modeling_mot.py

# coding=utf-8
# Copyright 2024 The OpenAI Team Authors and HuggingFace Inc. team.
# Copyright (c) 2018, NVIDIA CORPORATION.  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.
"""PyTorch MixtureOfTokens model."""

import math
import warnings
from typing import Optional, Tuple, Union

import torch
import torch.nn.functional as F
import torch.utils.checkpoint
from torch import nn
from torch.cuda.amp import autocast
from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
from torch.nn.init import trunc_normal_

from transformers.activations import ACT2FN
from transformers.modeling_outputs import (
    BaseModelOutputWithPastAndCrossAttentions,
    CausalLMOutputWithCrossAttentions,
    QuestionAnsweringModelOutput,
    SequenceClassifierOutputWithPast,
    TokenClassifierOutput,
)
from transformers.modeling_utils import PreTrainedModel
from transformers.pytorch_utils import Conv1D, find_pruneable_heads_and_indices, prune_conv1d_layer
from transformers.utils import (
    add_code_sample_docstrings,
    add_start_docstrings,
    add_start_docstrings_to_model_forward,
    logging,
)
from transformers.utils.model_parallel_utils import assert_device_map, get_device_map
from .configuration_mot import MoTConfig


logger = logging.get_logger(__name__)

_CHECKPOINT_FOR_DOC = "jaszczur/mixture_of_tokens"
_CONFIG_FOR_DOC = "MoTConfig"


def with_batch_size_alignment(forward_fn):
    def _forward(self, x):
        """assumed ordering (batch, seq_len, dmodel)"""
        size = x.size(self.sparsity_dim)
        if size % self.group_size != 0:
            if self.sparsity_dim == 1:
                x = x.transpose(0, 1)

            x = self.pad(x)

            if self.sparsity_dim == 1:
                x = forward_fn(self, x.transpose(0, 1))
                return x[:, :size, :]
            else:
                x = forward_fn(self, x)
                return x[:size, :, :]
        else:
            return forward_fn(self, x)

    return _forward


# Copied from transformers.models.gpt2.modeling_gpt2.GPT2Attention with GPT2->MoT
class MoTAttention(nn.Module):
    def __init__(self, config, is_cross_attention=False, layer_idx=None):
        super().__init__()
        self.config = config
        max_positions = config.max_position_embeddings
        self.register_buffer(
            "bias",
            torch.tril(torch.ones((max_positions, max_positions), dtype=torch.bool)).view(
                1, 1, max_positions, max_positions
            ),
            persistent=False,
        )
        self.register_buffer("masked_bias", torch.tensor(-1e4), persistent=False)

        self.embed_dim = config.hidden_size
        self.num_heads = config.num_attention_heads
        self.head_dim = self.embed_dim // self.num_heads
        self.split_size = self.embed_dim
        if self.head_dim * self.num_heads != self.embed_dim:
            raise ValueError(
                f"`embed_dim` must be divisible by num_heads (got `embed_dim`: {self.embed_dim} and `num_heads`:"
                f" {self.num_heads})."
            )

        self.scale_attn_weights = config.scale_attn_weights
        self.is_cross_attention = is_cross_attention

        # Layer-wise attention scaling, reordering, and upcasting
        self.scale_attn_by_inverse_layer_idx = config.scale_attn_by_inverse_layer_idx
        self.layer_idx = layer_idx
        self.reorder_and_upcast_attn = config.reorder_and_upcast_attn

        if self.is_cross_attention:
            self.c_attn = Conv1D(2 * self.embed_dim, self.embed_dim)
            self.q_attn = Conv1D(self.embed_dim, self.embed_dim)
        else:
            self.c_attn = Conv1D(3 * self.embed_dim, self.embed_dim)
        self.c_proj = Conv1D(self.embed_dim, self.embed_dim)

        self.attn_dropout = nn.Dropout(config.attn_pdrop)
        self.resid_dropout = nn.Dropout(config.resid_pdrop)
        self.is_causal = True

        self.pruned_heads = set()

    def prune_heads(self, heads):
        if len(heads) == 0:
            return
        heads, index = find_pruneable_heads_and_indices(heads, self.num_heads, self.head_dim, self.pruned_heads)
        index_attn = torch.cat([index, index + self.split_size, index + (2 * self.split_size)])

        # Prune conv1d layers
        self.c_attn = prune_conv1d_layer(self.c_attn, index_attn, dim=1)
        self.c_proj = prune_conv1d_layer(self.c_proj, index, dim=0)

        # Update hyper params
        self.split_size = (self.split_size // self.num_heads) * (self.num_heads - len(heads))
        self.num_heads = self.num_heads - len(heads)
        self.pruned_heads = self.pruned_heads.union(heads)

    def _attn(self, query, key, value, attention_mask=None, head_mask=None):
        attn_weights = torch.matmul(query, key.transpose(-1, -2))

        if self.scale_attn_weights:
            attn_weights = attn_weights / torch.full(
                [], value.size(-1) ** 0.5, dtype=attn_weights.dtype, device=attn_weights.device
            )

        # Layer-wise attention scaling
        if self.scale_attn_by_inverse_layer_idx:
            attn_weights = attn_weights / float(self.layer_idx + 1)

        if not self.is_cross_attention:
            # if only "normal" attention layer implements causal mask
            query_length, key_length = query.size(-2), key.size(-2)
            causal_mask = self.bias[:, :, key_length - query_length : key_length, :key_length]
            mask_value = torch.finfo(attn_weights.dtype).min
            # Need to be a tensor, otherwise we get error: `RuntimeError: expected scalar type float but found double`.
            # Need to be on the same device, otherwise `RuntimeError: ..., x and y to be on the same device`
            mask_value = torch.full([], mask_value, dtype=attn_weights.dtype, device=attn_weights.device)
            attn_weights = torch.where(causal_mask, attn_weights.to(attn_weights.dtype), mask_value)

        if attention_mask is not None:
            # Apply the attention mask
            attn_weights = attn_weights + attention_mask

        attn_weights = nn.functional.softmax(attn_weights, dim=-1)

        # Downcast (if necessary) back to V's dtype (if in mixed-precision) -- No-Op otherwise
        attn_weights = attn_weights.type(value.dtype)
        attn_weights = self.attn_dropout(attn_weights)

        # Mask heads if we want to
        if head_mask is not None:
            attn_weights = attn_weights * head_mask

        attn_output = torch.matmul(attn_weights, value)

        return attn_output, attn_weights

    def _upcast_and_reordered_attn(self, query, key, value, attention_mask=None, head_mask=None):
        # Use `torch.baddbmm` (a bit more efficient w/ alpha param for scaling -- from Megatron-LM)
        bsz, num_heads, q_seq_len, dk = query.size()
        _, _, k_seq_len, _ = key.size()

        # Preallocate attn_weights for `baddbmm`
        attn_weights = torch.empty(bsz * num_heads, q_seq_len, k_seq_len, dtype=torch.float32, device=query.device)

        # Compute Scale Factor
        scale_factor = 1.0
        if self.scale_attn_weights:
            scale_factor /= float(value.size(-1)) ** 0.5

        if self.scale_attn_by_inverse_layer_idx:
            scale_factor /= float(self.layer_idx + 1)

