fusion-bench 0.2.9__py3-none-any.whl

fusion_bench/models/modeling_smile_mistral/modeling_smile_mistral.py

@@ -0,0 +1,1034 @@
+import logging
+import math
+from typing import List, Optional, Tuple, Union
+
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+from torch import Tensor, nn
+from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
+from transformers.cache_utils import (
+    Cache,
+    DynamicCache,
+    SlidingWindowCache,
+    StaticCache,
+)
+from transformers.modeling_attn_mask_utils import AttentionMaskConverter
+from transformers.modeling_outputs import (
+    BaseModelOutputWithPast,
+    CausalLMOutputWithPast,
+    SequenceClassifierOutputWithPast,
+    TokenClassifierOutput,
+)
+from transformers.modeling_utils import PreTrainedModel
+from transformers.models.mistral.modeling_mistral import (
+    ACT2FN,
+    MistralRMSNorm,
+    MistralRotaryEmbedding,
+)
+
+from .configuration_smile_mistral import SmileMistralConfig
+
+logger = logging.getLogger(__name__)
+
+
+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)
+
+
+class SmileGate(nn.Module):
+    __constants__ = ["in_features", "num_experts", "k"]
+    in_features: int
+    num_experts: int
+    k: int
+    weight: Tensor
+
+    def __init__(
+        self,
+        in_features: int,
+        num_experts: int,
+        k: int,
+        device=None,
+        dtype=None,
+    ):
+        factory_kwargs = {"device": device, "dtype": dtype}
+        super().__init__()
+        self.input_features = in_features
+        self.num_experts = num_experts
+        self.k = k
+
+        self.weight = nn.Parameter(
+            torch.empty(num_experts * k, in_features, **factory_kwargs)
+        )
+
+    def forward(self, x: Tensor):
+        batch_size = x.size(0)
+        if self.num_experts == 1:
+            return torch.ones(batch_size, 1, device=x.device, dtype=x.dtype)
+
+        routing_weights = F.linear(x, self.weight).view(
+            batch_size, self.num_experts, self.k
+        )
+        routing_weights = routing_weights.norm(p=2, dim=2)
+        return routing_weights
+
+
+class SmileLinearExpert(nn.Module):
+    __constants__ = ["in_features", "out_features", "k"]
+    in_features: int
+    out_features: int
+    k: int
+
+    def __init__(
+        self,
+        in_features,
+        out_features,
+        k: int,
+        bias: bool,
+        device=None,
+        dtype=None,
+    ):
+        factory_kwargs = {"device": device, "dtype": dtype}
+        super().__init__()
+        self.in_features = in_features
+        self.out_features = out_features
+        self.k = k
+
+        self.u = nn.Parameter(torch.empty(out_features, k, **factory_kwargs))
+        self.svh = nn.Parameter(torch.empty(k, in_features, **factory_kwargs))
+
+        if bias:
+            self.bias = nn.Parameter(torch.empty(out_features, **factory_kwargs))
+        else:
+            self.register_parameter("bias", None)
+
+    def forward(self, x):
+        x = F.linear(x, self.svh)
+        x = F.linear(x, self.u, self.bias)
+        return x
+
+
+class SmileLinear(nn.Module):
+    @torch.no_grad()
+    def __init__(
+        self,
+        config: SmileMistralConfig,
+        in_features,
+        out_features,
+        bias: bool,
+        device=None,
+        dtype=None,
+    ):
+        factory_kwargs = {"device": device, "dtype": dtype}
+        super().__init__()
+        self.num_local_experts = config.num_local_experts
+        self.num_experts_per_tok = config.num_experts_per_tok
+        self.rank_of_expert = config.rank_of_expert
+        self.rank_of_router = config.rank_of_router
+        self.in_features = in_features
+        self.out_features = out_features
+
+        # construct the gate network
+        self.gate = SmileGate(
+            in_features=in_features,
+            num_experts=self.num_local_experts,
+            k=self.rank_of_router,
+            **factory_kwargs,
+        )
+
+        # the shared linear
+        self.shared_linear = nn.Linear(
+            in_features, out_features, bias=bias, **factory_kwargs
+        )
+
+        # construct experts
+        if self.rank_of_expert > 0:
+            self.experts = nn.ModuleList(
+                [
+                    SmileLinearExpert(
+                        in_features=in_features,
+                        out_features=out_features,
+                        bias=bias,
+                        k=self.rank_of_expert,
