
+        num_attention_heads (`int`, *optional*, defaults to 128):
+            Number of attention heads for each attention layer in the Transformer decoder.
+        num_key_value_heads (`int`, *optional*, defaults to 128):
+            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
+            `num_attention_heads`.
+        n_shared_experts (`int`, *optional*, defaults to 1):
+            Number of shared experts.
+        n_routed_experts (`int`, *optional*, defaults to 256):
+            Number of routed experts.
+        routed_scaling_factor (`float`, *optional*, defaults to 2.5):
+            Scaling factor or routed experts.
+        kv_lora_rank (`int`, *optional*, defaults to 512):
+            Rank of the LoRA matrices for key and value projections.
+        q_lora_rank (`int`, *optional*, defaults to 1536):
+            Rank of the LoRA matrices for query projections.
+        qk_rope_head_dim (`int`, *optional*, defaults to 64):
+            Dimension of the query/key heads that use rotary position embeddings.
+        v_head_dim (`int`, *optional*, defaults to 128):
+            Dimension of the value heads.
+        qk_nope_head_dim (`int`, *optional*, defaults to 128):
+            Dimension of the query/key heads that don't use rotary position embeddings.
+        n_group (`int`, *optional*, defaults to 8):
+            Number of groups for routed experts.
+        topk_group (`int`, *optional*, defaults to 4):
+            Number of selected groups for each token(for each token, ensuring the selected experts is only within `topk_group` groups).
+        num_experts_per_tok (`int`, *optional*, defaults to 8):
+            Number of selected experts, None means dense model.
+        first_k_dense_replace (`int`, *optional*, defaults to 3):
+            Number of dense layers in shallow layers(embed->dense->dense->...->dense->moe->moe...->lm_head).
+                                                            \--k dense layers--/
+        norm_topk_prob (`bool`, *optional*, defaults to `True`):
+            Whether to normalize the weights of the routed experts.
+        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):
+            The maximum sequence length that this model might ever be used with.
+        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*):
+            Padding token id.
+        bos_token_id (`int`, *optional*, defaults to 0):
+            Beginning of stream token id.
+        eos_token_id (`int`, *optional*, defaults to 1):
+            End of stream token id.
+        pretraining_tp (`int`, *optional*, defaults to 1):
+            Experimental feature. Tensor parallelism rank used during pretraining. Please refer to [this
+            document](https://huggingface.co/docs/transformers/parallelism) to understand more about it. This value is
+            necessary to ensure exact reproducibility of the pretraining results. Please refer to [this
+            issue](https://github.com/pytorch/pytorch/issues/76232).
+        tie_word_embeddings (`bool`, *optional*, defaults to `False`):
+            Whether to tie weight embeddings
+        rope_theta (`float`, *optional*, defaults to 10000.0):
+            The base period of the RoPE embeddings.
+        rope_scaling (`Dict`, *optional*):
+            Dictionary containing the scaling configuration for the RoPE embeddings. Currently supports two scaling
+            strategies: linear and dynamic. Their scaling factor must be a float greater than 1. The expected format is
+            `{"type": strategy name, "factor": scaling factor}`. When using this flag, don't update
+            `max_position_embeddings` to the expected new maximum.
+        rope_interleave (`bool`, *optional*, defaults to `True`):
+            Whether to interleave the rotary position embeddings.
+        attention_bias (`bool`, defaults to `False`, *optional*, defaults to `False`):
+            Whether to use a bias in the query, key, value and output projection layers during self-attention.
+        attention_dropout (`float`, *optional*, defaults to 0.0):
+            The dropout ratio for the attention probabilities.
+
+    ```python
+    >>> from transformers import DeepseekV3Model, DeepseekV3Config
+
+    >>> # Initializing a Deepseek-V3 style configuration
+    >>> configuration = DeepseekV3Config()
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "deepseek_v3"
+    keys_to_ignore_at_inference = ["past_key_values"]
+    # Default tensor parallel plan for base model `DeepseekV3Model`
+    base_model_tp_plan = {
+        "layers.*.self_attn.q_b_proj": "colwise",
+        "layers.*.self_attn.kv_b_proj": "colwise",
+        "layers.*.self_attn.o_proj": "rowwise",
+        "layers.*.gate_proj": "colwise",
+        "layers.*.up_proj": "colwise",
+        "layers.*.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=129280,
+        hidden_size=7168,
+        intermediate_size=18432,
+        moe_intermediate_size=2048,
+        num_hidden_layers=61,
+        num_attention_heads=128,
+        num_key_value_heads=128,
+        n_shared_experts=1,
+        n_routed_experts=256,
+        routed_scaling_factor=2.5,
+        kv_lora_rank=512,
+        q_lora_rank=1536,
+        qk_rope_head_dim=64,
+        v_head_dim=128,
+        qk_nope_head_dim=128,
+        n_group=8,
+        topk_group=4,
+        num_experts_per_tok=8,
+        first_k_dense_replace=3,
+        norm_topk_prob=True,
+        hidden_act="silu",
+        max_position_embeddings=4096,
+        initializer_range=0.02,
+        rms_norm_eps=1e-6,
+        use_cache=True,
+        pad_token_id=None,
+        bos_token_id=0,
+        eos_token_id=1,
+        pretraining_tp=1,
+        tie_word_embeddings=False,
+        rope_theta=10000.0,
+        rope_scaling=None,
+        rope_interleave=True,
+        attention_bias=False,
+        attention_dropout=0.0,
+        **kwargs,
+    ):
+        self.vocab_size = vocab_size
+        self.max_position_embeddings = max_position_embeddings
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.moe_intermediate_size = moe_intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.n_shared_experts = n_shared_experts
+        self.n_routed_experts = n_routed_experts
+        self.routed_scaling_factor = routed_scaling_factor
+        self.kv_lora_rank = kv_lora_rank
+        self.q_lora_rank = q_lora_rank
+        self.qk_rope_head_dim = qk_rope_head_dim
+        self.v_head_dim = v_head_dim
+        self.qk_nope_head_dim = qk_nope_head_dim
+        self.qk_head_dim = qk_nope_head_dim + qk_rope_head_dim
+        self.head_dim = qk_rope_head_dim
+        self.n_group = n_group
+        self.topk_group = topk_group
+        self.num_experts_per_tok = num_experts_per_tok
+        self.first_k_dense_replace = first_k_dense_replace
+        self.norm_topk_prob = norm_topk_prob
+        self.rope_interleave = rope_interleave
+
+        # 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.pretraining_tp = pretraining_tp
+        self.use_cache = use_cache
+        self.rope_theta = rope_theta
+        self.rope_scaling = rope_scaling
+        self.attention_bias = attention_bias
+        self.attention_dropout = attention_dropout
+        # Validate the correctness of rotary position embeddings parameters
+        # BC: if there is a 'type' field, copy it it to 'rope_type'.
+        if self.rope_scaling is not None and "type" in self.rope_scaling:
+            self.rope_scaling["rope_type"] = self.rope_scaling["type"]
+        rope_config_validation(self)
+
+        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,
+        )
+
+
+__all__ = ["DeepseekV3Config"]
diff --git a/src/transformers/models/deepseek_v3/modeling_deepseek_v3.py b/src/transformers/models/deepseek_v3/modeling_deepseek_v3.py
new file mode 100644
index 000000000000..872eb89e77c1
--- /dev/null
+++ b/src/transformers/models/deepseek_v3/modeling_deepseek_v3.py
@@ -0,0 +1,1035 @@
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+#           This file was automatically generated from src/transformers/models/deepseek_v3/modular_deepseek_v3.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_deepseek_v3.py file directly. One of our CI enforces this.
+#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
+import math
+from typing import Callable, List, Optional, Tuple, Union
+
+import torch
+import torch.nn.functional as F
+from torch import nn
+
+from ...activations import ACT2FN
+from ...cache_utils import Cache, DynamicCache, StaticCache
+from ...generation import GenerationMixin
+from ...modeling_attn_mask_utils import AttentionMaskConverter
+from ...modeling_flash_attention_utils import FlashAttentionKwargs
+from ...modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast
+from ...modeling_rope_utils import ROPE_INIT_FUNCTIONS
+from ...modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
+from ...processing_utils import Unpack
+from ...utils import (
+    LossKwargs,
+    add_start_docstrings,
+    add_start_docstrings_to_model_forward,
+    logging,
+    replace_return_docstrings,
+)
+from ...utils.deprecation import deprecate_kwarg
+from .configuration_deepseek_v3 import DeepseekV3Config
+
+
+logger = logging.get_logger(__name__)
+_CONFIG_FOR_DOC = "DeepseekV3Config"
+
+
+class DeepseekV3RMSNorm(nn.Module):
+    def __init__(self, hidden_size, eps=1e-6):
+        """
+        DeepseekV3RMSNorm is equivalent to T5LayerNorm
+        """
+        super().__init__()
+        self.weight = nn.Parameter(torch.ones(hidden_size))
+        self.variance_epsilon = eps
+
+    def forward(self, hidden_states):
+        input_dtype = hidden_states.dtype
+        hidden_states = hidden_states.to(torch.float32)
+        variance = hidden_states.pow(2).mean(-1, keepdim=True)
+        hidden_states = hidden_states * torch.rsqrt(variance + self.variance_epsilon)
+        return self.weight * hidden_states.to(input_dtype)
+
+    def extra_repr(self):
+        return f"{tuple(self.weight.shape)}, eps={self.variance_epsilon}"
+
+
+class DeepseekV3RotaryEmbedding(nn.Module):
+    def __init__(self, config: DeepseekV3Config, device=None):
+        super().__init__()
+        # BC: "rope_type" was originally "type"
+        if hasattr(config, "rope_scaling") and config.rope_scaling is not None:
+            self.rope_type = config.rope_scaling.get("rope_type", config.rope_scaling.get("type"))
+        else:
+            self.rope_type = "default"
+        self.max_seq_len_cached = config.max_position_embeddings
+        self.original_max_seq_len = config.max_position_embeddings
+
+        self.config = config
+        self.rope_init_fn = ROPE_INIT_FUNCTIONS[self.rope_type]
+
+        inv_freq, self.attention_scaling = self.rope_init_fn(self.config, device)
+        self.register_buffer("inv_freq", inv_freq, persistent=False)
+        self.original_inv_freq = self.inv_freq
+
+    def _dynamic_frequency_update(self, position_ids, device):
+        """
+        dynamic RoPE layers should recompute `inv_freq` in the following situations:
+        1 - growing beyond the cached sequence length (allow scaling)
+        2 - the current sequence length is in the original scale (avoid losing precision with small sequences)
+        """
+        seq_len = torch.max(position_ids) + 1
+        if seq_len > self.max_seq_len_cached:  # growth
+            inv_freq, self.attention_scaling = self.rope_init_fn(self.config, device, seq_len=seq_len)
+            self.register_buffer("inv_freq", inv_freq, persistent=False)  # TODO joao: may break with compilation
+            self.max_seq_len_cached = seq_len
+
+        if seq_len < self.original_max_seq_len and self.max_seq_len_cached > self.original_max_seq_len:  # reset
+            # This .to() is needed if the model has been moved to a device after being initialized (because
+            # the buffer is automatically moved, but not the original copy)
+            self.original_inv_freq = self.original_inv_freq.to(device)
+            self.register_buffer("inv_freq", self.original_inv_freq, persistent=False)
+            self.max_seq_len_cached = self.original_max_seq_len
+
+    @torch.no_grad()
+    def forward(self, x, position_ids):
+        if "dynamic" in self.rope_type:
+            self._dynamic_frequency_update(position_ids, device=x.device)
+
+        # Core RoPE block
+        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
+        position_ids_expanded = position_ids[:, None, :].float()
+        # Force float32 (see https://github.com/huggingface/transformers/pull/29285)
+        device_type = x.device.type
+        device_type = device_type if isinstance(device_type, str) and device_type != "mps" else "cpu"
+        with torch.autocast(device_type=device_type, enabled=False):
+            freqs = (inv_freq_expanded.float() @ position_ids_expanded.float()).transpose(1, 2)
+            emb = torch.cat((freqs, freqs), dim=-1)
+            cos = emb.cos()
+            sin = emb.sin()
+
+        # Advanced RoPE types (e.g. yarn) apply a post-processing scaling factor, equivalent to scaling attention
+        cos = cos * self.attention_scaling
+        sin = sin * self.attention_scaling
+
+        return cos.to(dtype=x.dtype), sin.to(dtype=x.dtype)
+
+
+class DeepseekV3MLP(nn.Module):
+    def __init__(self, config, hidden_size=None, intermediate_size=None):
+        super().__init__()
+        self.config = config
+        self.hidden_size = config.hidden_size if hidden_size is None else hidden_size
+        self.intermediate_size = config.intermediate_size if intermediate_size is None else intermediate_size