        # Upcast (turn off autocast) and reorder (Scale K by 1 / root(dk))
        with autocast(enabled=False):
            q, k = query.reshape(-1, q_seq_len, dk), key.transpose(-1, -2).reshape(-1, dk, k_seq_len)
            attn_weights = torch.baddbmm(attn_weights, q.float(), k.float(), beta=0, alpha=scale_factor)
            attn_weights = attn_weights.reshape(bsz, num_heads, q_seq_len, k_seq_len)

        if not self.is_cross_attention:
            # if only "normal" attention layer implements causal mask
            query_length, key_length = query.size(-2), key.size(-2)
            causal_mask = self.bias[:, :, key_length - query_length : key_length, :key_length]
            mask_value = torch.finfo(attn_weights.dtype).min
            # Need to be a tensor, otherwise we get error: `RuntimeError: expected scalar type float but found double`.
            # Need to be on the same device, otherwise `RuntimeError: ..., x and y to be on the same device`
            mask_value = torch.tensor(mask_value, dtype=attn_weights.dtype).to(attn_weights.device)
            attn_weights = torch.where(causal_mask, attn_weights, mask_value)

        if attention_mask is not None:
            # Apply the attention mask
            attn_weights = attn_weights + attention_mask

        attn_weights = nn.functional.softmax(attn_weights, dim=-1)

        # Downcast (if necessary) back to V's dtype (if in mixed-precision) -- No-Op if otherwise
        if attn_weights.dtype != torch.float32:
            raise RuntimeError("Error with upcasting, attn_weights does not have dtype torch.float32")
        attn_weights = attn_weights.type(value.dtype)
        attn_weights = self.attn_dropout(attn_weights)

        # Mask heads if we want to
        if head_mask is not None:
            attn_weights = attn_weights * head_mask

        attn_output = torch.matmul(attn_weights, value)

        return attn_output, attn_weights

    def _split_heads(self, tensor, num_heads, attn_head_size):
        """
        Splits hidden_size dim into attn_head_size and num_heads
        """
        new_shape = tensor.size()[:-1] + (num_heads, attn_head_size)
        tensor = tensor.view(new_shape)
        return tensor.permute(0, 2, 1, 3)  # (batch, head, seq_length, head_features)

    def _merge_heads(self, tensor, num_heads, attn_head_size):
        """
        Merges attn_head_size dim and num_attn_heads dim into hidden_size
        """
        tensor = tensor.permute(0, 2, 1, 3).contiguous()
        new_shape = tensor.size()[:-2] + (num_heads * attn_head_size,)
        return tensor.view(new_shape)

    def forward(
        self,
        hidden_states: Optional[Tuple[torch.FloatTensor]],
        layer_past: Optional[Tuple[torch.Tensor]] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        encoder_hidden_states: Optional[torch.Tensor] = None,
        encoder_attention_mask: Optional[torch.FloatTensor] = None,
        use_cache: Optional[bool] = False,
        output_attentions: Optional[bool] = False,
    ) -> Tuple[Union[torch.Tensor, Tuple[torch.Tensor]], ...]:
        if encoder_hidden_states is not None:
            if not hasattr(self, "q_attn"):
                raise ValueError(
                    "If class is used as cross attention, the weights `q_attn` have to be defined. "
                    "Please make sure to instantiate class with `MoTAttention(..., is_cross_attention=True)`."
                )

            query = self.q_attn(hidden_states)
            key, value = self.c_attn(encoder_hidden_states).split(self.split_size, dim=2)
            attention_mask = encoder_attention_mask
        else:
            query, key, value = self.c_attn(hidden_states).split(self.split_size, dim=2)

        query = self._split_heads(query, self.num_heads, self.head_dim)
        key = self._split_heads(key, self.num_heads, self.head_dim)
        value = self._split_heads(value, self.num_heads, self.head_dim)

        if layer_past is not None:
            past_key, past_value = layer_past
            key = torch.cat((past_key, key), dim=-2)
            value = torch.cat((past_value, value), dim=-2)

        if use_cache is True:
            present = (key, value)
        else:
            present = None

        if self.reorder_and_upcast_attn:
            attn_output, attn_weights = self._upcast_and_reordered_attn(query, key, value, attention_mask, head_mask)
        else:
            attn_output, attn_weights = self._attn(query, key, value, attention_mask, head_mask)

        attn_output = self._merge_heads(attn_output, self.num_heads, self.head_dim)
        attn_output = self.c_proj(attn_output)
        attn_output = self.resid_dropout(attn_output)

        outputs = (attn_output, present)
        if output_attentions:
            outputs += (attn_weights,)

        return outputs  # a, present, (attentions)


class MoTMLP(nn.Module):
    r"""
    Implementation of the Mixture of Tokens Sparse MLP module.
    """

    def __init__(self, inner_dim: int, config: MoTConfig, sparsity_dim: int = 0, init_type: str = "kaiming_uniform"):
        super().__init__()

        self.d_model: int = config.n_embd
        self.d_ff: int = config.n_inner if inner_dim is None else inner_dim
        self.n_expert: int = config.n_expert
        self.group_size: int = config.group_size
        self.sparsity_dim: int = sparsity_dim
        self.expert_size: int = config.expert_size
        self.temperature: float = config.temperature
        self.act = ACT2FN[config.activation_function]

        self.init_type: str = init_type
        self.init_scale: float = config.init_scale

        self.emit_softmax_over_experts: bool = config.emit_softmax_over_experts
        self.use_discrete_routing: bool = config.use_discrete_routing

        if self.n_expert is not None:
            if self.d_ff % self.n_expert:
                self.d_ff += self.n_expert - (self.d_ff % self.n_expert)
                warnings.warn("d_ff should be divisible by n_expert, padding d_ff to be divisible by n_expert")
            self.expert_size = self.d_ff // self.n_expert
        elif self.expert_size is not None:
            if self.d_ff % self.expert_size:
                self.d_ff += self.expert_size - (self.d_ff % self.expert_size)
                warnings.warn("d_ff should be divisible by expert_size, padding d_ff to be divisible by expert_size")
            self.n_expert = self.d_ff // self.expert_size
        else:
            raise ValueError("Either expert_size or n_expert should be provided")

        self.lin1 = nn.Parameter(
            self.get_init_weight(
                (self.n_expert, self.d_model, self.expert_size),
                fan_in=self.d_model,
                init_type=self.init_type,
                scale=self.init_scale,
            )
        )

        self.lin2 = nn.Parameter(
            self.get_init_weight(
                (self.n_expert, self.expert_size, self.d_model),
                fan_in=self.expert_size,
                init_type=self.init_type,
                scale=self.init_scale,
            )
        )

        self.controller = nn.Parameter(
            self.get_init_weight(
                (self.d_model, self.n_expert),
                fan_in=self.d_model,
                init_type=self.init_type,
                scale=self.init_scale,
            )
        )
        self.dropout = nn.Dropout(config.resid_pdrop)