+                        **factory_kwargs,
+                    )
+                    for _ in range(self.num_local_experts)
+                ]
+            )
+        else:
+            self.experts = nn.ModuleList(
+                [
+                    nn.Linear(in_features, out_features, bias=bias, **factory_kwargs)
+                    for _ in range(self.num_local_experts)
+                ]
+            )
+
+    def forward(self, hidden_states: Tensor):
+        pretrained_out = self.shared_linear(hidden_states)
+
+        input_shape = hidden_states.size()
+        hidden_states = hidden_states.view(-1, self.in_features)
+
+        router_logits = self.gate(hidden_states)
+        routing_weights = F.softmax(router_logits, dim=1)
+        # sample the expert according to the routing weights
+        routing_weights, selected_experts = torch.topk(
+            routing_weights, self.num_experts_per_tok, dim=-1
+        )
+        routing_weights /= routing_weights.sum(dim=-1, keepdim=True)
+
+        final_hidden_states = torch.zeros(
+            (hidden_states.size(0), self.out_features),
+            dtype=hidden_states.dtype,
+            device=hidden_states.device,
+        )
+
+        # One hot encode the selected experts to create an expert mask
+        # this will be used to easily index which expert is going to be sollicitated
+        expert_mask = torch.nn.functional.one_hot(
+            selected_experts, num_classes=self.num_local_experts
+        ).permute(2, 1, 0)
+
+        # Loop over all available experts in the model and perform the computation on each expert
+        for expert_idx in range(self.num_local_experts):
+            expert_layer = self.experts[expert_idx]
+            idx, top_x = torch.where(expert_mask[expert_idx])
+
+            # Index the correct hidden states and compute the expert hidden state for
+            # the current expert. We need to make sure to multiply the output hidden
+            # states by `routing_weights` on the corresponding tokens (top-1 and top-2)
+            current_state = hidden_states[None, top_x].reshape(-1, self.in_features)
+            if current_state.numel() == 0:
+                continue
+            current_hidden_states = (
+                expert_layer(current_state) * routing_weights[top_x, idx, None]
+            )
+
+            # However `index_add_` only support torch tensors for indexing so we'll use
+            # the `top_x` tensor here.
+            final_hidden_states.index_add_(
+                0, top_x, current_hidden_states.to(hidden_states.dtype)
+            )
+        final_hidden_states = final_hidden_states.reshape(
+            *input_shape[:-1], self.out_features
+        )
+        final_hidden_states = pretrained_out + final_hidden_states
+        return final_hidden_states
+
+    @property
+    def weight(self):
+        """
+        Mimic linear layer. Bacause in some cases, user might indicate the device (or dtype of parameters) of the linear layer using `linear_layer.weight.device`
+        """
+        return self.shared_linear.weight
+
+    @property
+    def bias(self):
+        return self.shared_linear.bias
+
+    def __repr__(self):
+        return (
+            f"SingularMoELinear("
+            f"in_features={self.shared_linear.in_features}, "
+            f"out_features={self.shared_linear.out_features}, "
+            f"num_local_experts={self.num_local_experts}, "
+            f"num_experts_per_tok={self.num_experts_per_tok}, "
+            f"rank_of_router={self.rank_of_router}, "
+            f"rank_of_expert={self.rank_of_expert}"
+            f")"
+        )
+
+
+class SmileMistralAttention(nn.Module):
+    """
+    Multi-headed attention from 'Attention Is All You Need' paper. Modified to use sliding window attention: Longformer
+    and "Generating Long Sequences with Sparse Transformers".
+    """
+
+    def __init__(self, config: SmileMistralConfig, layer_idx: Optional[int] = None):
+        super().__init__()
+        self.config = config
+        self.layer_idx = layer_idx
+        if layer_idx is None:
+            logger.warning_once(
+                f"Instantiating {self.__class__.__name__} without passing a `layer_idx` is not recommended and will "
+                "lead to errors during the forward call if caching is used. Please make sure to provide a `layer_idx` "
+                "when creating this class."