+
+        self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
+        self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
+        self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=False)
+        self.act_fn = ACT2FN[config.hidden_act]
+
+    def forward(self, x):
+        down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
+        return down_proj
+
+
+class DeepseekV3TopkRouter(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        self.top_k = config.num_experts_per_tok
+        self.n_routed_experts = config.n_routed_experts
+        self.routed_scaling_factor = config.routed_scaling_factor
+        self.n_group = config.n_group
+        self.topk_group = config.topk_group
+        self.norm_topk_prob = config.norm_topk_prob
+
+        self.weight = nn.Parameter(torch.empty((self.n_routed_experts, config.hidden_size)))
+        self.register_buffer("e_score_correction_bias", torch.zeros((self.n_routed_experts)))
+
+    def forward(self, hidden_states):
+        hidden_states = hidden_states.view(-1, self.config.hidden_size)
+        router_logits = F.linear(hidden_states.type(torch.float32), self.weight.type(torch.float32))
+        scores = router_logits.sigmoid()
+        topk_indices = self.get_topk_indices(scores)
+        topk_weights = scores.gather(1, topk_indices)
+        if self.norm_topk_prob:
+            denominator = topk_weights.sum(dim=-1, keepdim=True) + 1e-20
+            topk_weights /= denominator
+        topk_weights = topk_weights * self.routed_scaling_factor
+        return topk_indices, topk_weights
+
+    def get_topk_indices(self, scores):
+        scores_for_choice = scores.view(-1, self.n_routed_experts) + self.e_score_correction_bias.unsqueeze(0)
+        group_scores = (
+            scores_for_choice.view(-1, self.n_group, self.n_routed_experts // self.n_group)
+            .topk(2, dim=-1)[0]
+            .sum(dim=-1)
+        )
+        group_idx = torch.topk(group_scores, k=self.topk_group, dim=-1, sorted=False)[1]
+        group_mask = torch.zeros_like(group_scores)
+        group_mask.scatter_(1, group_idx, 1)
+        score_mask = (
+            group_mask.unsqueeze(-1)
+            .expand(-1, self.n_group, self.n_routed_experts // self.n_group)
+            .reshape(-1, self.n_routed_experts)
+        )
+        scores_for_choice = scores_for_choice.masked_fill(~score_mask.bool(), 0.0)
+        topk_indices = torch.topk(scores_for_choice, k=self.top_k, dim=-1, sorted=False)[1]
+        return topk_indices
+
+
+class DeepseekV3MoE(nn.Module):
+    """
+    A mixed expert module containing shared experts.
+    """
+
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        self.experts = nn.ModuleList(
+            [
+                DeepseekV3MLP(config, intermediate_size=config.moe_intermediate_size)
+                for _ in range(config.n_routed_experts)
+            ]
+        )
+        self.gate = DeepseekV3TopkRouter(config)
+        self.shared_experts = DeepseekV3MLP(
+            config=config, intermediate_size=config.moe_intermediate_size * config.n_shared_experts
+        )
+
+    def forward(self, hidden_states):
+        residuals = hidden_states
+        orig_shape = hidden_states.shape
+        topk_indices, topk_weights = self.gate(hidden_states)
+        hidden_states = hidden_states.view(-1, hidden_states.shape[-1])
+        hidden_states = self.moe(hidden_states, topk_indices, topk_weights).view(*orig_shape)
+        hidden_states = hidden_states + self.shared_experts(residuals)
+        return hidden_states
+
+    def moe(self, hidden_states: torch.Tensor, topk_indices: torch.Tensor, topk_weights: torch.Tensor):
+        final_hidden_states = torch.zeros_like(hidden_states, dtype=topk_weights.dtype)
+        expert_mask = torch.nn.functional.one_hot(topk_indices, num_classes=len(self.experts))
+        expert_mask = expert_mask.permute(2, 0, 1)
+
+        for expert_idx in range(len(self.experts)):
+            expert = self.experts[expert_idx]
+            mask = expert_mask[expert_idx]
+            token_indices, weight_indices = torch.where(mask)
+
+            if token_indices.numel() > 0:
+                expert_weights = topk_weights[token_indices, weight_indices]
+                expert_input = hidden_states[token_indices]
+                expert_output = expert(expert_input)
+                weighted_output = expert_output * expert_weights.unsqueeze(-1)
+                final_hidden_states.index_add_(0, token_indices, weighted_output)
+        return final_hidden_states.type(hidden_states.dtype)
+
+
+def rotate_half(x):
+    """Rotates half the hidden dims of the input."""
+    x1 = x[..., : x.shape[-1] // 2]
+    x2 = x[..., x.shape[-1] // 2 :]
+    return torch.cat((-x2, x1), dim=-1)
+
+
+def apply_rotary_pos_emb(q, k, cos, sin, position_ids=None, unsqueeze_dim=1):
+    """Applies Rotary Position Embedding to the query and key tensors.
+
+    Args:
+        q (`torch.Tensor`): The query tensor.
+        k (`torch.Tensor`): The key tensor.
+        cos (`torch.Tensor`): The cosine part of the rotary embedding.
+        sin (`torch.Tensor`): The sine part of the rotary embedding.
+        position_ids (`torch.Tensor`, *optional*):
+            Deprecated and unused.
+        unsqueeze_dim (`int`, *optional*, defaults to 1):
+            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
+            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
+            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
+            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
+            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
+            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
+    Returns:
+        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
+    """
+    cos = cos.unsqueeze(unsqueeze_dim)
+    sin = sin.unsqueeze(unsqueeze_dim)
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+
+
+def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
+    """
+    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
+    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
+    """
+    batch, num_key_value_heads, slen, head_dim = hidden_states.shape
+    if n_rep == 1:
+        return hidden_states
+    hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
+    return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)
+
+
+def eager_attention_forward(
+    module: nn.Module,
+    query: torch.Tensor,
+    key: torch.Tensor,
+    value: torch.Tensor,
+    attention_mask: Optional[torch.Tensor],
+    scaling: float,
+    dropout: float = 0.0,
+    **kwargs,
+):
+    key_states = repeat_kv(key, module.num_key_value_groups)
+    value_states = repeat_kv(value, module.num_key_value_groups)
+
+    attn_weights = torch.matmul(query, key_states.transpose(2, 3)) * scaling
+    if attention_mask is not None:
+        causal_mask = attention_mask[:, :, :, : key_states.shape[-2]]
+        attn_weights = attn_weights + causal_mask
+
+    attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query.dtype)
+    attn_weights = nn.functional.dropout(attn_weights, p=dropout, training=module.training)
+    attn_output = torch.matmul(attn_weights, value_states)
+    attn_output = attn_output.transpose(1, 2).contiguous()
+
+    return attn_output, attn_weights
+
+
+def apply_rotary_pos_emb_interleave(q, k, cos, sin, position_ids=None, unsqueeze_dim=1):
+    """Applies Rotary Position Embedding to the query and key tensors.
+    Args:
+        q (`torch.Tensor`): The query tensor.
+        k (`torch.Tensor`): The key tensor.
+        cos (`torch.Tensor`): The cosine part of the rotary embedding.
+        sin (`torch.Tensor`): The sine part of the rotary embedding.
+        position_ids (`torch.Tensor`):
+            The position indices of the tokens corresponding to the query and key tensors. For example, this can be
+            used to pass offsetted position ids when working with a KV-cache.
+        unsqueeze_dim (`int`, *optional*, defaults to 1):
+            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
+            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
+            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
+            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
+            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
+            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
+    Returns:
+        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
+    """
+    cos = cos.unsqueeze(unsqueeze_dim)
+    sin = sin.unsqueeze(unsqueeze_dim)
+
+    b, h, s, d = q.shape
+    q = q.view(b, h, s, d // 2, 2).transpose(4, 3).reshape(b, h, s, d)
+
+    b, h, s, d = k.shape
+    k = k.view(b, h, s, d // 2, 2).transpose(4, 3).reshape(b, h, s, d)
+
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+
+
+def yarn_get_mscale(scale=1, mscale=1):
+    if scale <= 1:
+        return 1.0
+    return 0.1 * mscale * math.log(scale) + 1.0
+
+
+class DeepseekV3Attention(nn.Module):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
+
+    def __init__(self, config: DeepseekV3Config, layer_idx: int):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+        self.num_key_value_groups = config.num_attention_heads // config.num_key_value_heads
+        self.attention_dropout = config.attention_dropout
+        self.num_heads = config.num_attention_heads
+        self.rope_theta = config.rope_theta
+        self.q_lora_rank = config.q_lora_rank
+        self.qk_rope_head_dim = config.qk_rope_head_dim
+        self.kv_lora_rank = config.kv_lora_rank
+        self.v_head_dim = config.v_head_dim
+        self.qk_nope_head_dim = config.qk_nope_head_dim
+        self.qk_head_dim = config.qk_head_dim
+
+        self.is_causal = True
+        self.q_a_proj = nn.Linear(config.hidden_size, config.q_lora_rank, bias=config.attention_bias)
+        self.q_a_layernorm = DeepseekV3RMSNorm(config.q_lora_rank)
+        self.q_b_proj = nn.Linear(config.q_lora_rank, self.num_heads * self.qk_head_dim, bias=False)
+
+        self.kv_a_proj_with_mqa = nn.Linear(
+            config.hidden_size,
+            self.kv_lora_rank + self.qk_rope_head_dim,
+            bias=config.attention_bias,
+        )
+        self.kv_a_layernorm = DeepseekV3RMSNorm(self.kv_lora_rank)
+        self.kv_b_proj = nn.Linear(
+            self.kv_lora_rank,
+            self.num_heads * (self.qk_nope_head_dim + self.v_head_dim),
+            bias=False,
+        )
+
+        self.o_proj = nn.Linear(
+            self.num_heads * self.v_head_dim,
+            config.hidden_size,
+            bias=config.attention_bias,
+        )
+
+        self.scaling = self.qk_head_dim ** (-0.5)
+        if self.config.rope_scaling is not None:
+            mscale_all_dim = self.config.rope_scaling.get("mscale_all_dim", 0)
+            scaling_factor = self.config.rope_scaling["factor"]
+            if mscale_all_dim:
+                mscale = yarn_get_mscale(scaling_factor, mscale_all_dim)
+                self.scaling = self.scaling * mscale * mscale
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        position_embeddings: Tuple[torch.Tensor, torch.Tensor],
+        attention_mask: Optional[torch.Tensor],
+        past_key_value: Optional[Cache] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Unpack[FlashAttentionKwargs],
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        batch_size, seq_length = hidden_states.shape[:-1]
+        query_shape = (batch_size, seq_length, -1, self.qk_head_dim)
+        key_shape = (batch_size, seq_length, -1, self.qk_nope_head_dim + self.v_head_dim)
+
+        q_states = self.q_b_proj(self.q_a_layernorm(self.q_a_proj(hidden_states))).view(query_shape).transpose(1, 2)
+        q_pass, q_rot = torch.split(q_states, [self.qk_nope_head_dim, self.qk_rope_head_dim], dim=-1)
+
+        compressed_kv = self.kv_a_proj_with_mqa(hidden_states)
+        k_pass, k_rot = torch.split(compressed_kv, [self.kv_lora_rank, self.qk_rope_head_dim], dim=-1)
+
+        k_pass = self.kv_b_proj(self.kv_a_layernorm(k_pass)).view(key_shape).transpose(1, 2)
+        k_pass, value_states = torch.split(k_pass, [self.qk_nope_head_dim, self.v_head_dim], dim=-1)
+
+        k_rot = k_rot.view(batch_size, 1, seq_length, self.qk_rope_head_dim)
+
+        cos, sin = position_embeddings
+        if self.config.rope_interleave:
+            q_rot, k_rot = apply_rotary_pos_emb_interleave(q_rot, k_rot, cos, sin)
+        else:
+            q_rot, k_rot = apply_rotary_pos_emb(q_rot, k_rot, cos, sin)
+        k_rot = k_rot.expand(*k_pass.shape[:-1], -1)
+
+        query_states = torch.cat((q_pass, q_rot), dim=-1)
+        key_states = torch.cat((k_pass, k_rot), dim=-1)
+
+        if past_key_value is not None:
+            # sin and cos are specific to RoPE models; cache_position needed for the static cache
+            cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        if self.config._attn_implementation == "flash_attention_2" and self.qk_head_dim != self.v_head_dim:
+            value_states = F.pad(value_states, [0, self.qk_head_dim - self.v_head_dim])
+
+        attention_interface: Callable = eager_attention_forward
+        if self.config._attn_implementation != "eager":
+            if self.config._attn_implementation == "sdpa" and kwargs.get("output_attentions", False):
+                logger.warning_once(
+                    "`torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to "
+                    'eager attention. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.'