    @staticmethod
    def argmax_one_hot(x: torch.Tensor, dim: int):
        max_values, _ = x.max(dim=dim, keepdim=True)
        return torch.where(
            condition=x == max_values,
            input=torch.Tensor([1.0]).to(dtype=x.dtype, device=x.device),
            other=torch.Tensor([0.0]).to(dtype=x.dtype, device=x.device),
        )  # potentially make it the value itself? torch.where(x == max_values, x, 0.0)

    def get_init_weight(self, shape, fan_in, init_type, scale, dtype=torch.float32):
        if init_type == "kaiming_uniform":
            return self.init_kaiming_uniform(shape=shape, fan_in=fan_in, scale=scale, dtype=dtype)
        elif init_type == "truncated_normal":
            return self.init_truncated_normal(shape=shape, fan_in=fan_in, scale=scale, dtype=dtype)
        else:
            raise ValueError(f"Unknown init_type: {init_type}")

    @staticmethod
    def init_kaiming_uniform(shape, fan_in, scale, dtype=torch.float32):
        range_ = scale * (3 / fan_in) ** 0.5
        return torch.zeros(shape, dtype=dtype).uniform_(-range_, range_)

    @staticmethod
    def init_truncated_normal(shape, fan_in, scale, dtype=torch.float32):
        std = (scale / fan_in) ** 0.5
        low = -2 * scale
        high = 2 * scale
        t = torch.zeros(shape, dtype=dtype)
        return trunc_normal_(t, mean=0.0, std=std, a=low, b=high)

    @staticmethod
    def stable_softmax_temperature(x: torch.Tensor, temperature: float, dim: int = -1) -> torch.Tensor:
        return F.softmax(x / temperature, dim=dim)

    def pad(self, x):
        size = x.size(0)
        ceiling = torch.ceil(torch.tensor(size / self.group_size).float())
        new_batch_size = self.group_size * ceiling.int().item()
        padding_size = new_batch_size - size
        logger.debug("Padding batch size from %d to %d", size, new_batch_size)

        # Create a zero sequence for padding
        zero_sequence = torch.zeros_like(x[0:1])
        padding_sequences = zero_sequence.repeat(padding_size, 1, 1)

        return torch.cat([x, padding_sequences], dim=0)

    @with_batch_size_alignment
    def forward(self, x):
        x = self.group_tokens(x)
        merge_weights, emit_weights = self.calculate_mixed_tokens_with_weights(x)
        x = self.merge_map_emit(x, merge_weights, emit_weights)
        x = self.redistribute_tokens(x)
        x = self.dropout(x)
        return x

    def group_tokens(self, x):
        """
        Reshape code so the axis to split into groups is on position 1, and then group over said axis.
        e.g.:
         - if we group tokens from different sequences in a batch (sparsity = 0), we need to put the batch dimension to position 1.
         - if we group tokens within one sequence, the dimension to split into groups is already on position 1, hence we leave it as is.

        free_dimension is the dimension on position 0 after reshape
        split_dimension is the dimension on position 1 - the one to split into groups

        :param x: normal input tensor of shape (batch, seq_len, dmodel)
        :return: x of shape (free_dimension, split_dimension // group_size, group_size , dmodel)
        """
        assert len(x.shape) == 3, "incorrect shape of a tensor, expected a 3D tensor"
        assert (
            x.size(-1) == self.d_model
        ), f"expected the last dimension of input tensor to be d_model = {self.d_model}"

        if self.sparsity_dim == 0:
            x = x.transpose(0, 1)
        elif self.sparsity_dim != 1:
            raise NotImplementedError

        free_dimension = x.size(1)
        assert (
            free_dimension % self.group_size == 0
        ), f"free dimension = {free_dimension} should be divisible by group size = {self.group_size}"

        x = x.view(x.size(0), -1, self.group_size, self.d_model)
        return x

    def redistribute_tokens(self, x):
        """
        An inverse operation to group_tokens.
        """
        assert len(x.shape) == 4, "incorrect shape of a tensor, expected a 4D tensor"

        x = x.view(x.size(0), -1, self.d_model)
        if self.sparsity_dim == 0:
            x = x.transpose(0, 1)
        elif self.sparsity_dim != 1:
            raise NotImplementedError

        return x

    def calculate_mixed_tokens_with_weights(self, x):
        """
        This function calculates merge and emit weights based on the input tensor, using a controller matrix.
        The merge weights determine the aggregation of tokens within a group, and emit weights govern the redistribution
        of the aggregated token back to the original tokens. Temperature scaling is applied to the logits, and optional
        discrete routing can be used to obtain one-hot representations of the weights.
        """
        # shape of x is (free_dimension, split_dimension // group_size, group_size, dmodel)
        merge_logits = torch.matmul(x, self.controller)
        # self.update_cache_for_logging("merge_logits", merge_logits)

        # shape of merge_logits is (free_dimension, aggr_dimension // group_size, group_size, n_expert)
        temp_merge = self.temperature
        temp_emit = self.temperature

        merge_softmax_dim = -2
        emit_softmax_dim = -1 if self.emit_softmax_over_experts else -2

        merge_weights = self.stable_softmax_temperature(merge_logits, temp_merge, dim=merge_softmax_dim)

        # by default we use the same weights for emitting and merging, but if the temperature is learnable or we want to take softmax over experts for emitting, we will use different weights
        if isinstance(temp_merge, torch.nn.Parameter) or self.emit_softmax_over_experts:
            emit_weights = self.stable_softmax_temperature(merge_logits, temp_emit, dim=emit_softmax_dim)
        else:
            emit_weights = merge_weights

        if self.use_discrete_routing:
            merge_weights = self.argmax_one_hot(merge_weights, dim=merge_softmax_dim)
            emit_weights = self.argmax_one_hot(emit_weights, dim=emit_softmax_dim)
        return merge_weights, emit_weights

    def merge_map_emit(self, x, merge_weights, emit_weights):
        """
        :param x: input reshaped to (free_dimension, split_dimension // group_size, group_size, dmodel)
        :param merge_weights: weights for merging tokens within a group, shape (free_dimension, split_dimension // group_size, group_size, n_expert)
        :param emit_weights: weights for emitting tokens within a group, shape (free_dimension, split_dimension // group_size, group_size, n_expert)
        :return: tensor of token updates of shape (free_dimension, split_dimension // group_size, group_size, dmodel)
        """
        x = torch.matmul(
            merge_weights.transpose(-1, -2),
            x,
        )
        # x shape is (free_dimension, split_dimension // group_size, n_expert, dmodel) ||| lin1 shape is (n_expert, dmodel, expert_size)
        x = torch.bmm(x.view(-1, self.n_expert, self.d_model).transpose(0, 1), self.lin1)
        x = self.act(x)
        # x shape is (n_expert, free_dimension * aggr_dimension // group_size, expert_size) ||| lin2 shape is (n_expert, expert_size, dmodel)
        x = torch.bmm(x, self.lin2)
        # x shape is (n_expert, free_dimension * aggr_dimension // group_size, dmodel)