+            )
+
+        self.attention_dropout = config.attention_dropout
+        self.hidden_size = config.hidden_size
+        self.num_heads = config.num_attention_heads
+        self.head_dim = config.head_dim
+        self.num_key_value_heads = config.num_key_value_heads
+        self.num_key_value_groups = self.num_heads // self.num_key_value_heads
+        self.max_position_embeddings = config.max_position_embeddings
+        self.rope_theta = config.rope_theta
+        self.is_causal = True
+
+        self.q_proj = SmileLinear(
+            config,
+            self.hidden_size,
+            self.num_heads * self.head_dim,
+            bias=False,
+        )
+        self.k_proj = SmileLinear(
+            config,
+            self.hidden_size,
+            self.num_key_value_heads * self.head_dim,
+            bias=False,
+        )
+        self.v_proj = SmileLinear(
+            config,
+            self.hidden_size,
+            self.num_key_value_heads * self.head_dim,
+            bias=False,
+        )
+        self.o_proj = SmileLinear(
+            config,
+            self.num_heads * self.head_dim,
+            self.hidden_size,
+            bias=False,
+        )
+
+        self.rotary_emb = MistralRotaryEmbedding(
+            self.head_dim,
+            max_position_embeddings=self.max_position_embeddings,
+            base=self.rope_theta,
+        )
+
+    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: bool = False,
+        use_cache: bool = False,
+        cache_position: Optional[torch.LongTensor] = None,
+    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
+        bsz, q_len, _ = hidden_states.size()
+
+        query_states = self.q_proj(hidden_states)
+        key_states = self.k_proj(hidden_states)
+        value_states = self.v_proj(hidden_states)
+
+        query_states = query_states.view(
+            bsz, q_len, self.num_heads, self.head_dim
+        ).transpose(1, 2)
+        key_states = key_states.view(
+            bsz, q_len, self.num_key_value_heads, self.head_dim
+        ).transpose(1, 2)
+        value_states = value_states.view(
+            bsz, q_len, self.num_key_value_heads, self.head_dim
+        ).transpose(1, 2)
+
+        cos, sin = self.rotary_emb(value_states, position_ids)
+        query_states, key_states = apply_rotary_pos_emb(
+            query_states, key_states, cos, sin
+        )
+
+        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
+            )
+
+        key_states = repeat_kv(key_states, self.num_key_value_groups)
+        value_states = repeat_kv(value_states, self.num_key_value_groups)
+
+        attn_weights = torch.matmul(
+            query_states, key_states.transpose(2, 3)
+        ) / math.sqrt(self.head_dim)
+
+        if attention_mask is not None:  # no matter the length, we just slice it
+            causal_mask = attention_mask[:, :, :, : key_states.shape[-2]]
+            attn_weights = attn_weights + causal_mask
+
+        # upcast attention to fp32
+        attn_weights = nn.functional.softmax(
+            attn_weights, dim=-1, dtype=torch.float32
+        ).to(query_states.dtype)
+        attn_weights = nn.functional.dropout(
+            attn_weights, p=self.attention_dropout, training=self.training
+        )
+        attn_output = torch.matmul(attn_weights, value_states)
+
+        if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
+            raise ValueError(
+                f"`attn_output` should be of size {(bsz, self.num_heads, q_len, self.head_dim)}, but is"
+                f" {attn_output.size()}"
+            )
+
+        attn_output = attn_output.transpose(1, 2).contiguous()
+
+        attn_output = attn_output.view(bsz, q_len, -1)
+        attn_output = self.o_proj(attn_output)
+
+        if not output_attentions:
+            attn_weights = None
+
+        return attn_output, attn_weights, past_key_value
+
+
+class SmileMistralMLP(nn.Module):
+    def __init__(self, config: SmileMistralConfig):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+        self.intermediate_size = config.intermediate_size
+        self.gate_proj = SmileLinear(
+            config,
+            in_features=self.hidden_size,
+            out_features=self.intermediate_size,
+            bias=False,
+        )
+        self.up_proj = SmileLinear(
+            config,
+            in_features=self.hidden_size,
+            out_features=self.intermediate_size,
+            bias=False,
+        )
+        self.down_proj = SmileLinear(
+            config,
+            in_features=self.intermediate_size,
+            out_features=self.hidden_size,
+            bias=False,
+        )
+        self.act_fn = ACT2FN[config.hidden_act]
+
+    def forward(self, hidden_state):
+        return self.down_proj(
+            self.act_fn(self.gate_proj(hidden_state)) * self.up_proj(hidden_state)
+        )
+
+
+SMILE_MISTRAL_ATTENTION_CLASSES = {
+    "eager": SmileMistralAttention,
+    # "flash_attention_2": MistralFlashAttention2,
+    # "sdpa": MistralSdpaAttention,
+}
+
+
+class SmileMistralDecoderLayer(nn.Module):
+    def __init__(self, config: SmileMistralConfig, layer_idx: int):
+        super().__init__()
+        self.hidden_size = config.hidden_size
+
+        self.self_attn = SMILE_MISTRAL_ATTENTION_CLASSES[config._attn_implementation](
+            config=config, layer_idx=layer_idx
+        )
+
+        self.mlp = SmileMistralMLP(config)
+        self.input_layernorm = MistralRMSNorm(
+            config.hidden_size, eps=config.rms_norm_eps
+        )
+        self.post_attention_layernorm = MistralRMSNorm(
+            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,