+                )
+            else:
+                attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
+
+        attn_output, attn_weights = attention_interface(
+            self,
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            dropout=0.0 if not self.training else self.attention_dropout,
+            scaling=self.scaling,
+            **kwargs,
+        )
+
+        if self.config._attn_implementation == "flash_attention_2" and self.qk_head_dim != self.v_head_dim:
+            attn_output = attn_output[:, :, :, : self.v_head_dim]
+
+        attn_output = attn_output.reshape(batch_size, seq_length, -1).contiguous()
+        attn_output = self.o_proj(attn_output)
+        return attn_output, attn_weights
+
+
+class DeepseekV3DecoderLayer(nn.Module):
+    def __init__(self, config: DeepseekV3Config, layer_idx: int):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+
+        self.self_attn = DeepseekV3Attention(config=config, layer_idx=layer_idx)
+
+        if layer_idx >= config.first_k_dense_replace:
+            self.mlp = DeepseekV3MoE(config)
+        else:
+            self.mlp = DeepseekV3MLP(config)
+
+        self.input_layernorm = DeepseekV3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.post_attention_layernorm = DeepseekV3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: Optional[bool] = False,
+        use_cache: Optional[bool] = False,
+        cache_position: Optional[torch.LongTensor] = None,
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # necessary, but kept here for BC
+        **kwargs: Unpack[FlashAttentionKwargs],
+    ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
+        residual = hidden_states
+
+        hidden_states = self.input_layernorm(hidden_states)
+
+        # Self Attention
+        hidden_states, self_attn_weights = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            output_attentions=output_attentions,
+            use_cache=use_cache,
+            cache_position=cache_position,
+            position_embeddings=position_embeddings,
+            **kwargs,
+        )
+        hidden_states = residual + hidden_states
+
+        # Fully Connected
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = residual + hidden_states
+
+        outputs = (hidden_states,)
+        if output_attentions:
+            outputs += (self_attn_weights,)
+
+        return outputs
+
+
+DEEPSEEK_V3_START_DOCSTRING = r"""
+    This model inherits from [`PreTrainedModel`]. Check the superclass documentation for the generic methods the
+    library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads
+    etc.)
+
+    This model is also a PyTorch [torch.nn.Module](https://pytorch.org/docs/stable/nn.html#torch.nn.Module) subclass.
+    Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage
+    and behavior.
+
+    Parameters:
+        config ([`DeepseekV3Config`]):
+            Model configuration class with all the parameters of the model. Initializing with a config file does not
+            load the weights associated with the model, only the configuration. Check out the
+            [`~PreTrainedModel.from_pretrained`] method to load the model weights.
+"""
+
+
+@add_start_docstrings(
+    "The bare DeepseekV3 Model outputting raw hidden-states without any specific head on top.",
+    DEEPSEEK_V3_START_DOCSTRING,
+)
+class DeepseekV3PreTrainedModel(PreTrainedModel):
+    config_class = DeepseekV3Config
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["DeepseekV3DecoderLayer"]
+    _skip_keys_device_placement = ["past_key_values"]
+    _supports_flash_attn_2 = True
+    _supports_sdpa = True
+    _supports_flex_attn = True
+    _supports_cache_class = True
+    _supports_quantized_cache = True
+    _supports_static_cache = True
+    _supports_attention_backend = True
+
+    def _init_weights(self, module):
+        std = self.config.initializer_range
+        if isinstance(module, nn.Linear):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.bias is not None:
+                module.bias.data.zero_()
+        elif isinstance(module, nn.Embedding):
+            module.weight.data.normal_(mean=0.0, std=std)
+            if module.padding_idx is not None:
+                module.weight.data[module.padding_idx].zero_()
+
+
+DEEPSEEK_V3_INPUTS_DOCSTRING = r"""
+    Args:
+        input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
+            Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
+            it.
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            [What are input IDs?](../glossary#input-ids)
+        attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
+
+            - 1 for tokens that are **not masked**,
+            - 0 for tokens that are **masked**.
+
+            [What are attention masks?](../glossary#attention-mask)
+
+            Indices can be obtained using [`AutoTokenizer`]. See [`PreTrainedTokenizer.encode`] and
+            [`PreTrainedTokenizer.__call__`] for details.
+
+            If `past_key_values` is used, optionally only the last `input_ids` have to be input (see
+            `past_key_values`).
+
+            If you want to change padding behavior, you should read [`modeling_opt._prepare_decoder_attention_mask`]
+            and modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more
+            information on the default strategy.
+
+            - 1 indicates the head is **not masked**,
+            - 0 indicates the head is **masked**.
+        position_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range `[0,
+            config.n_positions - 1]`.
+
+            [What are position IDs?](../glossary#position-ids)
+        past_key_values (`Cache` or `tuple(tuple(torch.FloatTensor))`, *optional*):
+            Pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention
+            blocks) that can be used to speed up sequential decoding. This typically consists in the `past_key_values`
+            returned by the model at a previous stage of decoding, when `use_cache=True` or `config.use_cache=True`.
+
+            Two formats are allowed:
+            - a [`~cache_utils.Cache`] instance, see our
+            [kv cache guide](https://huggingface.co/docs/transformers/en/kv_cache);
+            - Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of
+            shape `(batch_size, num_heads, sequence_length, embed_size_per_head)`). This is also known as the legacy
+            cache format.
+
+            The model will output the same cache format that is fed as input. If no `past_key_values` are passed, the
+            legacy cache format will be returned.
+
+            If `past_key_values` are used, the user can optionally input only the last `input_ids` (those that don't
+            have their past key value states given to this model) of shape `(batch_size, 1)` instead of all `input_ids`
+            of shape `(batch_size, sequence_length)`.
+        inputs_embeds (`torch.FloatTensor` of shape `(batch_size, sequence_length, hidden_size)`, *optional*):
+            Optionally, instead of passing `input_ids` you can choose to directly pass an embedded representation. This
+            is useful if you want more control over how to convert `input_ids` indices into associated vectors than the
+            model's internal embedding lookup matrix.
+        use_cache (`bool`, *optional*):
+            If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding (see
+            `past_key_values`).
+        output_attentions (`bool`, *optional*):
+            Whether or not to return the attentions tensors of all attention layers. See `attentions` under returned
+            tensors for more detail.
+        output_hidden_states (`bool`, *optional*):
+            Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors for
+            more detail.
+        return_dict (`bool`, *optional*):
+            Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
+        cache_position (`torch.LongTensor` of shape `(sequence_length)`, *optional*):
+            Indices depicting the position of the input sequence tokens in the sequence. Contrarily to `position_ids`,
+            this tensor is not affected by padding. It is used to update the cache in the correct position and to infer
+            the complete sequence length.
+"""
+
+
+@add_start_docstrings(
+    "The bare DeepseekV3 Model outputting raw hidden-states without any specific head on top.",
+    DEEPSEEK_V3_START_DOCSTRING,
+)
+class DeepseekV3Model(DeepseekV3PreTrainedModel):
+    """
+    Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`DeepseekV3DecoderLayer`]
+
+    Args:
+        config: DeepseekV3Config
+    """
+
+    def __init__(self, config: DeepseekV3Config):
+        super().__init__(config)
+        self.padding_idx = config.pad_token_id
+        self.vocab_size = config.vocab_size
+
+        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
+        self.layers = nn.ModuleList(
+            [DeepseekV3DecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
+        )
+        self.norm = DeepseekV3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.rotary_emb = DeepseekV3RotaryEmbedding(config=config)
+        self.gradient_checkpointing = False
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.embed_tokens = value
+
+    @add_start_docstrings_to_model_forward(DEEPSEEK_V3_INPUTS_DOCSTRING)
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Cache] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **flash_attn_kwargs: Unpack[FlashAttentionKwargs],
+    ) -> Union[Tuple, BaseModelOutputWithPast]:
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        use_cache = use_cache if use_cache is not None else self.config.use_cache
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        if (input_ids is None) ^ (inputs_embeds is not None):
+            raise ValueError("You must specify exactly one of input_ids or inputs_embeds")
+
+        if self.gradient_checkpointing and self.training and use_cache:
+            logger.warning_once(
+                "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`."
