Upload 8 files

README.md

@@ -0,0 +1,199 @@
+---
+library_name: transformers
+tags: []
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+This is the model card of a 🤗 transformers model that has been pushed on the Hub. This model card has been automatically generated.
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]

config.json

@@ -0,0 +1,77 @@
+{
+  "architectures": [
+    "Mistral4ForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "auto_map": {
+    "AutoConfig": "configuration_mistral4.Mistral4Config",
+    "AutoModelForCausalLM": "modeling_mistral4.Mistral4ForCausalLM"
+  },
+  "bos_token_id": 1,
+  "dtype": "float32",
+  "eos_token_id": 2,
+  "first_k_dense_replace": 0,
+  "head_dim": 24,
+  "hidden_act": "silu",
+  "hidden_size": 64,
+  "initializer_range": 0.02,
+  "intermediate_size": 32,
+  "kv_lora_rank": 16,
+  "max_position_embeddings": 1048576,
+  "mlp_bias": false,
+  "model_type": "mistral4",
+  "moe_intermediate_size": 16,
+  "n_group": 1,
+  "n_routed_experts": 4,
+  "n_shared_experts": 1,
+  "norm_topk_prob": true,
+  "num_attention_heads": 4,
+  "num_experts_per_tok": 2,
+  "num_hidden_layers": 3,
+  "num_key_value_heads": 4,
+  "pad_token_id": 11,
+  "pretraining_tp": 1,
+  "q_lora_rank": 32,
+  "qk_head_dim": 24,
+  "qk_nope_head_dim": 16,
+  "qk_rope_head_dim": 8,
+  "rms_norm_eps": 1e-06,
+  "rope_interleave": true,
+  "rope_parameters": {
+    "beta_fast": 32.0,
+    "beta_slow": 1.0,
+    "factor": 128.0,
+    "llama_4_scaling_beta": 0.1,
+    "max_position_embeddings": 1048576,
+    "mscale": 1.0,
+    "mscale_all_dim": 1.0,
+    "original_max_position_embeddings": 8192,
+    "partial_rotary_factor": 0.3333333333333333,
+    "rope_theta": 10000.0,
+    "rope_type": "yarn",
+    "type": "yarn"
+  },
+  "rope_scaling": {
+    "beta_fast": 32.0,
+    "beta_slow": 1.0,
+    "factor": 128.0,
+    "llama_4_scaling_beta": 0.1,
+    "max_position_embeddings": 1048576,
+    "mscale": 1.0,
+    "mscale_all_dim": 1.0,
+    "original_max_position_embeddings": 8192,
+    "partial_rotary_factor": 0.3333333333333333,
+    "rope_theta": 10000.0,
+    "rope_type": "yarn",
+    "type": "yarn"
+  },
+  "rope_theta": 10000.0,
+  "routed_scaling_factor": 1.0,
+  "tie_word_embeddings": false,
+  "topk_group": 1,
+  "transformers_version": "4.57.1",
+  "use_cache": true,
+  "v_head_dim": 16,
+  "vocab_size": 32000
+}