+        **kwargs,
+    ) -> Tuple[
+        torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]
+    ]:
+        """
+        Args:
+            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
+            attention_mask (`torch.FloatTensor`, *optional*):
+                attention mask of size `(batch_size, sequence_length)` if flash attention is used or `(batch_size, 1,
+                query_sequence_length, key_sequence_length)` if default attention is used.
+            output_attentions (`bool`, *optional*):
+                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
+                returned tensors for more detail.
+            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`).
+            past_key_value (`Tuple(torch.FloatTensor)`, *optional*): cached past key and value projection states
+            cache_position (`torch.LongTensor` of shape `(sequence_length)`, *optional*):
+                Indices depicting the position of the input sequence tokens in the sequence
+            kwargs (`dict`, *optional*):
+                Arbitrary kwargs to be ignored, used for FSDP and other methods that injects code
+                into the model
+        """
+        residual = hidden_states
+
+        hidden_states = self.input_layernorm(hidden_states)
+
+        # Self Attention
+        hidden_states, self_attn_weights, present_key_value = 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,
+            **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,)
+
+        if use_cache:
+            outputs += (present_key_value,)
+
+        return outputs
+
+
+class SmileMistralPreTrainedModel(PreTrainedModel):
+    config_class = SmileMistralConfig
+    base_model_prefix = "model"
+    supports_gradient_checkpointing = True
+    _no_split_modules = ["SmileMistralDecoderLayer"]
+    _skip_keys_device_placement = "past_key_values"
+    _supports_flash_attn_2 = False
+    _supports_sdpa = False
+    _supports_cache_class = True
+    _supports_static_cache = 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_()
+
+
+class SmileMistralModel(SmileMistralPreTrainedModel):
+    def __init__(self, config: SmileMistralConfig):
+        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(
+            [
+                SmileMistralDecoderLayer(config, layer_idx)
+                for layer_idx in range(config.num_hidden_layers)
+            ]
+        )
+        self._attn_implementation = config._attn_implementation
+        self.norm = MistralRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+        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
+
+    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,
+        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,
+    ) -> 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
+        )
+
+        # retrieve input_ids and inputs_embeds
+        if (input_ids is None) ^ (inputs_embeds is not None):
+            raise ValueError(
+                "You cannot specify both input_ids and inputs_embeds at the same time, and must specify either one"
+            )
+
+        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)
+
+        return_legacy_cache = False
+        if use_cache and not isinstance(past_key_values, Cache) and not self.training:
+            past_key_values = DynamicCache.from_legacy_cache(past_key_values)
+            return_legacy_cache = True
+            logger.warning_once(
+                "We detected that you are passing `past_key_values` as a tuple and this is deprecated and will be removed in v4.43. "
+                "Please use an appropriate `Cache` class (https://huggingface.co/docs/transformers/v4.41.3/en/internal/generation_utils#transformers.Cache)"
+            )
+
+        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,
+            use_cache,
+            output_attentions,
+        )
+
+        hidden_states = inputs_embeds
+
+        # decoder layers
+        all_hidden_states = () if output_hidden_states else None
+        all_self_attns = () if output_attentions else None
+        next_decoder_cache = None
+
+        for decoder_layer in self.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,
+                )
+            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,
+                )
+
+            hidden_states = layer_outputs[0]
+
+            if use_cache:
+                next_decoder_cache = layer_outputs[2 if output_attentions else 1]
+
+            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,)
+
+        next_cache = next_decoder_cache if use_cache else None
+        if return_legacy_cache:
+            next_cache = next_cache.to_legacy_cache()
+
+        if not return_dict:
+            return tuple(
+                v
+                for v in [hidden_states, next_cache, all_hidden_states, all_self_attns]
+                if v is not None
+            )
+        return BaseModelOutputWithPast(
+            last_hidden_state=hidden_states,
+            past_key_values=next_cache,
+            hidden_states=all_hidden_states,
+            attentions=all_self_attns,
+        )
+
+    def _update_causal_mask(
+        self,
+        attention_mask: torch.Tensor,
+        input_tensor: torch.Tensor,
+        cache_position: torch.Tensor,
+        past_key_values: Cache,
+        use_cache: bool,
+        output_attentions: bool,
+    ):
+        # TODO: As of torch==2.2.0, the `attention_mask` passed to the model in `generate` is 2D and of dynamic length even when the static
+        # KV cache is used. This is an issue for torch.compile which then recaptures cudagraphs at each decode steps due to the dynamic shapes.