+            )
+            use_cache = False
+
+        if inputs_embeds is None:
+            inputs_embeds = self.embed_tokens(input_ids)
+
+        if use_cache and past_key_values is None:
+            past_key_values = DynamicCache()
+
+        if cache_position is None:
+            past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
+            cache_position = torch.arange(
+                past_seen_tokens, past_seen_tokens + inputs_embeds.shape[1], device=inputs_embeds.device
+            )
+
+        if position_ids is None:
+            position_ids = cache_position.unsqueeze(0)
+
+        causal_mask = self._update_causal_mask(
+            attention_mask, inputs_embeds, cache_position, past_key_values, output_attentions
+        )
+
+        hidden_states = inputs_embeds
+
+        # create position embeddings to be shared across the decoder layers
+        position_embeddings = self.rotary_emb(hidden_states, position_ids)
+
+        # decoder layers
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attns = () if output_attentions else None
+
+        for decoder_layer in self.layers[: self.config.num_hidden_layers]:
+            if output_hidden_states:
+                all_hidden_states += (hidden_states,)
+
+            if self.gradient_checkpointing and self.training:
+                layer_outputs = self._gradient_checkpointing_func(
+                    decoder_layer.__call__,
+                    hidden_states,
+                    causal_mask,
+                    position_ids,
+                    past_key_values,
+                    output_attentions,
+                    use_cache,
+                    cache_position,
+                    position_embeddings,
+                )
+            else:
+                layer_outputs = decoder_layer(
+                    hidden_states,
+                    attention_mask=causal_mask,
+                    position_ids=position_ids,
+                    past_key_value=past_key_values,
+                    output_attentions=output_attentions,
+                    use_cache=use_cache,
+                    cache_position=cache_position,
+                    position_embeddings=position_embeddings,
+                    **flash_attn_kwargs,
+                )
+
+            hidden_states = layer_outputs[0]
+
+            if output_attentions:
+                all_self_attns += (layer_outputs[1],)
+
+        hidden_states = self.norm(hidden_states)
+
+        # add hidden states from the last decoder layer
+        if output_hidden_states:
+            all_hidden_states += (hidden_states,)
+
+        output = BaseModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=past_key_values if use_cache else None,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attns,
+        )
+        return output if return_dict else output.to_tuple()
+
+    def _update_causal_mask(
+        self,
+        attention_mask: torch.Tensor,
+        input_tensor: torch.Tensor,
+        cache_position: torch.Tensor,
+        past_key_values: Cache,
+        output_attentions: bool,
+    ):
+        if self.config._attn_implementation == "flash_attention_2":
+            if attention_mask is not None and (attention_mask == 0.0).any():
+                return attention_mask
+            return None
+
+        # For SDPA, when possible, we will rely on its `is_causal` argument instead of its `attn_mask` argument, in
+        # order to dispatch on Flash Attention 2. This feature is not compatible with static cache, as SDPA will fail
+        # to infer the attention mask.
+        past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
+        using_static_cache = isinstance(past_key_values, StaticCache)
+
+        # When output attentions is True, sdpa implementation's forward method calls the eager implementation's forward
+        if self.config._attn_implementation == "sdpa" and not using_static_cache and not output_attentions:
+            if AttentionMaskConverter._ignore_causal_mask_sdpa(
+                attention_mask,
+                inputs_embeds=input_tensor,
+                past_key_values_length=past_seen_tokens,
+                is_training=self.training,
+            ):
+                return None
+
+        dtype, device = input_tensor.dtype, input_tensor.device
+        sequence_length = input_tensor.shape[1]
+        if using_static_cache:
+            target_length = past_key_values.get_max_cache_shape()
+        else:
+            target_length = (
+                attention_mask.shape[-1]
+                if isinstance(attention_mask, torch.Tensor)
+                else past_seen_tokens + sequence_length + 1
+            )
+
+        # In case the provided `attention` mask is 2D, we generate a causal mask here (4D).
+        causal_mask = self._prepare_4d_causal_attention_mask_with_cache_position(
+            attention_mask,
+            sequence_length=sequence_length,
+            target_length=target_length,
+            dtype=dtype,
+            device=device,
+            cache_position=cache_position,
+            batch_size=input_tensor.shape[0],
+        )
+
+        if (
+            self.config._attn_implementation == "sdpa"
+            and attention_mask is not None
+            and attention_mask.device.type in ["cuda", "xpu"]
+            and not output_attentions
+        ):
+            # Attend to all tokens in fully masked rows in the causal_mask, for example the relevant first rows when
+            # using left padding. This is required by F.scaled_dot_product_attention memory-efficient attention path.
+            # Details: https://github.com/pytorch/pytorch/issues/110213
+            min_dtype = torch.finfo(dtype).min
+            causal_mask = AttentionMaskConverter._unmask_unattended(causal_mask, min_dtype)
+
+        return causal_mask
+
+    @staticmethod
+    def _prepare_4d_causal_attention_mask_with_cache_position(
+        attention_mask: torch.Tensor,
+        sequence_length: int,
+        target_length: int,
+        dtype: torch.dtype,
+        device: torch.device,
+        cache_position: torch.Tensor,
+        batch_size: int,
+        **kwargs,
+    ):
+        """
+        Creates a causal 4D mask of shape `(batch_size, 1, query_length, key_value_length)` from a 2D mask of shape
+        `(batch_size, key_value_length)`, or if the input `attention_mask` is already 4D, do nothing.
+
+        Args:
+            attention_mask (`torch.Tensor`):
+                A 2D attention mask of shape `(batch_size, key_value_length)` or a 4D attention mask of shape
+                `(batch_size, 1, query_length, key_value_length)`.
+            sequence_length (`int`):
+                The sequence length being processed.
+            target_length (`int`):
+                The target length: when generating with static cache, the mask should be as long as the static cache,
+                to account for the 0 padding, the part of the cache that is not filled yet.
+            dtype (`torch.dtype`):
+                The dtype to use for the 4D attention mask.
+            device (`torch.device`):
+                The device to plcae the 4D attention mask on.
+            cache_position (`torch.Tensor`):
+                Indices depicting the position of the input sequence tokens in the sequence.
+            batch_size (`torch.Tensor`):
+                Batch size.
+        """
+        if attention_mask is not None and attention_mask.dim() == 4:
+            # In this case we assume that the mask comes already in inverted form and requires no inversion or slicing.
+            causal_mask = attention_mask
+        else:
+            min_dtype = torch.finfo(dtype).min
+            causal_mask = torch.full(
+                (sequence_length, target_length), fill_value=min_dtype, dtype=dtype, device=device
+            )
+            if sequence_length != 1:
+                causal_mask = torch.triu(causal_mask, diagonal=1)
+            causal_mask *= torch.arange(target_length, device=device) > cache_position.reshape(-1, 1)
+            causal_mask = causal_mask[None, None, :, :].expand(batch_size, 1, -1, -1)
+            if attention_mask is not None:
+                causal_mask = causal_mask.clone()  # copy to contiguous memory for in-place edit
+                mask_length = attention_mask.shape[-1]
+                padding_mask = causal_mask[:, :, :, :mask_length] + attention_mask[:, None, None, :].to(
+                    causal_mask.device
+                )
+                padding_mask = padding_mask == 0
+                causal_mask[:, :, :, :mask_length] = causal_mask[:, :, :, :mask_length].masked_fill(
+                    padding_mask, min_dtype
+                )
+
+        return causal_mask
+
+
+class KwargsForCausalLM(FlashAttentionKwargs, LossKwargs): ...
+
+
+class DeepseekV3ForCausalLM(DeepseekV3PreTrainedModel, GenerationMixin):
+    _tied_weights_keys = ["lm_head.weight"]
+    _tp_plan = {"lm_head": "colwise_rep"}
+    _pp_plan = {"lm_head": (["hidden_states"], ["logits"])}
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.model = DeepseekV3Model(config)
+        self.vocab_size = config.vocab_size
+        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def get_input_embeddings(self):
+        return self.model.embed_tokens
+
+    def set_input_embeddings(self, value):
+        self.model.embed_tokens = value
+
+    def get_output_embeddings(self):
+        return self.lm_head
+
+    def set_output_embeddings(self, new_embeddings):
+        self.lm_head = new_embeddings
+
+    def set_decoder(self, decoder):
+        self.model = decoder
+
+    def get_decoder(self):
+        return self.model
+
+    @deprecate_kwarg("num_logits_to_keep", version="4.50", new_name="logits_to_keep")
+    @add_start_docstrings_to_model_forward(DEEPSEEK_V3_INPUTS_DOCSTRING)
+    @replace_return_docstrings(output_type=CausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC)
+    def forward(
+        self,
+        input_ids: torch.LongTensor = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_values: Optional[Union[Cache, List[torch.FloatTensor]]] = None,
+        inputs_embeds: Optional[torch.FloatTensor] = None,
+        labels: Optional[torch.LongTensor] = None,
+        use_cache: Optional[bool] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        logits_to_keep: Union[int, torch.Tensor] = 0,
+        **kwargs: Unpack[KwargsForCausalLM],
+    ) -> Union[Tuple, CausalLMOutputWithPast]:
+        r"""
+        Args:
+            labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
+                Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
+                config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
+                (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.
+
+            logits_to_keep (`int` or `torch.Tensor`, *optional*):
+                If an `int`, compute logits for the last `logits_to_keep` tokens. If `0`, calculate logits for all
+                `input_ids` (special case). Only last token logits are needed for generation, and calculating them only for that
+                token can save memory, which becomes pretty significant for long sequences or large vocabulary size.
+                If a `torch.Tensor`, must be 1D corresponding to the indices to keep in the sequence length dimension.
+                This is useful when using packed tensor format (single dimension for batch and sequence length).
+
+        Returns:
+
+        Example:
+
+        ```python
+        >>> from transformers import AutoTokenizer, DeepseekV3ForCausalLM
+
+        >>> model = DeepseekV3ForCausalLM.from_pretrained("meta-deepseek_v3/DeepseekV3-2-7b-hf")
+        >>> tokenizer = AutoTokenizer.from_pretrained("meta-deepseek_v3/DeepseekV3-2-7b-hf")
+
+        >>> prompt = "Hey, are you conscious? Can you talk to me?"
+        >>> inputs = tokenizer(prompt, return_tensors="pt")
+
+        >>> # Generate
+        >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
+        >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
+        "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
+        ```"""
+        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
+        output_hidden_states = (
+            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
+        )
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
+        outputs = self.model(
+            input_ids=input_ids,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_values=past_key_values,
+            inputs_embeds=inputs_embeds,
+            use_cache=use_cache,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+            cache_position=cache_position,
+            **kwargs,
+        )
+
+        hidden_states = outputs[0]
+        # Only compute necessary logits, and do not upcast them to float if we are not computing the loss
+        slice_indices = slice(-logits_to_keep, None) if isinstance(logits_to_keep, int) else logits_to_keep
+        logits = self.lm_head(hidden_states[:, slice_indices, :])
+
+        loss = None
+        if labels is not None:
+            loss = self.loss_function(logits=logits, labels=labels, vocab_size=self.config.vocab_size, **kwargs)
+
+        if not return_dict:
+            output = (logits,) + outputs[1:]
+            return (loss,) + output if loss is not None else output
+
+        return CausalLMOutputWithPast(
+            loss=loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+__all__ = ["DeepseekV3PreTrainedModel", "DeepseekV3Model", "DeepseekV3ForCausalLM"]
diff --git a/src/transformers/models/deepseek_v3/modular_deepseek_v3.py b/src/transformers/models/deepseek_v3/modular_deepseek_v3.py
new file mode 100644
index 000000000000..fe5ea154d8d0
--- /dev/null
+++ b/src/transformers/models/deepseek_v3/modular_deepseek_v3.py
@@ -0,0 +1,434 @@
+import math
+from typing import Callable, Optional, Tuple
+
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+from torch import nn
+
+from ...activations import ACT2FN
+from ...cache_utils import Cache
+from ...modeling_flash_attention_utils import FlashAttentionKwargs
+from ...modeling_utils import ALL_ATTENTION_FUNCTIONS
+from ...processing_utils import Unpack
+from ...utils import logging
+from ..llama.modeling_llama import (
+    LlamaForCausalLM,
+    LlamaModel,
+    LlamaPreTrainedModel,
+    LlamaRMSNorm,
+    LlamaRotaryEmbedding,
+    apply_rotary_pos_emb,
+    eager_attention_forward,
+    rotate_half,
+)
+from .configuration_deepseek_v3 import DeepseekV3Config
+
+
+logger = logging.get_logger(__name__)
+
+
+class DeepseekV3RMSNorm(LlamaRMSNorm):
+    pass
+
+
+class DeepseekV3RotaryEmbedding(LlamaRotaryEmbedding):
+    pass
+
+
+def apply_rotary_pos_emb_interleave(q, k, cos, sin, position_ids=None, unsqueeze_dim=1):
+    """Applies Rotary Position Embedding to the query and key tensors.