        # merge_weights shape is (free_dimension, aggr_dimension // group_size, group_size, n_expert)
        # view x to be (n_expert, free_dimension, aggr_dimension // group_size, dmodel)
        # permute it to be (free_dimension, aggr_dimension // group_size, n_expert, dmodel)
        x = torch.matmul(
            emit_weights,
            x.view(x.size(0), emit_weights.size(0), -1, self.d_model).permute(1, 2, 0, 3),
        )

        return x


MoT_ATTENTION_CLASSES = {
    "eager": MoTAttention,
}


# Copied from transformers.models.gpt2.modeling_gpt2.GPT2Block with GPT2->MoT
class MoTBlock(nn.Module):
    def __init__(self, config, layer_idx=None):
        super().__init__()
        hidden_size = config.hidden_size
        inner_dim = config.n_inner if config.n_inner is not None else 4 * hidden_size
        attention_class = MoT_ATTENTION_CLASSES[config._attn_implementation]

        self.ln_1 = nn.LayerNorm(hidden_size, eps=config.layer_norm_epsilon)
        self.attn = attention_class(config=config, layer_idx=layer_idx)
        self.ln_2 = nn.LayerNorm(hidden_size, eps=config.layer_norm_epsilon)

        if config.add_cross_attention:
            self.crossattention = attention_class(config=config, is_cross_attention=True, layer_idx=layer_idx)
            self.ln_cross_attn = nn.LayerNorm(hidden_size, eps=config.layer_norm_epsilon)

        self.mlp = MoTMLP(inner_dim, config)

    def forward(
        self,
        hidden_states: Optional[Tuple[torch.FloatTensor]],
        layer_past: Optional[Tuple[torch.Tensor]] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        encoder_hidden_states: Optional[torch.Tensor] = None,
        encoder_attention_mask: Optional[torch.FloatTensor] = None,
        use_cache: Optional[bool] = False,
        output_attentions: Optional[bool] = False,
    ) -> Union[Tuple[torch.Tensor], Optional[Tuple[torch.Tensor, Tuple[torch.FloatTensor, ...]]]]:
        residual = hidden_states
        hidden_states = self.ln_1(hidden_states)
        attn_outputs = self.attn(
            hidden_states,
            layer_past=layer_past,
            attention_mask=attention_mask,
            head_mask=head_mask,
            use_cache=use_cache,
            output_attentions=output_attentions,
        )
        attn_output = attn_outputs[0]  # output_attn: a, present, (attentions)
        outputs = attn_outputs[1:]
        # residual connection
        hidden_states = attn_output + residual

        if encoder_hidden_states is not None:
            # add one self-attention block for cross-attention
            if not hasattr(self, "crossattention"):
                raise ValueError(
                    f"If `encoder_hidden_states` are passed, {self} has to be instantiated with "
                    "cross-attention layers by setting `config.add_cross_attention=True`"
                )
            residual = hidden_states
            hidden_states = self.ln_cross_attn(hidden_states)
            cross_attn_outputs = self.crossattention(
                hidden_states,
                attention_mask=attention_mask,
                head_mask=head_mask,
                encoder_hidden_states=encoder_hidden_states,
                encoder_attention_mask=encoder_attention_mask,
                output_attentions=output_attentions,
            )
            attn_output = cross_attn_outputs[0]
            # residual connection
            hidden_states = residual + attn_output
            outputs = outputs + cross_attn_outputs[2:]  # add cross attentions if we output attention weights

        residual = hidden_states
        hidden_states = self.ln_2(hidden_states)
        feed_forward_hidden_states = self.mlp(hidden_states)
        # residual connection
        hidden_states = residual + feed_forward_hidden_states

        if use_cache:
            outputs = (hidden_states,) + outputs
        else:
            outputs = (hidden_states,) + outputs[1:]

        return outputs  # hidden_states, present, (attentions, cross_attentions)


# Mostly copied from transformers.models.gpt2.modeling_gpt2.GPT2PreTrainedModel with GPT2->MoT,gpt2->mot,OpenAI GPT-2->MixtureOfTokens, but removed references to TF and FlashAttention2
class MoTPreTrainedModel(PreTrainedModel):
    """
    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
    models.
    """

    config_class = MoTConfig
    base_model_prefix = "transformer"
    is_parallelizable = True
    supports_gradient_checkpointing = True
    _no_split_modules = ["MoTBlock"]
    _skip_keys_device_placement = "past_key_values"

    def __init__(self, *inputs, **kwargs):
        super().__init__(*inputs, **kwargs)

    # Copied from transformers.models.gpt2.modeling_gpt2.GPT2PreTrainedModel._init_weights with GPT2->MoT,gpt2->mot,OpenAI GPT-2->MixtureOfTokens
    def _init_weights(self, module):
        """Initialize the weights."""
        if isinstance(module, (nn.Linear, Conv1D)):
            # Slightly different from the TF version which uses truncated_normal for initialization
            # cf https://github.com/pytorch/pytorch/pull/5617
            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
            if module.bias is not None:
                module.bias.data.zero_()
        elif isinstance(module, nn.Embedding):
            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
            if module.padding_idx is not None:
                module.weight.data[module.padding_idx].zero_()
        elif isinstance(module, nn.LayerNorm):
            module.bias.data.zero_()
            module.weight.data.fill_(1.0)

        # Reinitialize selected weights subject to the OpenAI GPT-2 Paper Scheme:
        #   > A modified initialization which accounts for the accumulation on the residual path with model depth. Scale
        #   > the weights of residual layers at initialization by a factor of 1/√N where N is the # of residual layers.
        #   >   -- GPT-2 :: https://openai.com/blog/better-language-models/
        #
        # Reference (Megatron-LM): https://github.com/NVIDIA/Megatron-LM/blob/main/megatron/model/gpt_model.py
        for name, p in module.named_parameters():
            if name == "c_proj.weight":
                # Special Scaled Initialization --> There are 2 Layer Norms per Transformer Block
                p.data.normal_(mean=0.0, std=(self.config.initializer_range / math.sqrt(2 * self.config.n_layer)))


MOT_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 ([`MoTConfig`]): 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.
"""

MOT_INPUTS_DOCSTRING = r"""
    Args:
        input_ids (`torch.LongTensor` of shape `(batch_size, input_ids_length)`):
            `input_ids_length` = `sequence_length` if `past_key_values` is `None` else
            `past_key_values[0][0].shape[-2]` (`sequence_length` of input past key value states). Indices of input
            sequence tokens in the vocabulary.

            If `past_key_values` is used, only `input_ids` that do not have their past calculated should be passed as
            `input_ids`.

            Indices can be obtained using [`GPT2Tokenizer`]. See [`PreTrainedTokenizer.encode`] and
            [`PreTrainedTokenizer.__call__`] for details.

            [What are input IDs?](../glossary#input-ids)
        past_key_values (`Tuple[Tuple[torch.Tensor]]` of length `config.n_layers`):
            Contains precomputed hidden-states (key and values in the attention blocks) as computed by the model (see
            `past_key_values` output below). Can be used to speed up sequential decoding. The `input_ids` which have
            their past given to this model should not be passed as `input_ids` as they have already been computed.
        attention_mask (`torch.FloatTensor` 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**.

            If `past_key_values` is used, `attention_mask` needs to contain the masking strategy that was used for
            `past_key_values`. In other words, the `attention_mask` always has to have the length:
            `len(past_key_values) + len(input_ids)`

            [What are attention masks?](../glossary#attention-mask)
        token_type_ids (`torch.LongTensor` of shape `(batch_size, input_ids_length)`, *optional*):
            Segment token indices to indicate first and second portions of the inputs. Indices are selected in `[0,
            1]`:

            - 0 corresponds to a *sentence A* token,
            - 1 corresponds to a *sentence B* token.