configuration_mistral4.py

@@ -0,0 +1,162 @@
+from transformers.configuration_utils import PretrainedConfig
+
+class Mistral4Config(PretrainedConfig):
+    r"""
+    n_group (`int`, *optional*, defaults to 1):
+        Number of groups for routed experts.
+    first_k_dense_replace (`int`, *optional*, defaults to 0):
+        Number of dense layers in shallow layers(embed->dense->dense->...->dense->moe->moe...->lm_head).
+                                                        \--k dense layers--/
+    rope_interleave (`bool`, *optional*, defaults to `True`):
+        Whether to interleave the rotary position embeddings.
+
+    Example:
+
+    ```python
+    >>> from transformers import Mistral4Model, Mistral4Config
+
+    >>> # Initializing a Mistral4 style configuration
+    >>> configuration = Mistral4Config()
+
+    >>> # Accessing the model configuration
+    >>> configuration = model.config
+    ```"""
+
+    model_type = "mistral4"
+    keys_to_ignore_at_inference = ["past_key_values"]
+    base_model_tp_plan = {
+        "layers.*.mlp.experts.gate_up_proj": "packed_colwise",
+        "layers.*.mlp.experts.down_proj": "rowwise",
+        "layers.*.mlp.experts": "moe_tp_experts",
+        "layers.*.mlp.shared_experts.gate_proj": "colwise",
+        "layers.*.mlp.shared_experts.up_proj": "colwise",
+        "layers.*.mlp.shared_experts.down_proj": "rowwise",
+        "layers.*.mlp.gate_proj": "colwise",
+        "layers.*.mlp.up_proj": "colwise",
+        "layers.*.mlp.down_proj": "rowwise",
+    }
+    base_model_pp_plan = {
+        "embed_tokens": (["input_ids"], ["inputs_embeds"]),
+        "layers": (["hidden_states", "attention_mask"], ["hidden_states"]),
+        "norm": (["hidden_states"], ["hidden_states"]),
+    }
+    attribute_map = {
+        "num_local_experts": "n_routed_experts",
+    }
+
+    def __init__(
+        self,
+        vocab_size: int = 131072,
+        hidden_size: int = 4096,
+        intermediate_size: int = 12288,
+        moe_intermediate_size: int = 2048,
+        num_hidden_layers: int = 36,
+        num_attention_heads: int = 32,
+        num_key_value_heads: int | None = 32,
+        n_shared_experts: int = 1,
+        n_routed_experts: int = 128,
+        routed_scaling_factor: float = 1.0,
+        kv_lora_rank: int = 256,
+        q_lora_rank: int | None = 1024,
+        qk_rope_head_dim: int = 64,
+        v_head_dim: int | None = 128,
+        qk_nope_head_dim: int = 64,
+        n_group: int | None = 1,
+        topk_group: int | None = 1,
+        num_experts_per_tok: int | None = 4,
+        first_k_dense_replace: int | None = 0,
+        norm_topk_prob: bool | None = True,
+        hidden_act: str = "silu",
+        max_position_embeddings: int = 1048576,
+        initializer_range: float = 0.02,
+        rms_norm_eps: float = 1e-6,
+        use_cache: bool = True,
+        pad_token_id: int | None = 11,
+        bos_token_id: int | None = 1,
+        eos_token_id: int | list[int] | None = 2,
+        pretraining_tp: int | None = 1,
+        tie_word_embeddings: bool = False,
+        rope_parameters: dict | None = None,
+        rope_interleave: bool | None = True,
+        attention_bias: bool = False,
+        attention_dropout: float | int | None = 0.0,
+        mlp_bias: bool = False,
+        **kwargs
+    ):
+        super().__init__(**kwargs)
+        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.hidden_act = hidden_act
+        self.max_position_embeddings = max_position_embeddings
+        self.initializer_range = initializer_range
+        self.rms_norm_eps = rms_norm_eps
+        self.use_cache = use_cache
+        self.pad_token_id = pad_token_id
+        self.bos_token_id = bos_token_id
+        self.eos_token_id = eos_token_id
+        self.pretraining_tp = pretraining_tp
+        self.mlp_bias = mlp_bias
+        if rope_parameters is None:
+            rope_parameters = {
+                "type": "yarn",
+                "rope_theta": 10000.0,
+                "factor": 128.0,
+                "original_max_position_embeddings": 8192,
+                "max_position_embeddings": self.max_position_embeddings,
+                "beta_fast": 32.0,
+                "beta_slow": 1.0,
+                "mscale_all_dim": 1.0,
+                "mscale": 1.0,
+                "llama_4_scaling_beta": 0.1,
+                "partial_rotary_factor": self.qk_rope_head_dim / (self.qk_nope_head_dim + self.qk_rope_head_dim),
+            }
+        if "partial_rotary_factor" not in rope_parameters:
+            rope_parameters["partial_rotary_factor"] = self.qk_rope_head_dim / (self.qk_nope_head_dim + self.qk_rope_head_dim)
+        self.rope_parameters = rope_parameters
+        self.rope_theta = rope_parameters["rope_theta"]
+        self.rope_scaling = rope_parameters
+        self.rope_interleave = rope_interleave
+        self.attention_bias = attention_bias
+        self.attention_dropout = attention_dropout
+        self.tie_word_embeddings = tie_word_embeddings
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+__all__ = ["Mistral4Config"]