+    Args:
+        q (`torch.Tensor`): The query tensor.
+        k (`torch.Tensor`): The key tensor.
+        cos (`torch.Tensor`): The cosine part of the rotary embedding.
+        sin (`torch.Tensor`): The sine part of the rotary embedding.
+        position_ids (`torch.Tensor`):
+            The position indices of the tokens corresponding to the query and key tensors. For example, this can be
+            used to pass offsetted position ids when working with a KV-cache.
+        unsqueeze_dim (`int`, *optional*, defaults to 1):
+            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
+            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
+            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
+            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
+            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
+            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
+    Returns:
+        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
+    """
+    cos = cos.unsqueeze(unsqueeze_dim)
+    sin = sin.unsqueeze(unsqueeze_dim)
+
+    b, h, s, d = q.shape
+    q = q.view(b, h, s, d // 2, 2).transpose(4, 3).reshape(b, h, s, d)
+
+    b, h, s, d = k.shape
+    k = k.view(b, h, s, d // 2, 2).transpose(4, 3).reshape(b, h, s, d)
+
+    q_embed = (q * cos) + (rotate_half(q) * sin)
+    k_embed = (k * cos) + (rotate_half(k) * sin)
+    return q_embed, k_embed
+
+
+def yarn_get_mscale(scale=1, mscale=1):
+    if scale <= 1:
+        return 1.0
+    return 0.1 * mscale * math.log(scale) + 1.0
+
+
+class DeepseekV3MLP(nn.Module):
+    def __init__(self, config, hidden_size=None, intermediate_size=None):
+        super().__init__()
+        self.config = config
+        self.hidden_size = config.hidden_size if hidden_size is None else hidden_size
+        self.intermediate_size = config.intermediate_size if intermediate_size is None else intermediate_size
+
+        self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
+        self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
+        self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=False)
+        self.act_fn = ACT2FN[config.hidden_act]
+
+    def forward(self, x):
+        down_proj = self.down_proj(self.act_fn(self.gate_proj(x)) * self.up_proj(x))
+        return down_proj
+
+
+class DeepseekV3TopkRouter(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        self.top_k = config.num_experts_per_tok
+        self.n_routed_experts = config.n_routed_experts
+        self.routed_scaling_factor = config.routed_scaling_factor
+        self.n_group = config.n_group
+        self.topk_group = config.topk_group
+        self.norm_topk_prob = config.norm_topk_prob
+
+        self.weight = nn.Parameter(torch.empty((self.n_routed_experts, config.hidden_size)))
+        self.register_buffer("e_score_correction_bias", torch.zeros((self.n_routed_experts)))
+
+    def forward(self, hidden_states):
+        hidden_states = hidden_states.view(-1, self.config.hidden_size)
+        router_logits = F.linear(hidden_states.type(torch.float32), self.weight.type(torch.float32))
+        scores = router_logits.sigmoid()
+        topk_indices = self.get_topk_indices(scores)
+        topk_weights = scores.gather(1, topk_indices)
+        if self.norm_topk_prob:
+            denominator = topk_weights.sum(dim=-1, keepdim=True) + 1e-20
+            topk_weights /= denominator
+        topk_weights = topk_weights * self.routed_scaling_factor
+        return topk_indices, topk_weights
+
+    def get_topk_indices(self, scores):
+        scores_for_choice = scores.view(-1, self.n_routed_experts) + self.e_score_correction_bias.unsqueeze(0)
+        group_scores = (
+            scores_for_choice.view(-1, self.n_group, self.n_routed_experts // self.n_group)
+            .topk(2, dim=-1)[0]
+            .sum(dim=-1)
+        )
+        group_idx = torch.topk(group_scores, k=self.topk_group, dim=-1, sorted=False)[1]
+        group_mask = torch.zeros_like(group_scores)
+        group_mask.scatter_(1, group_idx, 1)
+        score_mask = (
+            group_mask.unsqueeze(-1)
+            .expand(-1, self.n_group, self.n_routed_experts // self.n_group)
+            .reshape(-1, self.n_routed_experts)
+        )
+        scores_for_choice = scores_for_choice.masked_fill(~score_mask.bool(), 0.0)
+        topk_indices = torch.topk(scores_for_choice, k=self.top_k, dim=-1, sorted=False)[1]
+        return topk_indices
+
+
+class DeepseekV3MoE(nn.Module):
+    """
+    A mixed expert module containing shared experts.
+    """
+
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        self.experts = nn.ModuleList(
+            [
+                DeepseekV3MLP(config, intermediate_size=config.moe_intermediate_size)
+                for _ in range(config.n_routed_experts)
+            ]
+        )
+        self.gate = DeepseekV3TopkRouter(config)
+        self.shared_experts = DeepseekV3MLP(
+            config=config, intermediate_size=config.moe_intermediate_size * config.n_shared_experts
+        )
+
+    def forward(self, hidden_states):
+        residuals = hidden_states
+        orig_shape = hidden_states.shape
+        topk_indices, topk_weights = self.gate(hidden_states)
+        hidden_states = hidden_states.view(-1, hidden_states.shape[-1])
+        hidden_states = self.moe(hidden_states, topk_indices, topk_weights).view(*orig_shape)
+        hidden_states = hidden_states + self.shared_experts(residuals)
+        return hidden_states
+
+    def moe(self, hidden_states: torch.Tensor, topk_indices: torch.Tensor, topk_weights: torch.Tensor):
+        final_hidden_states = torch.zeros_like(hidden_states, dtype=topk_weights.dtype)
+        expert_mask = torch.nn.functional.one_hot(topk_indices, num_classes=len(self.experts))
+        expert_mask = expert_mask.permute(2, 0, 1)
+
+        for expert_idx in range(len(self.experts)):
+            expert = self.experts[expert_idx]
+            mask = expert_mask[expert_idx]
+            token_indices, weight_indices = torch.where(mask)
+
+            if token_indices.numel() > 0:
+                expert_weights = topk_weights[token_indices, weight_indices]
+                expert_input = hidden_states[token_indices]
+                expert_output = expert(expert_input)
+                weighted_output = expert_output * expert_weights.unsqueeze(-1)
+                final_hidden_states.index_add_(0, token_indices, weighted_output)
+        return final_hidden_states.type(hidden_states.dtype)
+
+
+class DeepseekV3Attention(nn.Module):
+    """Multi-headed attention from 'Attention Is All You Need' paper"""
+
+    def __init__(self, config: DeepseekV3Config, layer_idx: int):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+        self.num_key_value_groups = config.num_attention_heads // config.num_key_value_heads
+        self.attention_dropout = config.attention_dropout
+        self.num_heads = config.num_attention_heads
+        self.rope_theta = config.rope_theta
+        self.q_lora_rank = config.q_lora_rank
+        self.qk_rope_head_dim = config.qk_rope_head_dim
+        self.kv_lora_rank = config.kv_lora_rank
+        self.v_head_dim = config.v_head_dim
+        self.qk_nope_head_dim = config.qk_nope_head_dim
+        self.qk_head_dim = config.qk_head_dim
+
+        self.is_causal = True
+        self.q_a_proj = nn.Linear(config.hidden_size, config.q_lora_rank, bias=config.attention_bias)
+        self.q_a_layernorm = DeepseekV3RMSNorm(config.q_lora_rank)
+        self.q_b_proj = nn.Linear(config.q_lora_rank, self.num_heads * self.qk_head_dim, bias=False)
+
+        self.kv_a_proj_with_mqa = nn.Linear(
+            config.hidden_size,
+            self.kv_lora_rank + self.qk_rope_head_dim,
+            bias=config.attention_bias,
+        )
+        self.kv_a_layernorm = DeepseekV3RMSNorm(self.kv_lora_rank)
+        self.kv_b_proj = nn.Linear(
+            self.kv_lora_rank,
+            self.num_heads * (self.qk_nope_head_dim + self.v_head_dim),
+            bias=False,
+        )
+
+        self.o_proj = nn.Linear(
+            self.num_heads * self.v_head_dim,
+            config.hidden_size,
+            bias=config.attention_bias,
+        )
+
+        self.scaling = self.qk_head_dim ** (-0.5)
+        if self.config.rope_scaling is not None:
+            mscale_all_dim = self.config.rope_scaling.get("mscale_all_dim", 0)
+            scaling_factor = self.config.rope_scaling["factor"]
+            if mscale_all_dim:
+                mscale = yarn_get_mscale(scaling_factor, mscale_all_dim)
+                self.scaling = self.scaling * mscale * mscale
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        position_embeddings: Tuple[torch.Tensor, torch.Tensor],
+        attention_mask: Optional[torch.Tensor],
+        past_key_value: Optional[Cache] = None,
+        cache_position: Optional[torch.LongTensor] = None,
+        **kwargs: Unpack[FlashAttentionKwargs],
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        batch_size, seq_length = hidden_states.shape[:-1]
+        query_shape = (batch_size, seq_length, -1, self.qk_head_dim)
+        key_shape = (batch_size, seq_length, -1, self.qk_nope_head_dim + self.v_head_dim)
+
+        q_states = self.q_b_proj(self.q_a_layernorm(self.q_a_proj(hidden_states))).view(query_shape).transpose(1, 2)
+        q_pass, q_rot = torch.split(q_states, [self.qk_nope_head_dim, self.qk_rope_head_dim], dim=-1)
+
+        compressed_kv = self.kv_a_proj_with_mqa(hidden_states)
+        k_pass, k_rot = torch.split(compressed_kv, [self.kv_lora_rank, self.qk_rope_head_dim], dim=-1)
+
+        k_pass = self.kv_b_proj(self.kv_a_layernorm(k_pass)).view(key_shape).transpose(1, 2)
+        k_pass, value_states = torch.split(k_pass, [self.qk_nope_head_dim, self.v_head_dim], dim=-1)
+
+        k_rot = k_rot.view(batch_size, 1, seq_length, self.qk_rope_head_dim)
+
+        cos, sin = position_embeddings
+        if self.config.rope_interleave:
+            q_rot, k_rot = apply_rotary_pos_emb_interleave(q_rot, k_rot, cos, sin)
+        else:
+            q_rot, k_rot = apply_rotary_pos_emb(q_rot, k_rot, cos, sin)
+        k_rot = k_rot.expand(*k_pass.shape[:-1], -1)
+
+        query_states = torch.cat((q_pass, q_rot), dim=-1)
+        key_states = torch.cat((k_pass, k_rot), dim=-1)
+
+        if past_key_value is not None:
+            # sin and cos are specific to RoPE models; cache_position needed for the static cache
+            cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
+            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)
+
+        if self.config._attn_implementation == "flash_attention_2" and self.qk_head_dim != self.v_head_dim:
+            value_states = F.pad(value_states, [0, self.qk_head_dim - self.v_head_dim])
+
+        attention_interface: Callable = eager_attention_forward
+        if self.config._attn_implementation != "eager":
+            if self.config._attn_implementation == "sdpa" and kwargs.get("output_attentions", False):
+                logger.warning_once(
+                    "`torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to "
+                    'eager attention. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.'
+                )
+            else:
+                attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]
+
+        attn_output, attn_weights = attention_interface(
+            self,
+            query_states,
+            key_states,
+            value_states,
+            attention_mask,
+            dropout=0.0 if not self.training else self.attention_dropout,
+            scaling=self.scaling,
+            **kwargs,
+        )
+
+        if self.config._attn_implementation == "flash_attention_2" and self.qk_head_dim != self.v_head_dim:
+            attn_output = attn_output[:, :, :, : self.v_head_dim]
+
+        attn_output = attn_output.reshape(batch_size, seq_length, -1).contiguous()
+        attn_output = self.o_proj(attn_output)
+        return attn_output, attn_weights
+
+
+class DeepseekV3DecoderLayer(nn.Module):
+    def __init__(self, config: DeepseekV3Config, layer_idx: int):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+
+        self.self_attn = DeepseekV3Attention(config=config, layer_idx=layer_idx)
+
+        if layer_idx >= config.first_k_dense_replace:
+            self.mlp = DeepseekV3MoE(config)
+        else:
+            self.mlp = DeepseekV3MLP(config)
+
+        self.input_layernorm = DeepseekV3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.post_attention_layernorm = DeepseekV3RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        attention_mask: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.LongTensor] = None,
+        past_key_value: Optional[Cache] = None,
+        output_attentions: Optional[bool] = False,
+        use_cache: Optional[bool] = False,
+        cache_position: Optional[torch.LongTensor] = None,
+        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # necessary, but kept here for BC
+        **kwargs: Unpack[FlashAttentionKwargs],
+    ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:
+        residual = hidden_states
+
+        hidden_states = self.input_layernorm(hidden_states)
+
+        # Self Attention
+        hidden_states, self_attn_weights = self.self_attn(
+            hidden_states=hidden_states,
+            attention_mask=attention_mask,
+            position_ids=position_ids,
+            past_key_value=past_key_value,
+            output_attentions=output_attentions,
+            use_cache=use_cache,
+            cache_position=cache_position,
+            position_embeddings=position_embeddings,
+            **kwargs,
+        )
+        hidden_states = residual + hidden_states
+
+        # Fully Connected
+        residual = hidden_states
+        hidden_states = self.post_attention_layernorm(hidden_states)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = residual + hidden_states
+
+        outputs = (hidden_states,)
+        if output_attentions:
+            outputs += (self_attn_weights,)
+
+        return outputs
+
+
+class DeepseekV3PreTrainedModel(LlamaPreTrainedModel):
+    pass
+
+
+class DeepseekV3Model(LlamaModel):
+    pass
+    # def __init__(self, config):
+    #     super().__init__(config)
+    #     self._register_load_state_dict_pre_hook(self.load_pre_hook)
+    #     self._register_state_dict_hook(self.load_hook)
+    #     self.post_init()
+
+    # def load_pre_hook(self, state_dict, prefix, *args):
+    #     """
+    #     Weights have to be permuted for correct rope formulation. We can't do this in the weights
+    #     as every other framework already uses the `Llama` original function (which is copyrighted btw).
+    #     And I am not even sure it's better.... anyways end of my rant
+    #     """
+
+    #     def permute_for_rope(input_tensor):
+    #         """
+    #         When you go from the complex ROPE formulation to sin and cos one, you need
+    #         to permute the query and key weights (to avoid doing it on the fly)
+    #         """
+    #         n_heads, dim1, dim2 = input_tensor.shape[0], input_tensor.shape[1], input_tensor.shape[2]
+    #         input_tensor = input_tensor.reshape(n_heads * dim1, dim2)
+    #         input_tensor = input_tensor.view(n_heads, dim1 // 2, 2, dim2)
+    #         input_tensor = input_tensor.transpose(1, 2).reshape(n_heads, dim1, dim2)
+    #         return input_tensor
+
+    #     def permute_layer_for_rope(key, num_heads, head_dim, rope_dim):
+    #         weight = state_dict[key]
+    #         weight = weight.view(num_heads, head_dim, -1)
+    #         weight_rot = weight[:, -rope_dim:]
+    #         weight_rot = permute_for_rope(weight_rot)
+    #         weight[:, -rope_dim:] = weight_rot
+    #         weight = weight.view(-1, weight.shape[-1])
+    #         state_dict[key] = weight
+
+    #     for k in state_dict:
+    #         if "q_b_proj." in k:
+    #             permute_layer_for_rope(
+    #                 k,
+    #                 num_heads=self.config.num_attention_heads,
+    #                 head_dim=self.config.qk_head_dim,
+    #                 rope_dim=self.config.qk_rope_head_dim,
+    #             )
+    #         if "kv_a_proj_with_mqa." in k:
+    #             permute_layer_for_rope(
+    #                 k,
+    #                 num_heads=1,
+    #                 head_dim=self.config.kv_lora_rank + self.config.qk_rope_head_dim,
+    #                 rope_dim=self.config.qk_rope_head_dim,
+    #             )
+
+    # def load_hook(self, module, state_dict, prefix, *args):
+    #     self.load_pre_hook(state_dict, prefix, *args)
+
+
+class DeepseekV3ForCausalLM(LlamaForCausalLM):
+    pass
+
+
+__all__ = [
+    "DeepseekV3PreTrainedModel",
+    "DeepseekV3Model",
+    "DeepseekV3ForCausalLM",
+]
diff --git a/src/transformers/utils/dummy_pt_objects.py b/src/transformers/utils/dummy_pt_objects.py
index a13e63ce99c2..1f005fb4fe11 100644
--- a/src/transformers/utils/dummy_pt_objects.py
+++ b/src/transformers/utils/dummy_pt_objects.py
@@ -2783,6 +2783,27 @@ def __init__(self, *args, **kwargs):
         requires_backends(self, ["torch"])
 
 
+class DeepseekV3ForCausalLM(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
+class DeepseekV3Model(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
+class DeepseekV3PreTrainedModel(metaclass=DummyObject):
+    _backends = ["torch"]
+
+    def __init__(self, *args, **kwargs):
+        requires_backends(self, ["torch"])
+
+
 class DeformableDetrForObjectDetection(metaclass=DummyObject):
     _backends = ["torch"]
 
diff --git a/tests/models/deepseek_v3/__init__.py b/tests/models/deepseek_v3/__init__.py
new file mode 100644
index 000000000000..e69de29bb2d1
diff --git a/tests/models/deepseek_v3/test_modeling_deepseek_v3.py b/tests/models/deepseek_v3/test_modeling_deepseek_v3.py
new file mode 100644
index 000000000000..c5efee3f9a6b
--- /dev/null
+++ b/tests/models/deepseek_v3/test_modeling_deepseek_v3.py
@@ -0,0 +1,669 @@
+# coding=utf-8
+# Copyright 2024 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.
+"""Testing suite for the PyTorch DeepseekV3 model."""
+
+import unittest
+
+from packaging import version
+from parameterized import parameterized
+
+from transformers import AutoTokenizer, DeepseekV3Config, is_torch_available, set_seed
+from transformers.testing_utils import (
+    require_read_token,
+    require_torch,
+    require_torch_accelerator,
+    require_torch_sdpa,
+    slow,
+    torch_device,
+)
+
+from ...generation.test_utils import GenerationTesterMixin
+from ...test_configuration_common import ConfigTester
+from ...test_modeling_common import ModelTesterMixin, ids_tensor
+from ...test_pipeline_mixin import PipelineTesterMixin
+
+
+if is_torch_available():
+    import torch
+
+    from transformers import (
+        DeepseekV3ForCausalLM,
+        DeepseekV3Model,
+    )
+    from transformers.models.deepseek_v3.modeling_deepseek_v3 import (
+        DeepseekV3RotaryEmbedding,
+    )
+
+
+class DeepseekV3ModelTester:
+    def __init__(
+        self,
+        parent,
+        batch_size=13,
+        seq_length=7,
+        is_training=True,
+        use_input_mask=True,
+        use_token_type_ids=False,
+        use_labels=True,
+        vocab_size=99,
+        hidden_size=32,
+        intermediate_size=37,
+        moe_intermediate_size=12,
+        num_hidden_layers=5,
+        num_attention_heads=4,
+        num_key_value_heads=4,
+        n_shared_experts=1,
+        n_routed_experts=8,
+        routed_scaling_factor=2.5,
+        kv_lora_rank=16,
+        q_lora_rank=32,
+        qk_rope_head_dim=16,
+        v_head_dim=32,
+        qk_nope_head_dim=32,
+        n_group=2,
+        topk_group=1,
+        num_experts_per_tok=2,
+        first_k_dense_replace=2,
+        norm_topk_prob=True,
+        aux_loss_alpha=0.001,
+        hidden_act="silu",
+        max_position_embeddings=512,
+        initializer_range=0.02,
+        attention_probs_dropout_prob=0.1,
+        type_vocab_size=16,
+        type_sequence_label_size=2,
+        num_labels=3,
+        num_choices=4,
+        pad_token_id=0,
+        scope=None,
+    ):
+        self.parent = parent
+        self.batch_size = batch_size
+        self.seq_length = seq_length
+        self.is_training = is_training
+        self.use_input_mask = use_input_mask
+        self.use_token_type_ids = use_token_type_ids
+        self.use_labels = use_labels
+        self.vocab_size = vocab_size
+        self.hidden_size = hidden_size
+        self.intermediate_size = intermediate_size
+        self.moe_intermediate_size = moe_intermediate_size
+        self.num_hidden_layers = num_hidden_layers
+        self.num_attention_heads = num_attention_heads
+        self.num_key_value_heads = num_key_value_heads
+        self.n_shared_experts = n_shared_experts
+        self.n_routed_experts = n_routed_experts
+        self.routed_scaling_factor = routed_scaling_factor
+        self.kv_lora_rank = kv_lora_rank
+        self.q_lora_rank = q_lora_rank
+        self.qk_rope_head_dim = qk_rope_head_dim
+        self.v_head_dim = v_head_dim
+        self.qk_nope_head_dim = qk_nope_head_dim
+        self.n_group = n_group
+        self.topk_group = topk_group
+        self.num_experts_per_tok = num_experts_per_tok
+        self.first_k_dense_replace = first_k_dense_replace
+        self.norm_topk_prob = norm_topk_prob
+        self.aux_loss_alpha = aux_loss_alpha
+        self.hidden_act = hidden_act
+        self.max_position_embeddings = max_position_embeddings
+        self.initializer_range = initializer_range
+        self.attention_probs_dropout_prob = attention_probs_dropout_prob
+        self.type_vocab_size = type_vocab_size
+        self.type_sequence_label_size = type_sequence_label_size
+        self.num_labels = num_labels
+        self.num_choices = num_choices
+        self.pad_token_id = pad_token_id
+        self.scope = scope
+
+    def prepare_config_and_inputs(self):
+        input_ids = ids_tensor([self.batch_size, self.seq_length], self.vocab_size)
+
+        input_mask = None
+        if self.use_input_mask:
+            input_mask = torch.tril(torch.ones_like(input_ids).to(torch_device))
+
+        token_type_ids = None
+        if self.use_token_type_ids:
+            token_type_ids = ids_tensor([self.batch_size, self.seq_length], self.type_vocab_size)
+
+        sequence_labels = None
+        token_labels = None
+        choice_labels = None
+        if self.use_labels:
+            sequence_labels = ids_tensor([self.batch_size], self.type_sequence_label_size)
+            token_labels = ids_tensor([self.batch_size, self.seq_length], self.num_labels)
+            choice_labels = ids_tensor([self.batch_size], self.num_choices)
+
+        config = self.get_config()
+
+        return config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
+
+    def get_config(self):
+        return DeepseekV3Config(
+            vocab_size=self.vocab_size,
+            hidden_size=self.hidden_size,
+            intermediate_size=self.intermediate_size,
+            moe_intermediate_size=self.moe_intermediate_size,
+            num_hidden_layers=self.num_hidden_layers,
+            num_attention_heads=self.num_attention_heads,
+            num_key_value_heads=self.num_key_value_heads,
+            n_shared_experts=self.n_shared_experts,
+            n_routed_experts=self.n_routed_experts,
+            routed_scaling_factor=self.routed_scaling_factor,
+            kv_lora_rank=self.kv_lora_rank,
+            q_lora_rank=self.q_lora_rank,
+            qk_rope_head_dim=self.qk_rope_head_dim,
+            v_head_dim=self.v_head_dim,
+            qk_nope_head_dim=self.qk_nope_head_dim,
+            n_group=self.n_group,
+            topk_group=self.topk_group,
+            num_experts_per_tok=self.num_experts_per_tok,
+            first_k_dense_replace=self.first_k_dense_replace,
+            norm_topk_prob=self.norm_topk_prob,
+            aux_loss_alpha=self.aux_loss_alpha,
+            hidden_act=self.hidden_act,
+            max_position_embeddings=self.max_position_embeddings,
+            initializer_range=self.initializer_range,
+            use_cache=True,
+            pad_token_id=self.pad_token_id,
+            attention_dropout=self.attention_probs_dropout_prob,
+        )
+
+    def create_and_check_model(
+        self, config, input_ids, token_type_ids, input_mask, sequence_labels, token_labels, choice_labels
+    ):
+        model = DeepseekV3Model(config=config)
+        model.to(torch_device)
+        model.eval()
+        result = model(input_ids, attention_mask=input_mask)
+        result = model(input_ids)
+        self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size))
+
+    def create_and_check_model_as_decoder(
+        self,
+        config,
+        input_ids,
+        token_type_ids,
+        input_mask,
+        sequence_labels,
+        token_labels,
+        choice_labels,
+        encoder_hidden_states,
+        encoder_attention_mask,
+    ):
+        config.add_cross_attention = True
+        model = DeepseekV3Model(config)
+        model.to(torch_device)
+        model.eval()
+        result = model(
+            input_ids,
+            attention_mask=input_mask,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_attention_mask,
+        )
+        result = model(
+            input_ids,
+            attention_mask=input_mask,
+            encoder_hidden_states=encoder_hidden_states,
+        )
+        result = model(input_ids, attention_mask=input_mask)
+        self.parent.assertEqual(result.last_hidden_state.shape, (self.batch_size, self.seq_length, self.hidden_size))
+
+    def create_and_check_for_causal_lm(
+        self,
+        config,
+        input_ids,
+        token_type_ids,
+        input_mask,
+        sequence_labels,
+        token_labels,
+        choice_labels,
+        encoder_hidden_states,
+        encoder_attention_mask,
+    ):
+        model = DeepseekV3ForCausalLM(config=config)
+        model.to(torch_device)
+        model.eval()
+        result = model(input_ids, attention_mask=input_mask, labels=token_labels)
+        self.parent.assertEqual(result.logits.shape, (self.batch_size, self.seq_length, self.vocab_size))
+
+    def create_and_check_decoder_model_past_large_inputs(
+        self,
+        config,
+        input_ids,
+        token_type_ids,
+        input_mask,
+        sequence_labels,
+        token_labels,
+        choice_labels,
+        encoder_hidden_states,
+        encoder_attention_mask,
+    ):
+        config.is_decoder = True
+        config.add_cross_attention = True
+        model = DeepseekV3ForCausalLM(config=config)
+        model.to(torch_device)
+        model.eval()
+
+        # first forward pass
+        outputs = model(
+            input_ids,
+            attention_mask=input_mask,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_attention_mask,
+            use_cache=True,
+        )
+        past_key_values = outputs.past_key_values
+
+        # create hypothetical multiple next token and extent to next_input_ids
+        next_tokens = ids_tensor((self.batch_size, 3), config.vocab_size)
+        next_mask = ids_tensor((self.batch_size, 3), vocab_size=2)
+
+        # append to next input_ids and
+        next_input_ids = torch.cat([input_ids, next_tokens], dim=-1)
+        next_attention_mask = torch.cat([input_mask, next_mask], dim=-1)
+
+        output_from_no_past = model(
+            next_input_ids,
+            attention_mask=next_attention_mask,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_attention_mask,
+            output_hidden_states=True,
+        )["hidden_states"][0]
+        output_from_past = model(
+            next_tokens,
+            attention_mask=next_attention_mask,
+            encoder_hidden_states=encoder_hidden_states,
+            encoder_attention_mask=encoder_attention_mask,
+            past_key_values=past_key_values,
+            output_hidden_states=True,
+        )["hidden_states"][0]
+
+        # select random slice
+        random_slice_idx = ids_tensor((1,), output_from_past.shape[-1]).item()
+        output_from_no_past_slice = output_from_no_past[:, -3:, random_slice_idx].detach()
+        output_from_past_slice = output_from_past[:, :, random_slice_idx].detach()
+
+        self.parent.assertTrue(output_from_past_slice.shape[1] == next_tokens.shape[1])
+
+        # test that outputs are equal for slice
+        self.parent.assertTrue(torch.allclose(output_from_past_slice, output_from_no_past_slice, atol=1e-3))
+
+    def prepare_config_and_inputs_for_common(self):
+        config_and_inputs = self.prepare_config_and_inputs()
+        (
+            config,
+            input_ids,
+            token_type_ids,
+            input_mask,
+            sequence_labels,
+            token_labels,
+            choice_labels,
+        ) = config_and_inputs
+        inputs_dict = {"input_ids": input_ids, "attention_mask": input_mask}
+        return config, inputs_dict
+
+
+@require_torch
+class DeepseekV3ModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin, unittest.TestCase):
+    all_model_classes = (
+        (
+            DeepseekV3Model,
+            DeepseekV3ForCausalLM,
+        )
+        if is_torch_available()
+        else ()
+    )
+    all_generative_model_classes = (DeepseekV3ForCausalLM,) if is_torch_available() else ()
+    pipeline_model_mapping = (
+        {
+            "feature-extraction": DeepseekV3Model,
+            "text-generation": DeepseekV3ForCausalLM,
+        }
+        if is_torch_available()
+        else {}
+    )
+    test_headmasking = False
+    test_pruning = False
+    fx_compatible = False
+
+    # Need to use `0.8` instead of `0.9` for `test_cpu_offload`
+    # This is because we are hitting edge cases with the causal_mask buffer
+    model_split_percents = [0.5, 0.7, 0.8]
+
+    # used in `test_torch_compile_for_training`
+    _torch_compile_train_cls = DeepseekV3ForCausalLM if is_torch_available() else None
+
+    def setUp(self):
+        self.model_tester = DeepseekV3ModelTester(self)
+        self.config_tester = ConfigTester(self, config_class=DeepseekV3Config, hidden_size=37)
+
+    @unittest.skip("Failing because of unique cache (HybridCache)")
+    def test_model_outputs_equivalence(self, **kwargs):
+        pass
+
+    @parameterized.expand([("random",), ("same",)])
+    @unittest.skip("DeepseekV3 has HybridCache which is not compatible with assisted decoding")
+    def test_assisted_decoding_matches_greedy_search(self, assistant_type):
+        pass
+
+    @unittest.skip("DeepseekV3 has HybridCache which is not compatible with assisted decoding")
+    def test_prompt_lookup_decoding_matches_greedy_search(self, assistant_type):
+        pass
+
+    @unittest.skip("DeepseekV3 has HybridCache which is not compatible with assisted decoding")
+    def test_assisted_decoding_sample(self):
+        pass
+
+    @unittest.skip("DeepseekV3 has HybridCache which is not compatible with dola decoding")
+    def test_dola_decoding_sample(self):
+        pass
+
+    @unittest.skip("DeepseekV3 has HybridCache and doesn't support continue from past kv")
+    def test_generate_continue_from_past_key_values(self):
+        pass
+
+    @unittest.skip("DeepseekV3 has HybridCache and doesn't support low_memory generation")
+    def test_beam_search_low_memory(self):
+        pass
+
+    @unittest.skip("DeepseekV3 has HybridCache and doesn't support contrastive generation")
+    def test_contrastive_generate(self):
+        pass
+
+    @unittest.skip("DeepseekV3 has HybridCache and doesn't support contrastive generation")
+    def test_contrastive_generate_dict_outputs_use_cache(self):
+        pass
+
+    @unittest.skip("DeepseekV3 has HybridCache and doesn't support contrastive generation")
+    def test_contrastive_generate_low_memory(self):
+        pass
+
+    @unittest.skip(
+        "DeepseekV3 has HybridCache and doesn't support StaticCache. Though it could, it shouldn't support."
+    )
+    def test_generate_with_static_cache(self):
+        pass
+
+    @unittest.skip(
+        "DeepseekV3 has HybridCache and doesn't support StaticCache. Though it could, it shouldn't support."
+    )
+    def test_generate_from_inputs_embeds_with_static_cache(self):
+        pass
+
+    @unittest.skip(
+        "DeepseekV3 has HybridCache and doesn't support StaticCache. Though it could, it shouldn't support."
+    )
+    def test_generate_continue_from_inputs_embeds(self):
+        pass
+
+    @unittest.skip("DeepseekV3's eager attn/sdpa attn outputs are expected to be different")
+    def test_sdpa_equivalence(self):
+        pass
+
+    @unittest.skip("Deepseek-V3 uses MLA so it is not compatible with the standard cache format")
+    def test_beam_search_generate_dict_outputs_use_cache(self):
+        pass
+
+    @unittest.skip("Deepseek-V3 uses MLA so it is not compatible with the standard cache format")
+    def test_generate_compilation_all_outputs(self):
+        pass
+
+    @unittest.skip("Deepseek-V3 uses MLA so it is not compatible with the standard cache format")
+    def test_generate_compile_model_forward(self):
+        pass
+
+    @unittest.skip("Deepseek-V3 uses MLA so it is not compatible with the standard cache format")
+    def test_greedy_generate_dict_outputs_use_cache(self):
+        pass
+
+    def test_config(self):
+        self.config_tester.run_common_tests()
+
+    def test_model(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        self.model_tester.create_and_check_model(*config_and_inputs)
+
+    def test_model_various_embeddings(self):
+        config_and_inputs = self.model_tester.prepare_config_and_inputs()
+        for type in ["absolute", "relative_key", "relative_key_query"]:
+            config_and_inputs[0].position_embedding_type = type
+            self.model_tester.create_and_check_model(*config_and_inputs)
+
+    @parameterized.expand([("yarn",)])
+    def test_model_rope_scaling_from_config(self, scaling_type):
+        config, _ = self.model_tester.prepare_config_and_inputs_for_common()
+        short_input = ids_tensor([1, 10], config.vocab_size)
+        long_input = ids_tensor([1, int(config.max_position_embeddings * 1.5)], config.vocab_size)
+
+        set_seed(42)  # Fixed seed at init time so the two models get the same random weights
+        original_model = DeepseekV3Model(config)
+        original_model.to(torch_device)
+        original_model.eval()
+        original_short_output = original_model(short_input).last_hidden_state
+        original_long_output = original_model(long_input).last_hidden_state
+
+        set_seed(42)  # Fixed seed at init time so the two models get the same random weights
+        config.rope_scaling = {"type": scaling_type, "factor": 10.0}
+        scaled_model = DeepseekV3Model(config)
+        scaled_model.to(torch_device)
+        scaled_model.eval()
+        scaled_short_output = scaled_model(short_input).last_hidden_state
+        scaled_long_output = scaled_model(long_input).last_hidden_state
+
+        # Dynamic scaling does not change the RoPE embeddings until it receives an input longer than the original
+        # maximum sequence length, so the outputs for the short input should match.
+        if scaling_type == "dynamic":
+            torch.testing.assert_close(original_short_output, scaled_short_output, rtol=1e-5, atol=1e-5)
+        else:
+            self.assertFalse(torch.allclose(original_short_output, scaled_short_output, atol=1e-5))
+
+        # The output should be different for long inputs
+        self.assertFalse(torch.allclose(original_long_output, scaled_long_output, atol=1e-5))
+
+    def test_model_rope_scaling(self):
+        config, _ = self.model_tester.prepare_config_and_inputs_for_common()
+        scaling_factor = 10
+        short_input_length = 10
+        long_input_length = int(config.max_position_embeddings * 1.5)
+
+        # Inputs
+        x = torch.randn(1, dtype=torch.float32, device=torch_device)  # used exlusively to get the dtype and the device
+        position_ids_short = torch.arange(short_input_length, dtype=torch.long, device=torch_device)
+        position_ids_short = position_ids_short.unsqueeze(0)
+        position_ids_long = torch.arange(long_input_length, dtype=torch.long, device=torch_device)
+        position_ids_long = position_ids_long.unsqueeze(0)
+
+        # Sanity check original RoPE
+        original_rope = DeepseekV3RotaryEmbedding(config=config).to(torch_device)
+        original_cos_short, original_sin_short = original_rope(x, position_ids_short)
+        original_cos_long, original_sin_long = original_rope(x, position_ids_long)
+        torch.testing.assert_close(original_cos_short, original_cos_long[:, :short_input_length, :])
+        torch.testing.assert_close(original_sin_short, original_sin_long[:, :short_input_length, :])
+
+        # Sanity check linear RoPE scaling
+        # New position "x" should match original position with index "x/scaling_factor"
+        config.rope_scaling = {"type": "linear", "factor": scaling_factor}
+        linear_scaling_rope = DeepseekV3RotaryEmbedding(config=config).to(torch_device)
+        linear_cos_short, linear_sin_short = linear_scaling_rope(x, position_ids_short)
+        linear_cos_long, linear_sin_long = linear_scaling_rope(x, position_ids_long)
+        torch.testing.assert_close(linear_cos_short, linear_cos_long[:, :short_input_length, :])
+        torch.testing.assert_close(linear_sin_short, linear_sin_long[:, :short_input_length, :])
+        for new_position in range(0, long_input_length, scaling_factor):
+            original_position = int(new_position // scaling_factor)
+            torch.testing.assert_close(linear_cos_long[:, new_position, :], original_cos_long[:, original_position, :])
+            torch.testing.assert_close(linear_sin_long[:, new_position, :], original_sin_long[:, original_position, :])
+
+        # Sanity check Dynamic NTK RoPE scaling
+        # Scaling should only be observed after a long input is fed. We can observe that the frequencies increase
+        # with scaling_factor (or that `inv_freq` decreases)
+        config.rope_scaling = {"type": "dynamic", "factor": scaling_factor}
+        ntk_scaling_rope = DeepseekV3RotaryEmbedding(config=config).to(torch_device)
+        ntk_cos_short, ntk_sin_short = ntk_scaling_rope(x, position_ids_short)
+        ntk_cos_long, ntk_sin_long = ntk_scaling_rope(x, position_ids_long)
+        torch.testing.assert_close(ntk_cos_short, original_cos_short)
+        torch.testing.assert_close(ntk_sin_short, original_sin_short)
+        with self.assertRaises(AssertionError):
+            torch.testing.assert_close(ntk_cos_long, original_cos_long)
+        with self.assertRaises(AssertionError):
+            torch.testing.assert_close(ntk_sin_long, original_sin_long)
+        self.assertTrue((ntk_scaling_rope.inv_freq <= original_rope.inv_freq).all())
+
+        # Sanity check Yarn RoPE scaling
+        # Scaling should be over the entire input
+        config.rope_scaling = {"type": "yarn", "factor": scaling_factor}
+        yarn_scaling_rope = DeepseekV3RotaryEmbedding(config=config).to(torch_device)
+        yarn_cos_short, yarn_sin_short = yarn_scaling_rope(x, position_ids_short)
+        yarn_cos_long, yarn_sin_long = yarn_scaling_rope(x, position_ids_long)
+        torch.testing.assert_close(yarn_cos_short, yarn_cos_long[:, :short_input_length, :])
+        torch.testing.assert_close(yarn_sin_short, yarn_sin_long[:, :short_input_length, :])
+        with self.assertRaises(AssertionError):
+            torch.testing.assert_close(yarn_cos_short, original_cos_short)
+        with self.assertRaises(AssertionError):
+            torch.testing.assert_close(yarn_sin_short, original_sin_short)
+        with self.assertRaises(AssertionError):
+            torch.testing.assert_close(yarn_cos_long, original_cos_long)
+        with self.assertRaises(AssertionError):
+            torch.testing.assert_close(yarn_sin_long, original_sin_long)
+
+    @unittest.skip(reason="Deepseek-V3 uses MLA on all models so the KV cache is a non standard format")
+    def test_past_key_values_format(self):
+        pass
+
+    @require_torch_sdpa
+    @slow
+    def test_eager_matches_sdpa_generate(self):
+        """
+        Overwritting the common test as the test is flaky on tiny models
+        """
+        max_new_tokens = 30
+
+        tokenizer = AutoTokenizer.from_pretrained("bzantium/tiny-deepseek-v3")
+
+        model_sdpa = DeepseekV3ForCausalLM.from_pretrained(
+            "bzantium/tiny-deepseek-v3",
+            torch_dtype=torch.float16,
+            low_cpu_mem_usage=True,
+        ).to(torch_device)
+
+        self.assertTrue(model_sdpa.config._attn_implementation == "sdpa")
+
+        model_eager = DeepseekV3ForCausalLM.from_pretrained(
+            "bzantium/tiny-deepseek-v3",
+            torch_dtype=torch.float16,
+            low_cpu_mem_usage=True,
+            attn_implementation="eager",
+        ).to(torch_device)
+
+        self.assertTrue(model_eager.config._attn_implementation == "eager")
+
+        for name, submodule in model_eager.named_modules():
+            if "SdpaAttention" in submodule.__class__.__name__:
+                raise ValueError("The eager model should not have SDPA attention layers")
+
+        has_sdpa = False
+        for name, submodule in model_sdpa.named_modules():
+            if "SdpaAttention" in submodule.__class__.__name__:
+                has_sdpa = True
+                break
+        if not has_sdpa:
+            raise ValueError("The SDPA model should have SDPA attention layers")
+
+        texts = [
+            "hi here's a longer context, getting longer and",
+            "Hello this is a very long sentence my friend, very long for real",
+            "Today I am in Paris and",
+        ]
+
+        for padding_side in ["left", "right"]:
+            tokenizer.padding_side = padding_side
+            tokenizer.pad_token = tokenizer.eos_token
+
+            inputs = tokenizer(texts, return_tensors="pt", padding=True).to(torch_device)
+
+            res_eager = model_eager.generate(**inputs, max_new_tokens=max_new_tokens, do_sample=False)
+            res_sdpa = model_sdpa.generate(**inputs, max_new_tokens=max_new_tokens, do_sample=False)
+
+            with self.subTest(f"{padding_side}"):
+                torch.testing.assert_close(
+                    res_eager,
+                    res_sdpa,
+                    msg=f"\n{tokenizer.batch_decode(res_eager)} \nvs\n{tokenizer.batch_decode(res_sdpa)}",
+                )
+
+
+@require_torch_accelerator
+class DeepseekV3IntegrationTest(unittest.TestCase):
+    # This variable is used to determine which CUDA device are we using for our runners (A10 or T4)
+    # Depending on the hardware we get different logits / generations
+    cuda_compute_capability_major_version = None
+
+    @classmethod
+    def setUpClass(cls):
+        if is_torch_available() and torch.cuda.is_available():
+            # 8 is for A100 / A10 and 7 for T4
+            cls.cuda_compute_capability_major_version = torch.cuda.get_device_capability()[0]
+
+    @slow
+    @require_torch_accelerator
+    @require_read_token
+    def test_compile_static_cache(self):
+        # `torch==2.2` will throw an error on this test (as in other compilation tests), but torch==2.1.2 and torch>2.2
+        # work as intended. See https://github.com/pytorch/pytorch/issues/121943
+        if version.parse(torch.__version__) < version.parse("2.3.0"):
+            self.skipTest(reason="This test requires torch >= 2.3 to run.")
+
+        NUM_TOKENS_TO_GENERATE = 40
+        # Note on `EXPECTED_TEXT_COMPLETION`'s diff: the current value matches the original test if the original test
+        # was changed to have a cache of 53 tokens (as opposed to 4096), on Ampere GPUs.
+        EXPECTED_TEXT_COMPLETION = [
+            "Simply put, the theory of relativity states that 1) the speed of light is constant in all inertial "
+            "reference frames, and 2) the laws of physics are the same for all inertial reference frames.\nThe "
+            "theory of relativ",
+            "My favorite all time favorite condiment is ketchup. I love it on everything. I love it on my eggs, "
+            "my fries, my chicken, my burgers, my hot dogs, my sandwiches, my salads, my p",
+        ]
+
+        prompts = [
+            "Simply put, the theory of relativity states that ",
+            "My favorite all time favorite condiment is ketchup.",
+        ]
+        tokenizer = AutoTokenizer.from_pretrained("bzantium/tiny-deepseek-v3", pad_token="</s>", padding_side="right")
+        model = DeepseekV3ForCausalLM.from_pretrained(
+            "bzantium/tiny-deepseek-v3", device_map=torch_device, torch_dtype=torch.float16
+        )
+        inputs = tokenizer(prompts, return_tensors="pt", padding=True).to(model.device)
+
+        # Dynamic Cache
+        generated_ids = model.generate(**inputs, max_new_tokens=NUM_TOKENS_TO_GENERATE, do_sample=False)
+        dynamic_text = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)
+        self.assertEqual(EXPECTED_TEXT_COMPLETION, dynamic_text)
+
+        # Static Cache
+        generated_ids = model.generate(
+            **inputs, max_new_tokens=NUM_TOKENS_TO_GENERATE, do_sample=False, cache_implementation="static"
+        )
+        static_text = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)
+        self.assertEqual(EXPECTED_TEXT_COMPLETION, static_text)
+
+        # Static Cache + compile
+        model._cache = None  # clear cache object, initialized when we pass `cache_implementation="static"`
+        model.forward = torch.compile(model.forward, mode="reduce-overhead", fullgraph=True)
+        generated_ids = model.generate(
+            **inputs, max_new_tokens=NUM_TOKENS_TO_GENERATE, do_sample=False, cache_implementation="static"
+        )
+        static_compiled_text = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)
+        self.assertEqual(EXPECTED_TEXT_COMPLETION, static_compiled_text)