            [What are token type IDs?](../glossary#token-type-ids)
        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.max_position_embeddings - 1]`.

            [What are position IDs?](../glossary#position-ids)
        head_mask (`torch.FloatTensor` of shape `(num_heads,)` or `(num_layers, num_heads)`, *optional*):
            Mask to nullify selected heads of the self-attention modules. Mask values selected in `[0, 1]`:

            - 1 indicates the head is **not masked**,
            - 0 indicates the head is **masked**.

        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.

            If `past_key_values` is used, optionally only the last `inputs_embeds` have to be input (see
            `past_key_values`).
        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.
"""
PARALLELIZE_DOCSTRING = r"""
    This is an experimental feature and is a subject to change at a moment's notice.

    Uses a device map to distribute attention modules of the model across several devices. If no device map is given,
    it will evenly distribute blocks across all devices.

    Args:
        device_map (`Dict[int, list]`, optional, defaults to None):
            A dictionary that maps attention modules to devices. Note that the embedding module and LMHead are always
            automatically mapped to the first device (for esoteric reasons). That means that the first device should
            have fewer attention modules mapped to it than other devices. For reference, the mot models have the
            following number of attention modules:

                - mot: 12
                - mot-medium: 24
                - mot-large: 36
                - mot-xl: 48

    Example:

    ```python
    # Here is an example of a device map on a machine with 4 GPUs using mot-xl, which has a total of 48 attention modules:
    model = MoTLMHeadModel.from_pretrained("jaszczur/mixture_of_tokens")
    device_map = {
        0: [0, 1, 2, 3, 4, 5, 6, 7, 8],
        1: [9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21],
        2: [22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34],
        3: [35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47],
    }
    model.parallelize(device_map)
    ```
"""
DEPARALLELIZE_DOCSTRING = r"""
    Moves the model to cpu from a model parallel state.

    Example:

    ```python
    # On a 4 GPU machine with mot-large:
    model = MoTLMHeadModel.from_pretrained("jaszczur/mixture_of_tokens")
    device_map = {
        0: [0, 1, 2, 3, 4, 5, 6, 7],
        1: [8, 9, 10, 11, 12, 13, 14, 15],
        2: [16, 17, 18, 19, 20, 21, 22, 23],
        3: [24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35],
    }
    model.parallelize(device_map)  # Splits the model across several devices
    model.deparallelize()  # Put the model back on cpu and cleans memory by calling torch.cuda.empty_cache()
    ```
"""


@add_start_docstrings(
    "The bare MOT Model transformer outputting raw hidden-states without any specific head on top.",
    MOT_START_DOCSTRING,
)
class MoTModel(MoTPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)

        self.embed_dim = config.hidden_size

        self.wte = nn.Embedding(config.vocab_size, self.embed_dim)
        self.wpe = nn.Embedding(config.max_position_embeddings, self.embed_dim)

        self.drop = nn.Dropout(config.embd_pdrop)
        self.h = nn.ModuleList([MoTBlock(config, layer_idx=i) for i in range(config.num_hidden_layers)])
        self.ln_f = nn.LayerNorm(self.embed_dim, eps=config.layer_norm_epsilon)

        # Model parallel
        self.model_parallel = False
        self.device_map = None
        self.gradient_checkpointing = False

        # Initialize weights and apply final processing
        self.post_init()

    @add_start_docstrings(PARALLELIZE_DOCSTRING)
    def parallelize(self, device_map=None):
        # Check validity of device_map
        warnings.warn(
            "`MoTModel.parallelize` is deprecated and will be removed in v5 of Transformers, you should load your"
            " model with `device_map='balanced'` in the call to `from_pretrained`. You can also provide your own"
            " `device_map` but it needs to be a dictionary module_name to device, so for instance {'h.0': 0, 'h.1': 1,"
            " ...}",
            FutureWarning,
        )
        self.device_map = (
            get_device_map(len(self.h), range(torch.cuda.device_count())) if device_map is None else device_map
        )
        assert_device_map(self.device_map, len(self.h))
        self.model_parallel = True
        self.first_device = "cpu" if "cpu" in self.device_map.keys() else "cuda:" + str(min(self.device_map.keys()))
        self.last_device = "cuda:" + str(max(self.device_map.keys()))
        self.wte = self.wte.to(self.first_device)
        self.wpe = self.wpe.to(self.first_device)
        # Load onto devices
        for k, v in self.device_map.items():
            for block in v:
                cuda_device = "cuda:" + str(k)
                self.h[block] = self.h[block].to(cuda_device)
        # ln_f to last
        self.ln_f = self.ln_f.to(self.last_device)

    @add_start_docstrings(DEPARALLELIZE_DOCSTRING)
    def deparallelize(self):
        warnings.warn(
            "Like `parallelize`, `deparallelize` is deprecated and will be removed in v5 of Transformers.",
            FutureWarning,
        )
        self.model_parallel = False
        self.device_map = None
        self.first_device = "cpu"
        self.last_device = "cpu"
        self.wte = self.wte.to("cpu")
        self.wpe = self.wpe.to("cpu")
        for index in range(len(self.h)):
            self.h[index] = self.h[index].to("cpu")
        self.ln_f = self.ln_f.to("cpu")
        torch.cuda.empty_cache()

    def get_input_embeddings(self):
        return self.wte

    def set_input_embeddings(self, new_embeddings):
        self.wte = new_embeddings

    def _prune_heads(self, heads_to_prune):
        """
        Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer}
        """
        for layer, heads in heads_to_prune.items():
            self.h[layer].attn.prune_heads(heads)

    @add_start_docstrings_to_model_forward(MOT_INPUTS_DOCSTRING)
    @add_code_sample_docstrings(
        checkpoint=_CHECKPOINT_FOR_DOC,
        output_type=BaseModelOutputWithPastAndCrossAttentions,
        config_class=_CONFIG_FOR_DOC,
    )
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[Tuple[Tuple[torch.Tensor]]] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        token_type_ids: Optional[torch.LongTensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        encoder_hidden_states: Optional[torch.Tensor] = None,
        encoder_attention_mask: 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,
    ) -> Union[Tuple, BaseModelOutputWithPastAndCrossAttentions]:
        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 not None and inputs_embeds is not None:
            raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
        elif input_ids is not None:
            self.warn_if_padding_and_no_attention_mask(input_ids, attention_mask)
            input_shape = input_ids.size()
            input_ids = input_ids.view(-1, input_shape[-1])
            batch_size = input_ids.shape[0]
        elif inputs_embeds is not None:
            input_shape = inputs_embeds.size()[:-1]
            batch_size = inputs_embeds.shape[0]
        else:
            raise ValueError("You have to specify either input_ids or inputs_embeds")

        device = input_ids.device if input_ids is not None else inputs_embeds.device

        if token_type_ids is not None:
            token_type_ids = token_type_ids.view(-1, input_shape[-1])

        if past_key_values is None:
            past_length = 0
            past_key_values = tuple([None] * len(self.h))
        else:
            past_length = past_key_values[0][0].size(-2)
        if position_ids is None:
            position_ids = torch.arange(past_length, input_shape[-1] + past_length, dtype=torch.long, device=device)
            position_ids = position_ids.unsqueeze(0)