+        # (`recording cudagraph tree for symint key 13`, etc.), which is VERY slow. A workaround is `@torch.compiler.disable`, but this prevents using
+        # `fullgraph=True`. See more context in https://github.com/huggingface/transformers/pull/29114
+
+        if self._attn_implementation == "flash_attention_2":
+            if attention_mask is not None and use_cache:
+                is_padding_right = (
+                    attention_mask[:, -1].sum().item() != input_tensor.size()[0]
+                )
+                if is_padding_right:
+                    raise ValueError(
+                        "You are attempting to perform batched generation with padding_side='right'"
+                        " this may lead to unexpected behaviour for Flash Attention version of Mistral. Make sure to "
+                        " call `tokenizer.padding_side  = 'left'` before tokenizing the input. "
+                    )
+            if attention_mask is not None and 0.0 in attention_mask:
+                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.
+
+        # cache_position must be valid here no matter which cache we use
+        past_seen_tokens = cache_position[0] if past_key_values is not None else 0
+        using_static_cache = isinstance(past_key_values, StaticCache)
+        using_sliding_window_cache = isinstance(past_key_values, SlidingWindowCache)
+
+        if (
+            self.config._attn_implementation == "sdpa"
+            and not (using_static_cache or using_sliding_window_cache)
+            and not output_attentions
+        ):
+            if AttentionMaskConverter._ignore_causal_mask_sdpa(
+                attention_mask,
+                inputs_embeds=input_tensor,
+                past_key_values_length=past_seen_tokens,
+                sliding_window=self.config.sliding_window,
+                is_training=self.training,
+            ):
+                return None
+
+        dtype, device = input_tensor.dtype, input_tensor.device
+        min_dtype = torch.finfo(dtype).min
+        sequence_length = input_tensor.shape[1]
+        # SlidingWindowCache
+        if using_sliding_window_cache:
+            target_length = max(sequence_length, self.config.sliding_window)
+        # StaticCache
+        elif using_static_cache:
+            target_length = past_key_values.get_max_length()
+        # DynamicCache or no cache
+        else:
+            target_length = (
+                attention_mask.shape[-1]
+                if isinstance(attention_mask, torch.Tensor)
+                else past_seen_tokens + sequence_length + 1
+            )
+
+        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
+            if attention_mask.max() != 0:
+                raise ValueError(
+                    "Custom 4D attention mask should be passed in inverted form with max==0`"
+                )
+            causal_mask = attention_mask
+        else:
+            causal_mask = torch.full(
+                (sequence_length, target_length),
+                fill_value=min_dtype,
+                dtype=dtype,
+                device=device,
+            )
+            exclude_mask = torch.arange(
+                target_length, device=device
+            ) > cache_position.reshape(-1, 1)
+            if self.config.sliding_window is not None:
+                if (
+                    not using_sliding_window_cache
+                    or sequence_length > self.config.sliding_window
+                ):
+                    exclude_mask.bitwise_or_(
+                        torch.arange(target_length, device=device)
+                        <= (cache_position.reshape(-1, 1) - self.config.sliding_window)
+                    )
+            causal_mask *= exclude_mask
+            causal_mask = causal_mask[None, None, :, :].expand(
+                input_tensor.shape[0], 1, -1, -1
+            )
+            if attention_mask is not None:
+                causal_mask = (
+                    causal_mask.clone()
+                )  # copy to contiguous memory for in-place edit
+                if attention_mask.dim() == 2:
+                    mask_length = attention_mask.shape[-1]
+                    padding_mask = (
+                        causal_mask[:, :, :, :mask_length]
+                        + attention_mask[:, None, None, :]
+                    )
+                    padding_mask = padding_mask == 0
+                    causal_mask[:, :, :, :mask_length] = causal_mask[
+                        :, :, :, :mask_length
+                    ].masked_fill(padding_mask, min_dtype)
+
+        if (
+            self.config._attn_implementation == "sdpa"
+            and attention_mask is not None
+            and attention_mask.device.type == "cuda"
+            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
+            causal_mask = AttentionMaskConverter._unmask_unattended(
+                causal_mask, min_dtype
+            )
+
+        return causal_mask
+
+
+class SmileMistralForCausalLM(SmileMistralPreTrainedModel):
+    _tied_weights_keys = ["lm_head.weight"]
+
+    def __init__(self, config):
+        super().__init__(config)
+        self.model = SmileMistralModel(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
+
+    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,
+    ) -> 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]`.
+
+        Returns:
+
+        Example:
+
+        ```python
+        >>> from transformers import AutoTokenizer, MistralForCausalLM
+
+        >>> model = MistralForCausalLM.from_pretrained("mistralai/Mistral-7B-v0.1")
+        >>> tokenizer = AutoTokenizer.from_pretrained("mistralai/Mistral-7B-v0.1")
+
+        >>> 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,
+        )
+
+        hidden_states = outputs[0]
+        logits = self.lm_head(hidden_states)
+        logits = logits.float()
+
+        loss = None
+        if labels is not None:
+            # Shift so that tokens < n predict n
+            shift_logits = logits[..., :-1, :].contiguous()
+            shift_labels = labels[..., 1:].contiguous()
+            # Flatten the tokens
+            shift_logits = shift_logits.view(-1, self.config.vocab_size)
+            shift_labels = shift_labels.view(-1)
+            # Ensure tensors are on the same device
+            shift_labels = shift_labels.to(shift_logits.device)
+            loss_fct = CrossEntropyLoss()
+            loss = loss_fct(shift_logits, shift_labels)
+
+        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,
+        )
+
+    # Copied from transformers.models.llama.modeling_llama.LlamaForCausalLM.prepare_inputs_for_generation
+    def prepare_inputs_for_generation(
+        self,
+        input_ids,
+        past_key_values=None,
+        attention_mask=None,
+        inputs_embeds=None,
+        cache_position=None,
+        position_ids=None,
+        use_cache=True,
+        **kwargs,
+    ):
+        # If we have cache: let's slice `input_ids` through `cache_position`, to keep only the unprocessed tokens
+        # Exception 1: when passing input_embeds, input_ids may be missing entries
+        # Exception 2: some generation methods do special slicing of input_ids, so we don't need to do it here
+        if past_key_values is not None:
+            if inputs_embeds is not None:  # Exception 1
+                input_ids = input_ids[:, -cache_position.shape[0] :]
+            elif (
+                input_ids.shape[1] != cache_position.shape[0]
+            ):  # Default case (the "else", a no op, is Exception 2)
+                input_ids = input_ids[:, cache_position]
+
+        if attention_mask is not None and position_ids is None:
+            # create position_ids on the fly for batch generation
+            position_ids = attention_mask.long().cumsum(-1) - 1
+            position_ids.masked_fill_(attention_mask == 0, 1)
+            if past_key_values:
+                position_ids = position_ids[:, -input_ids.shape[1] :]
+
+        # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
+        if inputs_embeds is not None and cache_position[0] == 0:
+            model_inputs = {"inputs_embeds": inputs_embeds}
+        else:
+            model_inputs = {
+                "input_ids": input_ids.contiguous()
+            }  # `contiguous()` needed for compilation use cases
+
+        model_inputs.update(
+            {
+                "position_ids": position_ids,
+                "cache_position": cache_position,
+                "past_key_values": past_key_values,
+                "use_cache": use_cache,
+                "attention_mask": attention_mask,
+            }
+        )
+        return model_inputs