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "pad_token_id": 11,
+  "transformers_version": "4.57.1"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bfdf912f19d09e9fbd44cafdfef1981e2d65945f39aa0c66de89c6aae128d4fb
+size 16880152

modeling_mistral4.py

@@ -0,0 +1,465 @@
+from collections.abc import Callable
+
+import torch
+import torch.nn.functional as F
+from torch import nn
+
+from transformers.cache_utils import Cache, DynamicCache
+from transformers.generation import GenerationMixin
+from transformers.modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast
+from transformers.activations import ACT2FN
+from transformers.modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
+from transformers.utils import logging
+from transformers.models.deepseek_v3.modeling_deepseek_v3 import (
+    DeepseekV3Attention,
+    DeepseekV3DecoderLayer,
+    apply_rotary_pos_emb_interleave,
+)
+from transformers.models.llama.modeling_llama import (
+    LlamaRMSNorm,
+    LlamaRotaryEmbedding,
+    apply_rotary_pos_emb,
+    eager_attention_forward,
+)
+from transformers.masking_utils import create_causal_mask
+from transformers.models.gemma.modeling_gemma import GemmaMLP
+from .configuration_mistral4 import Mistral4Config
+
+
+logger = logging.get_logger(__name__)
+
+
+def get_llama_4_attn_scale(positions_ids: torch.Tensor, beta: float, max_position_embeddings: int) -> torch.Tensor:
+    scaling = 1 + beta * torch.log(1 + torch.floor(positions_ids / max_position_embeddings))
+    return scaling[:, None, :, None]
+
+class Mistral4RMSNorm(LlamaRMSNorm):
+    pass
+
+
+class Mistral4RotaryEmbedding(LlamaRotaryEmbedding):
+    pass
+
+
+class Mistral4MLP(GemmaMLP):
+    def __init__(self, config, intermediate_size=None):
+        super().__init__(config)
+        self.config = config
+        self.hidden_size = config.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]
+
+
+class Mistral4TopkRouter(nn.Module):
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        self.n_routed_experts = config.n_routed_experts
+
+        self.weight = nn.Parameter(torch.empty((self.n_routed_experts, config.hidden_size)))
+
+    def forward(self, hidden_states):
+        hidden_states = hidden_states.view(-1, self.config.hidden_size)
+        router_logits = F.linear(hidden_states, self.weight)
+        return router_logits
+
+
+class Mistral4NaiveMoe(nn.Module):
+    """Collection of expert weights stored as 3D tensors."""
+
+    def __init__(self, config):
+        super().__init__()
+        self.num_experts = config.num_local_experts
+        self.hidden_dim = config.hidden_size
+        self.intermediate_dim = config.intermediate_size
+        self.gate_up_proj = nn.Parameter(torch.empty(self.num_experts, 2 * self.intermediate_dim, self.hidden_dim))
+        self.down_proj = nn.Parameter(torch.empty(self.num_experts, self.hidden_dim, self.intermediate_dim))
+        self.act_fn = ACT2FN[config.hidden_act]
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        top_k_index: torch.Tensor,
+        top_k_weights: torch.Tensor,
+    ) -> torch.Tensor:
+        final_hidden_states = torch.zeros_like(hidden_states)
+        with torch.no_grad():
+            expert_mask = torch.nn.functional.one_hot(top_k_index, num_classes=self.num_experts)
+            expert_mask = expert_mask.permute(2, 1, 0)
+            expert_hit = torch.greater(expert_mask.sum(dim=(-1, -2)), 0).nonzero()
+
+        for expert_idx in expert_hit:
+            expert_idx = expert_idx[0]
+            if expert_idx == self.num_experts:
+                continue
+            top_k_pos, token_idx = torch.where(expert_mask[expert_idx])
+            current_state = hidden_states[token_idx]
+            gate, up = nn.functional.linear(current_state, self.gate_up_proj[expert_idx]).chunk(2, dim=-1)
+            current_hidden_states = self.act_fn(gate) * up
+            current_hidden_states = nn.functional.linear(current_hidden_states, self.down_proj[expert_idx])
+            current_hidden_states = current_hidden_states * top_k_weights[token_idx, top_k_pos, None]
+            final_hidden_states.index_add_(0, token_idx, current_hidden_states.to(final_hidden_states.dtype))
+
+        return final_hidden_states
+
+
+class Mistral4MoE(nn.Module):
+    """
+    A mixed expert module containing shared experts.
+    """
+
+    def __init__(self, config):
+        super().__init__()
+        self.config = config
+        self.experts = Mistral4NaiveMoe(config)
+        self.gate = Mistral4TopkRouter(config)
+        self.shared_experts = Mistral4MLP(
+            config=config, intermediate_size=config.moe_intermediate_size * config.n_shared_experts
+        )
+        self.n_routed_experts = config.n_routed_experts
+        self.n_group = config.n_group
+        self.topk_group = config.topk_group
+        self.norm_topk_prob = config.norm_topk_prob