        # MoTAttention mask.
        if attention_mask is not None:
            if batch_size <= 0:
                raise ValueError("batch_size has to be defined and > 0")
            attention_mask = attention_mask.view(batch_size, -1)
            # We create a 3D attention mask from a 2D tensor mask.
            # Sizes are [batch_size, 1, 1, to_seq_length]
            # So we can broadcast to [batch_size, num_heads, from_seq_length, to_seq_length]
            # this attention mask is more simple than the triangular masking of causal attention
            # used in OpenAI GPT, we just need to prepare the broadcast dimension here.
            attention_mask = attention_mask[:, None, None, :]

            # Since attention_mask is 1.0 for positions we want to attend and 0.0 for
            # masked positions, this operation will create a tensor which is 0.0 for
            # positions we want to attend and the dtype's smallest value for masked positions.
            # Since we are adding it to the raw scores before the softmax, this is
            # effectively the same as removing these entirely.
            attention_mask = attention_mask.to(dtype=self.dtype)  # fp16 compatibility
            attention_mask = (1.0 - attention_mask) * torch.finfo(self.dtype).min

        # If a 2D or 3D attention mask is provided for the cross-attention
        # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
        if self.config.add_cross_attention and encoder_hidden_states is not None:
            encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.size()
            encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
            if encoder_attention_mask is None:
                encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device)
            encoder_attention_mask = self.invert_attention_mask(encoder_attention_mask)
        else:
            encoder_attention_mask = None

        # Prepare head mask if needed
        # 1.0 in head_mask indicate we keep the head
        # attention_probs has shape bsz x n_heads x N x N
        # head_mask has shape n_layer x batch x n_heads x N x N
        head_mask = self.get_head_mask(head_mask, self.config.n_layer)

        if inputs_embeds is None:
            inputs_embeds = self.wte(input_ids)
        position_embeds = self.wpe(position_ids)
        hidden_states = inputs_embeds + position_embeds

        if token_type_ids is not None:
            token_type_embeds = self.wte(token_type_ids)
            hidden_states = hidden_states + token_type_embeds

        hidden_states = self.drop(hidden_states)

        output_shape = (-1,) + input_shape[1:] + (hidden_states.size(-1),)

        if self.gradient_checkpointing and self.training:
            if use_cache:
                logger.warning_once(
                    "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
                )
                use_cache = False

        presents = () if use_cache else None
        all_self_attentions = () if output_attentions else None
        all_cross_attentions = () if output_attentions and self.config.add_cross_attention else None
        all_hidden_states = () if output_hidden_states else None
        for i, (block, layer_past) in enumerate(zip(self.h, past_key_values)):
            # Model parallel
            if self.model_parallel:
                torch.cuda.set_device(hidden_states.device)
                # Ensure layer_past is on same device as hidden_states (might not be correct)
                if layer_past is not None:
                    layer_past = tuple(past_state.to(hidden_states.device) for past_state in layer_past)
                # Ensure that attention_mask is always on the same device as hidden_states
                if attention_mask is not None:
                    attention_mask = attention_mask.to(hidden_states.device)
                if isinstance(head_mask, torch.Tensor):
                    head_mask = head_mask.to(hidden_states.device)
            if output_hidden_states:
                all_hidden_states = all_hidden_states + (hidden_states,)

            if self.gradient_checkpointing and self.training:
                outputs = self._gradient_checkpointing_func(
                    block.__call__,
                    hidden_states,
                    None,
                    attention_mask,
                    head_mask[i],
                    encoder_hidden_states,
                    encoder_attention_mask,
                    use_cache,
                    output_attentions,
                )
            else:
                outputs = block(
                    hidden_states,
                    layer_past=layer_past,
                    attention_mask=attention_mask,
                    head_mask=head_mask[i],
                    encoder_hidden_states=encoder_hidden_states,
                    encoder_attention_mask=encoder_attention_mask,
                    use_cache=use_cache,
                    output_attentions=output_attentions,
                )

            hidden_states = outputs[0]
            if use_cache is True:
                presents = presents + (outputs[1],)

            if output_attentions:
                all_self_attentions = all_self_attentions + (outputs[2 if use_cache else 1],)
                if self.config.add_cross_attention:
                    all_cross_attentions = all_cross_attentions + (outputs[3 if use_cache else 2],)

            # Model Parallel: If it's the last layer for that device, put things on the next device
            if self.model_parallel:
                for k, v in self.device_map.items():
                    if i == v[-1] and "cuda:" + str(k) != self.last_device:
                        hidden_states = hidden_states.to("cuda:" + str(k + 1))

        hidden_states = self.ln_f(hidden_states)

        hidden_states = hidden_states.view(output_shape)
        # Add last hidden state
        if output_hidden_states:
            all_hidden_states = all_hidden_states + (hidden_states,)

        if not return_dict:
            return tuple(
                v
                for v in [hidden_states, presents, all_hidden_states, all_self_attentions, all_cross_attentions]
                if v is not None
            )

        return BaseModelOutputWithPastAndCrossAttentions(
            last_hidden_state=hidden_states,
            past_key_values=presents,
            hidden_states=all_hidden_states,
            attentions=all_self_attentions,
            cross_attentions=all_cross_attentions,
        )


@add_start_docstrings(
    """
    The MOT Model transformer with a language modeling head on top (linear layer with weights tied to the input
    embeddings).
    """,
    MOT_START_DOCSTRING,
)
class MoTLMHeadModel(MoTPreTrainedModel):
    _tied_weights_keys = ["lm_head.weight"]

    def __init__(self, config):
        super().__init__(config)
        self.transformer = MoTModel(config)
        self.lm_head = nn.Linear(config.n_embd, config.vocab_size, bias=False)

        # Model parallel
        self.model_parallel = False
        self.device_map = None

        # Initialize weights and apply final processing
        self.post_init()

    @add_start_docstrings(PARALLELIZE_DOCSTRING)
    def parallelize(self, device_map=None):
        warnings.warn(
            "`MoTLMHeadModel.parallelize` is deprecated and will be removed in v5 of Transformers, you should load"
            " your model with `device_map='balanced'` in the call to `from_pretrained`. You can also provide your own"
            " `device_map` but it needs to be a dictionary module_name to device, so for instance {'transformer.h.0':"
            " 0, 'transformer.h.1': 1, ...}",
            FutureWarning,
        )
        self.device_map = (
            get_device_map(len(self.transformer.h), range(torch.cuda.device_count()))
            if device_map is None
            else device_map
        )
        assert_device_map(self.device_map, len(self.transformer.h))
        self.transformer.parallelize(self.device_map)
        self.lm_head = self.lm_head.to(self.transformer.first_device)
        self.model_parallel = True

    @add_start_docstrings(DEPARALLELIZE_DOCSTRING)
    def deparallelize(self):
        warnings.warn(
            "Like `parallelize`, `deparallelize` is deprecated and will be removed in v5 of Transformers.",
            FutureWarning,
        )
        self.transformer.deparallelize()
        self.transformer = self.transformer.to("cpu")
        self.lm_head = self.lm_head.to("cpu")
        self.model_parallel = False
        torch.cuda.empty_cache()

    def get_output_embeddings(self):
        return self.lm_head

    def set_output_embeddings(self, new_embeddings):
        self.lm_head = new_embeddings

    def prepare_inputs_for_generation(self, input_ids, past_key_values=None, inputs_embeds=None, **kwargs):
        token_type_ids = kwargs.get("token_type_ids", None)
        # Omit tokens covered by past_key_values
        if past_key_values:
            past_length = past_key_values[0][0].shape[2]

            # Some generation methods already pass only the last input ID
            if input_ids.shape[1] > past_length:
                remove_prefix_length = past_length
            else:
                # Default to old behavior: keep only final ID
                remove_prefix_length = input_ids.shape[1] - 1

            input_ids = input_ids[:, remove_prefix_length:]
            if token_type_ids is not None:
                token_type_ids = token_type_ids[:, -input_ids.shape[1] :]

        attention_mask = kwargs.get("attention_mask", None)
        position_ids = kwargs.get("position_ids", None)

        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] :]
        else:
            position_ids = None

        # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
        if inputs_embeds is not None and past_key_values is None:
            model_inputs = {"inputs_embeds": inputs_embeds}
        else:
            model_inputs = {"input_ids": input_ids}

        model_inputs.update(
            {
                "past_key_values": past_key_values,
                "use_cache": kwargs.get("use_cache"),
                "position_ids": position_ids,
                "attention_mask": attention_mask,
                "token_type_ids": token_type_ids,
            }
        )

        return model_inputs

    @add_start_docstrings_to_model_forward(MOT_INPUTS_DOCSTRING)
    @add_code_sample_docstrings(
        checkpoint=_CHECKPOINT_FOR_DOC,
        output_type=CausalLMOutputWithCrossAttentions,
        config_class=_CONFIG_FOR_DOC,
    )
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[Tuple[Tuple[torch.Tensor]]] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        token_type_ids: Optional[torch.LongTensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        encoder_hidden_states: Optional[torch.Tensor] = None,
        encoder_attention_mask: 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,
    ) -> Union[Tuple, CausalLMOutputWithCrossAttentions]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for language modeling. Note that the labels **are shifted** inside the model, i.e. you can set
            `labels = input_ids` Indices are selected in `[-100, 0, ..., config.vocab_size]` All labels set to `-100`
            are ignored (masked), the loss is only computed for labels in `[0, ..., config.vocab_size]`
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        transformer_outputs = self.transformer(
            input_ids,
            past_key_values=past_key_values,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            encoder_hidden_states=encoder_hidden_states,
            encoder_attention_mask=encoder_attention_mask,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )
        hidden_states = transformer_outputs[0]

        # Set device for model parallelism
        if self.model_parallel:
            torch.cuda.set_device(self.transformer.first_device)
            hidden_states = hidden_states.to(self.lm_head.weight.device)

        lm_logits = self.lm_head(hidden_states)

        loss = None
        if labels is not None:
            # move labels to correct device to enable model parallelism
            labels = labels.to(lm_logits.device)
            # Shift so that tokens < n predict n
            shift_logits = lm_logits[..., :-1, :].contiguous()
            shift_labels = labels[..., 1:].contiguous()
            # Flatten the tokens
            loss_fct = CrossEntropyLoss()
            loss = loss_fct(shift_logits.view(-1, shift_logits.size(-1)), shift_labels.view(-1))

        if not return_dict:
            output = (lm_logits,) + transformer_outputs[1:]
            return ((loss,) + output) if loss is not None else output

        return CausalLMOutputWithCrossAttentions(
            loss=loss,
            logits=lm_logits,
            past_key_values=transformer_outputs.past_key_values,
            hidden_states=transformer_outputs.hidden_states,
            attentions=transformer_outputs.attentions,
            cross_attentions=transformer_outputs.cross_attentions,
        )

    @staticmethod
    def _reorder_cache(
        past_key_values: Tuple[Tuple[torch.Tensor]], beam_idx: torch.Tensor
    ) -> Tuple[Tuple[torch.Tensor]]:
        """
        This function is used to re-order the `past_key_values` cache if [`~PreTrainedModel.beam_search`] or
        [`~PreTrainedModel.beam_sample`] is called. This is required to match `past_key_values` with the correct
        beam_idx at every generation step.
        """
        return tuple(
            tuple(past_state.index_select(0, beam_idx.to(past_state.device)) for past_state in layer_past)
            for layer_past in past_key_values
        )


@add_start_docstrings(
    """
The MoT Model transformer with a language modeling and a multiple-choice classification head on top e.g. for
RocStories/SWAG tasks. The two heads are two linear layers. The language modeling head has its weights tied to the
input embeddings, the classification head takes as input the input of a specified classification token index in the
input sequence).
""",
    MOT_START_DOCSTRING,
)
@add_start_docstrings(
    """
    The MOT Model transformer with a sequence classification head on top (linear layer).

    [`MoTForSequenceClassification`] uses the last token in order to do the classification, as other causal models
    (e.g. GPT-1) do.

    Since it does classification on the last token, it requires to know the position of the last token. If a
    `pad_token_id` is defined in the configuration, it finds the last token that is not a padding token in each row. If
    no `pad_token_id` is defined, it simply takes the last value in each row of the batch. Since it cannot guess the
    padding tokens when `inputs_embeds` are passed instead of `input_ids`, it does the same (take the last value in
    each row of the batch).
    """,
    MOT_START_DOCSTRING,
)
class MoTForSequenceClassification(MoTPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels
        self.transformer = MoTModel(config)
        self.score = nn.Linear(config.n_embd, self.num_labels, bias=False)

        # Model parallel
        self.model_parallel = False
        self.device_map = None

        # Initialize weights and apply final processing
        self.post_init()

    @add_start_docstrings_to_model_forward(MOT_INPUTS_DOCSTRING)
    @add_code_sample_docstrings(
        checkpoint=_CHECKPOINT_FOR_DOC,
        output_type=SequenceClassifierOutputWithPast,
        config_class=_CONFIG_FOR_DOC,
    )
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[Tuple[Tuple[torch.Tensor]]] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        token_type_ids: Optional[torch.LongTensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        head_mask: Optional[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,
    ) -> Union[Tuple, SequenceClassifierOutputWithPast]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        transformer_outputs = self.transformer(
            input_ids,
            past_key_values=past_key_values,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )
        hidden_states = transformer_outputs[0]
        logits = self.score(hidden_states)

        if input_ids is not None:
            batch_size, sequence_length = input_ids.shape[:2]
        else:
            batch_size, sequence_length = inputs_embeds.shape[:2]