+        self.routed_scaling_factor = config.routed_scaling_factor
+        self.top_k = config.num_experts_per_tok
+
+    def route_tokens_to_experts(self, router_logits: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:
+        router_logits = router_logits.softmax(-1)
+        group_scores = (
+            router_logits.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 = router_logits.masked_fill(~score_mask.bool(), 0.0)
+        topk_indices = torch.topk(scores_for_choice, k=self.top_k, dim=-1, sorted=False)[1]
+        topk_weights = router_logits.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 forward(self, hidden_states):
+        residuals = hidden_states
+        orig_shape = hidden_states.shape
+        router_logits = self.gate(hidden_states)
+        topk_indices, topk_weights = self.route_tokens_to_experts(router_logits)
+        hidden_states = hidden_states.view(-1, hidden_states.shape[-1])
+        hidden_states = self.experts(hidden_states, topk_indices, topk_weights).view(*orig_shape)
+        hidden_states = hidden_states + self.shared_experts(residuals)
+        return hidden_states
+
+
+class Mistral4Attention(DeepseekV3Attention):
+    def __init__(self, config: Mistral4Config, layer_idx: int):
+        nn.Module.__init__(self)
+        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.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
+        if self.q_lora_rank is None:
+            self.q_proj = nn.Linear(config.hidden_size, self.num_heads * self.qk_head_dim, bias=False)
+        else:
+            self.q_a_proj = nn.Linear(config.hidden_size, config.q_lora_rank, bias=config.attention_bias)
+            self.q_a_layernorm = Mistral4RMSNorm(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 = Mistral4RMSNorm(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)
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        position_embeddings: tuple[torch.Tensor, torch.Tensor],
+        attention_mask: torch.Tensor | None,
+        position_ids: torch.Tensor,
+        past_key_values: Cache | None = None,
+        **kwargs,
+    ) -> tuple[torch.Tensor, torch.Tensor | None, tuple[torch.Tensor] | None]:
+        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)
+
+        if self.q_lora_rank is None:
+            q_states = self.q_proj(hidden_states)
+        else:
+            q_states = self.q_b_proj(self.q_a_layernorm(self.q_a_proj(hidden_states)))
+        q_states = q_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:  # support using interleaved weights for efficiency
+            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)
+
+        query_states = query_states * get_llama_4_attn_scale(
+            position_ids,
+            self.config.rope_parameters.get("llama_4_scaling_beta"),
+            self.config.rope_parameters.get("original_max_position_embeddings"),
+        ).to(query_states.dtype)
+
+        if past_key_values is not None:
+            key_states, value_states = past_key_values.update(key_states, value_states, self.layer_idx)
+
+        attention_interface: Callable = ALL_ATTENTION_FUNCTIONS.get_interface(
+            self.config._attn_implementation, eager_attention_forward
+        )
+
+        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,
+        )
+
+        attn_output = attn_output.reshape(batch_size, seq_length, -1).contiguous()
+        attn_output = self.o_proj(attn_output)
+        return attn_output, attn_weights
+
+
+class Mistral4DecoderLayer(DeepseekV3DecoderLayer):
+    def __init__(self, config: Mistral4Config, layer_idx: int):
+        nn.Module.__init__(self)
+        self.hidden_size = config.hidden_size
+
+        self.self_attn = Mistral4Attention(config=config, layer_idx=layer_idx)
+
+        if layer_idx >= config.first_k_dense_replace:
+            self.mlp = Mistral4MoE(config)
+        else:
+            self.mlp = Mistral4MLP(config)
+
+        self.input_layernorm = Mistral4RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.post_attention_layernorm = Mistral4RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+
+class Mistral4PreTrainedModel(PreTrainedModel):
+    config: Mistral4Config
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["Mistral4DecoderLayer"]
+    _skip_keys_device_placement = ["past_key_values"]
+    _supports_flash_attn = True
+    _supports_sdpa = True
+    _supports_flex_attn = True
+
+    _can_compile_fullgraph = True
+    _supports_attention_backend = True
+    _can_record_outputs = {
+        "hidden_states": Mistral4DecoderLayer,
+        "attentions": Mistral4Attention,
+    }
+    _keep_in_fp32_modules_strict = []
+    _keys_to_ignore_on_load_unexpected = []
+
+    @torch.no_grad()
+    def _init_weights(self, module):
+        super()._init_weights(module)
+        if isinstance(module, Mistral4TopkRouter):
+            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
+        elif isinstance(module, Mistral4NaiveMoe):
+            module.gate_up_proj.data.normal_(mean=0.0, std=self.config.initializer_range)
+            module.down_proj.normal_(mean=0.0, std=self.config.initializer_range)
+
+
+class Mistral4Model(Mistral4PreTrainedModel):
+    def __init__(self, config: Mistral4Config):
+        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(
+            [Mistral4DecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
+        )
+        self.norm = Mistral4RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.rotary_emb = Mistral4RotaryEmbedding(config=config)
+        self.gradient_checkpointing = False
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def forward(
+        self,
+        input_ids: torch.LongTensor | None = None,
+        attention_mask: torch.Tensor | None = None,
+        position_ids: torch.LongTensor | None = None,
+        past_key_values: Cache | None = None,
+        inputs_embeds: torch.FloatTensor | None = None,
+        use_cache: bool | None = None,
+        **kwargs,
+    ) :
+        if (input_ids is None) ^ (inputs_embeds is not None):
+            raise ValueError("You must specify exactly one of input_ids or inputs_embeds")
+
+        if inputs_embeds is None:
+            inputs_embeds: torch.Tensor = self.embed_tokens(input_ids)
+
+        if use_cache and past_key_values is None:
+            past_key_values = DynamicCache(config=self.config)
+
+        if position_ids is None:
+            past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
+            position_ids = torch.arange(inputs_embeds.shape[1], device=inputs_embeds.device) + past_seen_tokens
+            position_ids = position_ids.unsqueeze(0)
+
+        causal_mask = create_causal_mask(
+            config=self.config,
+            inputs_embeds=inputs_embeds,
+            attention_mask=attention_mask,
+            past_key_values=past_key_values,
+            position_ids=position_ids,
+        )
+
+        hidden_states = inputs_embeds
+        position_embeddings = self.rotary_emb(hidden_states, position_ids=position_ids)
+
+        for decoder_layer in self.layers[: self.config.num_hidden_layers]:
+            hidden_states = decoder_layer(
+                hidden_states,
+                attention_mask=causal_mask,
+                position_embeddings=position_embeddings,
+                position_ids=position_ids,
+                past_key_values=past_key_values,
+                use_cache=use_cache,
+                **kwargs,
+            )
+
+        hidden_states = self.norm(hidden_states)
+        return BaseModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=past_key_values,
+        )
+
+
+class Mistral4ForCausalLM(Mistral4PreTrainedModel, GenerationMixin):
+    _tied_weights_keys = {"lm_head.weight": "model.embed_tokens.weight"}
+    _tp_plan = {"lm_head": "colwise_gather_output"}
+    _pp_plan = {"lm_head": (["hidden_states"], ["logits"])}
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.model = Mistral4Model(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 forward(
+        self,
+        input_ids: torch.LongTensor | None = None,
+        attention_mask: torch.Tensor | None = None,
+        position_ids: torch.LongTensor | None = None,
+        past_key_values: Cache | None = None,
+        inputs_embeds: torch.FloatTensor | None = None,
+        labels: torch.LongTensor | None = None,
+        use_cache: bool | None = None,
+        logits_to_keep: int | torch.Tensor = 0,
+        **kwargs,
+    ) -> CausalLMOutputWithPast:
+        r"""
+        Example:
+
+        ```python
+        >>> from transformers import AutoTokenizer, Mistral4ForCausalLM
+
+        >>> model = Mistral4ForCausalLM.from_pretrained("meta-mistral4/Mistral4-2-7b-hf")
+        >>> tokenizer = AutoTokenizer.from_pretrained("meta-mistral4/Mistral4-2-7b-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."
+        ```"""
+        outputs: BaseModelOutputWithPast = 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,
+            **kwargs,
+        )
+
+        hidden_states = outputs.last_hidden_state
+        # 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)
+
+        return CausalLMOutputWithPast(
+            loss=loss,
+            logits=logits,
+            past_key_values=outputs.past_key_values,
+            hidden_states=outputs.hidden_states,
+            attentions=outputs.attentions,
+        )
+
+
+
+
+__all__ = [
+    "Mistral4PreTrainedModel",
+    "Mistral4Model",
+    "Mistral4ForCausalLM",
+]

tokenizer_config.json

@@ -0,0 +1,17 @@
+{
+  "add_prefix_space": null,
+  "backend": "tokenizers",
+  "tokenizer_class": "PreTrainedTokenizer",
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "extra_special_tokens": {},
+  "is_local": false,
+  "legacy": false,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "</s>",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}