        assert (
            self.config.pad_token_id is not None or batch_size == 1
        ), "Cannot handle batch sizes > 1 if no padding token is defined."
        if self.config.pad_token_id is None:
            sequence_lengths = -1
        else:
            if input_ids is not None:
                # if no pad token found, use modulo instead of reverse indexing for ONNX compatibility
                sequence_lengths = torch.eq(input_ids, self.config.pad_token_id).int().argmax(-1) - 1
                sequence_lengths = sequence_lengths % input_ids.shape[-1]
                sequence_lengths = sequence_lengths.to(logits.device)
            else:
                sequence_lengths = -1
                logger.warning(
                    f"{self.__class__.__name__} will not detect padding tokens in `inputs_embeds`. Results may be "
                    "unexpected if using padding tokens in conjunction with `inputs_embeds.`"
                )

        pooled_logits = logits[torch.arange(batch_size, device=logits.device), sequence_lengths]

        loss = None
        if labels is not None:
            if self.config.problem_type is None:
                if self.num_labels == 1:
                    self.config.problem_type = "regression"
                elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
                    self.config.problem_type = "single_label_classification"
                else:
                    self.config.problem_type = "multi_label_classification"

            if self.config.problem_type == "regression":
                loss_fct = MSELoss()
                if self.num_labels == 1:
                    loss = loss_fct(pooled_logits.squeeze(), labels.squeeze())
                else:
                    loss = loss_fct(pooled_logits, labels)
            elif self.config.problem_type == "single_label_classification":
                loss_fct = CrossEntropyLoss()
                loss = loss_fct(pooled_logits.view(-1, self.num_labels), labels.view(-1))
            elif self.config.problem_type == "multi_label_classification":
                loss_fct = BCEWithLogitsLoss()
                loss = loss_fct(pooled_logits, labels)
        if not return_dict:
            output = (pooled_logits,) + transformer_outputs[1:]
            return ((loss,) + output) if loss is not None else output

        return SequenceClassifierOutputWithPast(
            loss=loss,
            logits=pooled_logits,
            past_key_values=transformer_outputs.past_key_values,
            hidden_states=transformer_outputs.hidden_states,
            attentions=transformer_outputs.attentions,
        )


@add_start_docstrings(
    """
    MOT Model with a token classification head on top (a linear layer on top of the hidden-states output) e.g. for
    Named-Entity-Recognition (NER) tasks.
    """,
    MOT_START_DOCSTRING,
)
class MoTForTokenClassification(MoTPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels

        self.transformer = MoTModel(config)
        if hasattr(config, "classifier_dropout") and config.classifier_dropout is not None:
            classifier_dropout = config.classifier_dropout
        elif hasattr(config, "hidden_dropout") and config.hidden_dropout is not None:
            classifier_dropout = config.hidden_dropout
        else:
            classifier_dropout = 0.1
        self.dropout = nn.Dropout(classifier_dropout)
        self.classifier = nn.Linear(config.hidden_size, config.num_labels)

        # Model parallel
        self.model_parallel = False
        self.device_map = None

        # Initialize weights and apply final processing
        self.post_init()

    @add_start_docstrings_to_model_forward(MOT_INPUTS_DOCSTRING)
    # fmt: off
    @add_code_sample_docstrings(
        checkpoint=_CHECKPOINT_FOR_DOC,
        output_type=TokenClassifierOutput,
        config_class=_CONFIG_FOR_DOC,
        expected_loss=0.25,
        expected_output=[
            "Lead",
            "Lead",
            "Lead",
            "Position",
            "Lead",
            "Lead",
            "Lead",
            "Lead",
            "Lead",
            "Lead",
            "Lead",
            "Lead",
        ],
    )
    # fmt: on
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[Tuple[Tuple[torch.Tensor]]] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        token_type_ids: Optional[torch.LongTensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        head_mask: Optional[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,
    ) -> Union[Tuple, TokenClassifierOutput]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        transformer_outputs = self.transformer(
            input_ids,
            past_key_values=past_key_values,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        hidden_states = transformer_outputs[0]
        hidden_states = self.dropout(hidden_states)
        logits = self.classifier(hidden_states)

        loss = None
        if labels is not None:
            labels = labels.to(logits.device)
            loss_fct = CrossEntropyLoss()
            loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))

        if not return_dict:
            output = (logits,) + transformer_outputs[2:]
            return ((loss,) + output) if loss is not None else output

        return TokenClassifierOutput(
            loss=loss,
            logits=logits,
            hidden_states=transformer_outputs.hidden_states,
            attentions=transformer_outputs.attentions,
        )


@add_start_docstrings(
    """
    The MixtureOfTokens transformer with a span classification head on top for extractive question-answering tasks like
    SQuAD (a linear layer on top of the hidden-states output to compute `span start logits` and `span end logits`).
    """,
    MOT_START_DOCSTRING,
)
class MoTForQuestionAnswering(MoTPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels
        self.transformer = MoTModel(config)
        self.qa_outputs = nn.Linear(config.hidden_size, 2)

        # Model parallel
        self.model_parallel = False
        self.device_map = None

        # Initialize weights and apply final processing
        self.post_init()

    @add_start_docstrings_to_model_forward(MOT_INPUTS_DOCSTRING.format("batch_size, sequence_length"))
    @add_code_sample_docstrings(
        checkpoint=_CHECKPOINT_FOR_DOC,
        output_type=QuestionAnsweringModelOutput,
        config_class=_CONFIG_FOR_DOC,
    )
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        token_type_ids: Optional[torch.LongTensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        start_positions: Optional[torch.LongTensor] = None,
        end_positions: Optional[torch.LongTensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple, QuestionAnsweringModelOutput]:
        r"""
        start_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for position (index) of the start of the labelled span for computing the token classification loss.
            Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
            are not taken into account for computing the loss.
        end_positions (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for position (index) of the end of the labelled span for computing the token classification loss.
            Positions are clamped to the length of the sequence (`sequence_length`). Position outside of the sequence
            are not taken into account for computing the loss.
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.transformer(
            input_ids,
            attention_mask=attention_mask,
            token_type_ids=token_type_ids,
            position_ids=position_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        sequence_output = outputs[0]

        logits = self.qa_outputs(sequence_output)
        start_logits, end_logits = logits.split(1, dim=-1)
        start_logits = start_logits.squeeze(-1).contiguous()
        end_logits = end_logits.squeeze(-1).contiguous()

        total_loss = None
        if start_positions is not None and end_positions is not None:
            # If we are on multi-GPU, split add a dimension
            if len(start_positions.size()) > 1:
                start_positions = start_positions.squeeze(-1).to(start_logits.device)
            if len(end_positions.size()) > 1:
                end_positions = end_positions.squeeze(-1).to(end_logits.device)
            # sometimes the start/end positions are outside our model inputs, we ignore these terms
            ignored_index = start_logits.size(1)
            start_positions = start_positions.clamp(0, ignored_index)
            end_positions = end_positions.clamp(0, ignored_index)

            loss_fct = CrossEntropyLoss(ignore_index=ignored_index)
            start_loss = loss_fct(start_logits, start_positions)
            end_loss = loss_fct(end_logits, end_positions)
            total_loss = (start_loss + end_loss) / 2

        if not return_dict:
            output = (start_logits, end_logits) + outputs[2:]
            return ((total_loss,) + output) if total_loss is not None else output

        return QuestionAnsweringModelOutput(
            loss=total_loss,
            start_logits=start_logits,
            end_logits=end_logits,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )