prefix2.py

""" PyTorch BART model."""
import math
import random
import warnings
from typing import List, Optional, Tuple, Union, Dict

import torch
import torch.utils.checkpoint
from torch import nn
from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
from transformers.models.bart.modeling_bart import *
from transformers.generation_utils import *
import ipdb

logger = logging.get_logger(__name__)

_CHECKPOINT_FOR_DOC = "facebook/bart-base"
_CONFIG_FOR_DOC = "BartConfig"
_TOKENIZER_FOR_DOC = "BartTokenizer"

# Base model docstring
_EXPECTED_OUTPUT_SHAPE = [1, 8, 768]

# SequenceClassification docstring
_CHECKPOINT_FOR_SEQUENCE_CLASSIFICATION = "valhalla/bart-large-sst2"
_SEQ_CLASS_EXPECTED_LOSS = 0.0
_SEQ_CLASS_EXPECTED_OUTPUT = "'POSITIVE'"

# QuestionAsnwering docstring
_CHECKPOINT_FOR_QA = "valhalla/bart-large-finetuned-squadv1"
_QA_EXPECTED_LOSS = 0.59
_QA_EXPECTED_OUTPUT = "' nice puppet'"


BART_PRETRAINED_MODEL_ARCHIVE_LIST = [
    "facebook/bart-large",
    # see all BART models at https://huggingface.co/models?filter=bart
]

def shift_tokens_right(input_ids: torch.Tensor, pad_token_id: int, decoder_start_token_id: int):
    """
    Shift input ids one token to the right.
    """
    shifted_input_ids = input_ids.new_zeros(input_ids.shape)
    shifted_input_ids[:, 1:] = input_ids[:, :-1].clone()
    shifted_input_ids[:, 0] = decoder_start_token_id

    if pad_token_id is None:
        raise ValueError("self.model.config.pad_token_id has to be defined.")
    # replace possible -100 values in labels by `pad_token_id`
    shifted_input_ids.masked_fill_(shifted_input_ids == -100, pad_token_id)

    return shifted_input_ids


def _make_causal_mask(input_ids_shape: torch.Size, dtype: torch.dtype, past_key_values_length: int = 0):
    """
    Make causal mask used for bi-directional self-attention.
    """
    bsz, tgt_len = input_ids_shape
    mask = torch.full((tgt_len, tgt_len), float("-inf"))
    mask_cond = torch.arange(mask.size(-1))
    mask.masked_fill_(mask_cond < (mask_cond + 1).view(mask.size(-1), 1), 0)
    mask = mask.to(dtype)

    if past_key_values_length > 0:
        mask = torch.cat([torch.zeros(tgt_len, past_key_values_length, dtype=dtype), mask], dim=-1)
    return mask[None, None, :, :].expand(bsz, 1, tgt_len, tgt_len + past_key_values_length)


def _expand_mask(mask: torch.Tensor, dtype: torch.dtype, tgt_len: Optional[int] = None):
    """
    Expands attention_mask from `[bsz, seq_len]` to `[bsz, 1, tgt_seq_len, src_seq_len]`.
    """
    bsz, src_len = mask.size()
    tgt_len = tgt_len if tgt_len is not None else src_len

    expanded_mask = mask[:, None, None, :].expand(bsz, 1, tgt_len, src_len).to(dtype)

    inverted_mask = 1.0 - expanded_mask

    return inverted_mask.masked_fill(inverted_mask.bool(), torch.finfo(dtype).min)


class PrefixGenBartAttention(nn.Module):
    """Multi-headed attention from 'Attention Is All You Need' paper"""

    def __init__(
        self,
        embed_dim: int,
        num_heads: int,
        dropout: float = 0.0,
        is_decoder: bool = False,
        bias: bool = True,
    ):
        super().__init__()
        self.embed_dim = embed_dim
        self.num_heads = num_heads
        self.dropout = dropout
        self.head_dim = embed_dim // num_heads

        if (self.head_dim * num_heads) != self.embed_dim:
            raise ValueError(
                f"embed_dim must be divisible by num_heads (got `embed_dim`: {self.embed_dim}"
                f" and `num_heads`: {num_heads})."
            )
        self.scaling = self.head_dim**-0.5
        self.is_decoder = is_decoder

        self.k_proj = nn.Linear(embed_dim, embed_dim, bias=bias)
        self.v_proj = nn.Linear(embed_dim, embed_dim, bias=bias)
        self.q_proj = nn.Linear(embed_dim, embed_dim, bias=bias)
        self.out_proj = nn.Linear(embed_dim, embed_dim, bias=bias)

    def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
        return tensor.view(bsz, seq_len, self.num_heads, self.head_dim).transpose(1, 2).contiguous()
    
    # Tag: Changed
    def apply_prefix(self, key_states, value_states, prefix, attention_mask=None):
        """
            key_states: (bz, head_num, src_len, dim)
            value_states: (bz, head_num, src_len, dim)
            prefix: Tuple( Tensor(bz, prefix_len, head_num, dim), Tensor(bz, prefix_len, head_num, dim))
        """
        batch_size = key_states.size(0)
        prefix_keys = prefix[0].transpose(1, 2).contiguous()
        prefix_values = prefix[1].transpose(1, 2).contiguous()
        key_states = torch.cat([prefix_keys, key_states], dim=2)
        value_states = torch.cat([prefix_values, value_states], dim=2)

        if attention_mask is not None:
            if attention_mask.dim() == 2:
                prefix_mask = torch.ones(batch_size, prefix_keys.size(2)).to(attention_mask.device)
            else:
                prefix_mask = torch.ones(batch_size, 1, attention_mask.size(2), prefix_keys.size(2)).to(
                    attention_mask.device
                )
            attention_mask = torch.cat([prefix_mask, attention_mask], dim=-1)
        
        return key_states, value_states, attention_mask

    def forward(
        self,
        hidden_states: torch.Tensor,
        prefix: Optional[Tuple[torch.FloatTensor]] = None, # Tag: Changed
        key_value_states: Optional[torch.Tensor] = None,
        past_key_value: Optional[Tuple[torch.Tensor]] = None,
        attention_mask: Optional[torch.Tensor] = None,
        layer_head_mask: Optional[torch.Tensor] = None,
        output_attentions: bool = False,
    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
        """Input shape: Batch x Time x Channel"""

        # if key_value_states are provided this layer is used as a cross-attention layer
        # for the decoder
        is_cross_attention = key_value_states is not None

        bsz, tgt_len, _ = hidden_states.size()

        # get query proj
        query_states = self.q_proj(hidden_states) * self.scaling
        # get key, value proj
        if is_cross_attention and past_key_value is not None:
            # reuse k,v, cross_attentions
            key_states = past_key_value[0]
            value_states = past_key_value[1]
        elif is_cross_attention:
            # cross_attentions
            key_states = self._shape(self.k_proj(key_value_states), -1, bsz)
            value_states = self._shape(self.v_proj(key_value_states), -1, bsz)
        elif past_key_value is not None:
            # reuse k, v, self_attention
            key_states = self._shape(self.k_proj(hidden_states), -1, bsz)
            value_states = self._shape(self.v_proj(hidden_states), -1, bsz)
            key_states = torch.cat([past_key_value[0], key_states], dim=2)
            value_states = torch.cat([past_key_value[1], value_states], dim=2)
        else:
            # self_attention
            key_states = self._shape(self.k_proj(hidden_states), -1, bsz)
            value_states = self._shape(self.v_proj(hidden_states), -1, bsz)

        if self.is_decoder:
            # if cross_attention save Tuple(torch.Tensor, torch.Tensor) of all cross attention key/value_states.
            # Further calls to cross_attention layer can then reuse all cross-attention
            # key/value_states (first "if" case)
            # if uni-directional self-attention (decoder) save Tuple(torch.Tensor, torch.Tensor) of
            # all previous decoder key/value_states. Further calls to uni-directional self-attention
            # can concat previous decoder key/value_states to current projected key/value_states (third "elif" case)
            # if encoder bi-directional self-attention `past_key_value` is always `None`
            past_key_value = (key_states, value_states)

        proj_shape = (bsz * self.num_heads, -1, self.head_dim)
        query_states = self._shape(query_states, tgt_len, bsz).view(*proj_shape)

        # Tag: Changed
        if prefix is not None:
            key_states, value_states, attention_mask = self.apply_prefix(
                key_states, value_states, prefix, attention_mask
            )

        key_states = key_states.view(*proj_shape)
        value_states = value_states.view(*proj_shape)

        src_len = key_states.size(1)
        attn_weights = torch.bmm(query_states, key_states.transpose(1, 2))

        if attn_weights.size() != (bsz * self.num_heads, tgt_len, src_len):
            raise ValueError(
                f"Attention weights should be of size {(bsz * self.num_heads, tgt_len, src_len)}, but is {attn_weights.size()}"
            )

        if attention_mask is not None:
            if attention_mask.size() != (bsz, 1, tgt_len, src_len):
                raise ValueError(
                    f"Attention mask should be of size {(bsz, 1, tgt_len, src_len)}, but is {attention_mask.size()}"
                )
            attn_weights = attn_weights.view(bsz, self.num_heads, tgt_len, src_len) + attention_mask
            attn_weights = attn_weights.view(bsz * self.num_heads, tgt_len, src_len)

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

        if layer_head_mask is not None:
            if layer_head_mask.size() != (self.num_heads,):
                raise ValueError(
                    f"Head mask for a single layer should be of size {(self.num_heads,)}, but is {layer_head_mask.size()}"
                )
            attn_weights = layer_head_mask.view(1, -1, 1, 1) * attn_weights.view(bsz, self.num_heads, tgt_len, src_len)
            attn_weights = attn_weights.view(bsz * self.num_heads, tgt_len, src_len)

        if output_attentions:
            # this operation is a bit awkward, but it's required to
            # make sure that attn_weights keeps its gradient.
            # In order to do so, attn_weights have to be reshaped
            # twice and have to be reused in the following
            attn_weights_reshaped = attn_weights.view(bsz, self.num_heads, tgt_len, src_len)
            attn_weights = attn_weights_reshaped.view(bsz * self.num_heads, tgt_len, src_len)
        else:
            attn_weights_reshaped = None

        attn_probs = nn.functional.dropout(attn_weights, p=self.dropout, training=self.training)

        attn_output = torch.bmm(attn_probs, value_states)

        if attn_output.size() != (bsz * self.num_heads, tgt_len, self.head_dim):
            raise ValueError(
                f"`attn_output` should be of size {(bsz, self.num_heads, tgt_len, self.head_dim)}, but is {attn_output.size()}"
            )

        attn_output = attn_output.view(bsz, self.num_heads, tgt_len, self.head_dim)
        attn_output = attn_output.transpose(1, 2)

        # Use the `embed_dim` from the config (stored in the class) rather than `hidden_state` because `attn_output` can be
        # partitioned aross GPUs when using tensor-parallelism.
        attn_output = attn_output.reshape(bsz, tgt_len, self.embed_dim)

        attn_output = self.out_proj(attn_output)

        return attn_output, attn_weights_reshaped, past_key_value


class PrefixGenBartEncoderLayer(nn.Module):
    def __init__(self, config: BartConfig):
        super().__init__()
        self.embed_dim = config.d_model
        self.self_attn = PrefixGenBartAttention(
            embed_dim=self.embed_dim,
            num_heads=config.encoder_attention_heads,
            dropout=config.attention_dropout,
        )
        self.self_attn_layer_norm = nn.LayerNorm(self.embed_dim)
        self.dropout = config.dropout
        self.activation_fn = ACT2FN[config.activation_function]
        self.activation_dropout = config.activation_dropout
        self.fc1 = nn.Linear(self.embed_dim, config.encoder_ffn_dim)
        self.fc2 = nn.Linear(config.encoder_ffn_dim, self.embed_dim)
        self.final_layer_norm = nn.LayerNorm(self.embed_dim)

    def forward(
        self,
        hidden_states: torch.FloatTensor,
        attention_mask: torch.FloatTensor,
        layer_head_mask: torch.FloatTensor,
        prefix: Optional[Tuple[torch.FloatTensor]] = None, # Tag: Changed
        output_attentions: Optional[bool] = False,
    ) -> Tuple[torch.FloatTensor, Optional[torch.FloatTensor]]:
        """
        Args:
            hidden_states (`torch.FloatTensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
            attention_mask (`torch.FloatTensor`): attention mask of size
                `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
            layer_head_mask (`torch.FloatTensor`): mask for attention heads in a given layer of size
                `(encoder_attention_heads,)`.
            output_attentions (`bool`, *optional*):
                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
                returned tensors for more detail.
        """
        residual = hidden_states
        hidden_states, attn_weights, _ = self.self_attn(
            hidden_states=hidden_states,
            prefix=prefix,
            attention_mask=attention_mask,
            layer_head_mask=layer_head_mask,
            output_attentions=output_attentions,
        )
        hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
        hidden_states = residual + hidden_states
        hidden_states = self.self_attn_layer_norm(hidden_states)

        residual = hidden_states
        hidden_states = self.activation_fn(self.fc1(hidden_states))
        hidden_states = nn.functional.dropout(hidden_states, p=self.activation_dropout, training=self.training)
        hidden_states = self.fc2(hidden_states)
        hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
        hidden_states = residual + hidden_states
        hidden_states = self.final_layer_norm(hidden_states)

        if hidden_states.dtype == torch.float16 and (
            torch.isinf(hidden_states).any() or torch.isnan(hidden_states).any()
        ):
            clamp_value = torch.finfo(hidden_states.dtype).max - 1000
            hidden_states = torch.clamp(hidden_states, min=-clamp_value, max=clamp_value)

        outputs = (hidden_states,)

        if output_attentions:
            outputs += (attn_weights,)

        return outputs


class PrefixGenBartDecoderLayer(nn.Module):
    def __init__(self, config: BartConfig):
        super().__init__()
        self.embed_dim = config.d_model

        self.self_attn = PrefixGenBartAttention(
            embed_dim=self.embed_dim,
            num_heads=config.decoder_attention_heads,
            dropout=config.attention_dropout,
            is_decoder=True,
        )
        self.dropout = config.dropout
        self.activation_fn = ACT2FN[config.activation_function]
        self.activation_dropout = config.activation_dropout

        self.self_attn_layer_norm = nn.LayerNorm(self.embed_dim)
        self.encoder_attn = PrefixGenBartAttention(
            self.embed_dim,
            config.decoder_attention_heads,
            dropout=config.attention_dropout,
            is_decoder=True,
        )
        self.encoder_attn_layer_norm = nn.LayerNorm(self.embed_dim)
        self.fc1 = nn.Linear(self.embed_dim, config.decoder_ffn_dim)
        self.fc2 = nn.Linear(config.decoder_ffn_dim, self.embed_dim)
        self.final_layer_norm = nn.LayerNorm(self.embed_dim)

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
        encoder_hidden_states: Optional[torch.Tensor] = None,
        encoder_attention_mask: Optional[torch.Tensor] = None,
        layer_head_mask: Optional[torch.Tensor] = None,
        cross_prefix: Optional[Tuple[torch.FloatTensor]] = None, # Tag: Changed
        decoder_prefix: Optional[Tuple[torch.FloatTensor]] = None, # Tag: Changed
        cross_attn_layer_head_mask: Optional[torch.Tensor] = None,
        past_key_value: Optional[Tuple[torch.Tensor]] = None,
        output_attentions: Optional[bool] = False,
        use_cache: Optional[bool] = True,
    ) -> 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`): attention mask of size
                `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
            encoder_hidden_states (`torch.FloatTensor`):
                cross attention input to the layer of shape `(batch, seq_len, embed_dim)`
            encoder_attention_mask (`torch.FloatTensor`): encoder attention mask of size
                `(batch, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
            layer_head_mask (`torch.FloatTensor`): mask for attention heads in a given layer of size
                `(encoder_attention_heads,)`.
            cross_attn_layer_head_mask (`torch.FloatTensor`): mask for cross-attention heads in a given layer of
                size `(decoder_attention_heads,)`.
            past_key_value (`Tuple(torch.FloatTensor)`): cached past key and value projection states
            output_attentions (`bool`, *optional*):
                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
                returned tensors for more detail.
        """
        residual = hidden_states

        # Self Attention
        # decoder uni-directional self-attention cached key/values tuple is at positions 1,2
        self_attn_past_key_value = past_key_value[:2] if past_key_value is not None else None
        # add present self-attn cache to positions 1,2 of present_key_value tuple
        hidden_states, self_attn_weights, present_key_value = self.self_attn(
            hidden_states=hidden_states,
            prefix=decoder_prefix,
            past_key_value=self_attn_past_key_value,
            attention_mask=attention_mask,
            layer_head_mask=layer_head_mask,
            output_attentions=output_attentions,
        )
        hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
        hidden_states = residual + hidden_states
        hidden_states = self.self_attn_layer_norm(hidden_states)

        # Cross-Attention Block
        cross_attn_present_key_value = None
        cross_attn_weights = None
        if encoder_hidden_states is not None:
            residual = hidden_states

            # cross_attn cached key/values tuple is at positions 3,4 of present_key_value tuple
            cross_attn_past_key_value = past_key_value[-2:] if past_key_value is not None else None
            hidden_states, cross_attn_weights, cross_attn_present_key_value = self.encoder_attn(
                hidden_states=hidden_states,
                prefix=cross_prefix,
                key_value_states=encoder_hidden_states,
                attention_mask=encoder_attention_mask,
                layer_head_mask=cross_attn_layer_head_mask,
                past_key_value=cross_attn_past_key_value,
                output_attentions=output_attentions,
            )
            hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
            hidden_states = residual + hidden_states
            hidden_states = self.encoder_attn_layer_norm(hidden_states)

            # add cross-attn to positions 3,4 of present_key_value tuple
            present_key_value = present_key_value + cross_attn_present_key_value

        # Fully Connected
        residual = hidden_states
        hidden_states = self.activation_fn(self.fc1(hidden_states))
        hidden_states = nn.functional.dropout(hidden_states, p=self.activation_dropout, training=self.training)
        hidden_states = self.fc2(hidden_states)
        hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
        hidden_states = residual + hidden_states
        hidden_states = self.final_layer_norm(hidden_states)

        outputs = (hidden_states,)

        if output_attentions:
            outputs += (self_attn_weights, cross_attn_weights)

        if use_cache:
            outputs += (present_key_value,)

        return outputs


BART_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 ([`BartConfig`]):
            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.
"""

BART_GENERATION_EXAMPLE = r"""
    Summarization example:
    ```python
    >>> from transformers import BartTokenizer, BartForConditionalGeneration
    >>> model = BartForConditionalGeneration.from_pretrained("facebook/bart-large-cnn")
    >>> tokenizer = BartTokenizer.from_pretrained("facebook/bart-large-cnn")
    >>> ARTICLE_TO_SUMMARIZE = (
    ...     "PG&E stated it scheduled the blackouts in response to forecasts for high winds "
    ...     "amid dry conditions. The aim is to reduce the risk of wildfires. Nearly 800 thousand customers were "
    ...     "scheduled to be affected by the shutoffs which were expected to last through at least midday tomorrow."
    ... )
    >>> inputs = tokenizer([ARTICLE_TO_SUMMARIZE], max_length=1024, return_tensors="pt")
    >>> # Generate Summary
    >>> summary_ids = model.generate(inputs["input_ids"], num_beams=2, max_length=20)
    >>> tokenizer.batch_decode(summary_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
    'PG&E scheduled the blackouts in response to forecasts for high winds amid dry conditions'
    ```
    Mask filling example:
    ```python
    >>> from transformers import BartTokenizer, BartForConditionalGeneration
    >>> tokenizer = BartTokenizer.from_pretrained("facebook/bart-base")
    >>> model = BartForConditionalGeneration.from_pretrained("facebook/bart-base")
    >>> TXT = "My friends are <mask> but they eat too many carbs."
    >>> input_ids = tokenizer([TXT], return_tensors="pt")["input_ids"]
    >>> logits = model(input_ids).logits
    >>> masked_index = (input_ids[0] == tokenizer.mask_token_id).nonzero().item()
    >>> probs = logits[0, masked_index].softmax(dim=0)
    >>> values, predictions = probs.topk(5)
    >>> tokenizer.decode(predictions).split()
    ['not', 'good', 'healthy', 'great', 'very']
    ```
"""

BART_INPUTS_DOCSTRING = r"""
    Args:
        input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
            Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you provide
            it.
            Indices can be obtained using [`BartTokenizer`]. See [`PreTrainedTokenizer.encode`] and
            [`PreTrainedTokenizer.__call__`] for details.
            [What are input IDs?](../glossary#input-ids)
        attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
            Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
            - 1 for tokens that are **not masked**,
            - 0 for tokens that are **masked**.
            [What are attention masks?](../glossary#attention-mask)
        decoder_input_ids (`torch.LongTensor` of shape `(batch_size, target_sequence_length)`, *optional*):
            Indices of decoder input sequence tokens in the vocabulary.
            Indices can be obtained using [`BartTokenizer`]. See [`PreTrainedTokenizer.encode`] and
            [`PreTrainedTokenizer.__call__`] for details.
            [What are decoder input IDs?](../glossary#decoder-input-ids)
            Bart uses the `eos_token_id` as the starting token for `decoder_input_ids` generation. If `past_key_values`
            is used, optionally only the last `decoder_input_ids` have to be input (see `past_key_values`).
            For translation and summarization training, `decoder_input_ids` should be provided. If no
            `decoder_input_ids` is provided, the model will create this tensor by shifting the `input_ids` to the right
            for denoising pre-training following the paper.
        decoder_attention_mask (`torch.LongTensor` of shape `(batch_size, target_sequence_length)`, *optional*):
            Default behavior: generate a tensor that ignores pad tokens in `decoder_input_ids`. Causal mask will also
            be used by default.
            If you want to change padding behavior, you should read [`modeling_bart._prepare_decoder_inputs`] and
            modify to your needs. See diagram 1 in [the paper](https://arxiv.org/abs/1910.13461) for more information
            on the default strategy.
        head_mask (`torch.Tensor` of shape `(encoder_layers, encoder_attention_heads)`, *optional*):
            Mask to nullify selected heads of the attention modules in the encoder. Mask values selected in `[0, 1]`:
            - 1 indicates the head is **not masked**,
            - 0 indicates the head is **masked**.
        decoder_head_mask (`torch.Tensor` of shape `(decoder_layers, decoder_attention_heads)`, *optional*):
            Mask to nullify selected heads of the attention modules in the decoder. Mask values selected in `[0, 1]`:
            - 1 indicates the head is **not masked**,
            - 0 indicates the head is **masked**.
        cross_attn_head_mask (`torch.Tensor` of shape `(decoder_layers, decoder_attention_heads)`, *optional*):
            Mask to nullify selected heads of the cross-attention modules in the decoder. Mask values selected in `[0,
            1]`:
            - 1 indicates the head is **not masked**,
            - 0 indicates the head is **masked**.
        encoder_outputs (`tuple(tuple(torch.FloatTensor)`, *optional*):
            Tuple consists of (`last_hidden_state`, *optional*: `hidden_states`, *optional*: `attentions`)
            `last_hidden_state` of shape `(batch_size, sequence_length, hidden_size)`, *optional*) is a sequence of
            hidden-states at the output of the last layer of the encoder. Used in the cross-attention of the decoder.
        past_key_values (`tuple(tuple(torch.FloatTensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`):
            Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of shape
            `(batch_size, num_heads, sequence_length, embed_size_per_head)`) and 2 additional tensors of shape
            `(batch_size, num_heads, encoder_sequence_length, embed_size_per_head)`.
            Contains pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention
            blocks) that can be used (see `past_key_values` input) to speed up sequential decoding.
            If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those that
            don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of all
            `decoder_input_ids` of shape `(batch_size, sequence_length)`. inputs_embeds (`torch.FloatTensor` of shape
            `(batch_size, sequence_length, hidden_size)`, *optional*): Optionally, instead of passing `input_ids` you
            can choose to directly pass an embedded representation. This is useful if you want more control over how to
            convert `input_ids` indices into associated vectors than the model's internal embedding lookup matrix.
        decoder_inputs_embeds (`torch.FloatTensor` of shape `(batch_size, target_sequence_length, hidden_size)`, *optional*):
            Optionally, instead of passing `decoder_input_ids` you can choose to directly pass an embedded
            representation. If `past_key_values` is used, optionally only the last `decoder_inputs_embeds` have to be
            input (see `past_key_values`). This is useful if you want more control over how to convert
            `decoder_input_ids` indices into associated vectors than the model's internal embedding lookup matrix.
            If `decoder_input_ids` and `decoder_inputs_embeds` are both unset, `decoder_inputs_embeds` takes the value
            of `inputs_embeds`.
        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.
"""

class PrefixGenBartEncoder(BartPretrainedModel):
    """
    Transformer encoder consisting of *config.encoder_layers* self attention layers. Each layer is a
    [`BartEncoderLayer`].
    Args:
        config: BartConfig
        embed_tokens (nn.Embedding): output embedding
    """

    def __init__(self, config: BartConfig, embed_tokens: Optional[nn.Embedding] = None):
        super().__init__(config)

        self.dropout = config.dropout
        self.layerdrop = config.encoder_layerdrop

        embed_dim = config.d_model
        self.padding_idx = config.pad_token_id
        self.max_source_positions = config.max_position_embeddings
        self.embed_scale = math.sqrt(embed_dim) if config.scale_embedding else 1.0

        if embed_tokens is not None:
            self.embed_tokens = embed_tokens
        else:
            self.embed_tokens = nn.Embedding(config.vocab_size, embed_dim, self.padding_idx)

        self.embed_positions = BartLearnedPositionalEmbedding(
            config.max_position_embeddings,
            embed_dim,
        )
        self.layers = nn.ModuleList([PrefixGenBartEncoderLayer(config) for _ in range(config.encoder_layers)])
        self.layernorm_embedding = nn.LayerNorm(embed_dim)

        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,
        encoder_prefix: Optional[Tuple[torch.Tensor]] = None,
        attention_mask: Optional[torch.Tensor] = None,
        head_mask: Optional[torch.Tensor] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        # **kwargs
    ) -> Union[Tuple, BaseModelOutput]:
        r"""
        Args:
            input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
                Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you
                provide it.
                Indices can be obtained using [`BartTokenizer`]. See [`PreTrainedTokenizer.encode`] and
                [`PreTrainedTokenizer.__call__`] for details.
                [What are input IDs?](../glossary#input-ids)
            attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
                Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
                - 1 for tokens that are **not masked**,
                - 0 for tokens that are **masked**.
                [What are attention masks?](../glossary#attention-mask)
            head_mask (`torch.Tensor` of shape `(encoder_layers, encoder_attention_heads)`, *optional*):
                Mask to nullify selected heads of the 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.
            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.
        """
        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

        # retrieve input_ids and inputs_embeds
        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:
            input_shape = input_ids.size()
            input_ids = input_ids.view(-1, input_shape[-1])
        elif inputs_embeds is not None:
            input_shape = inputs_embeds.size()[:-1]
        else:
            raise ValueError("You have to specify either input_ids or inputs_embeds")

        if inputs_embeds is None:
            inputs_embeds = self.embed_tokens(input_ids) * self.embed_scale

        embed_pos = self.embed_positions(input_shape)

        hidden_states = inputs_embeds + embed_pos
        hidden_states = self.layernorm_embedding(hidden_states)
        hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)

        # expand attention_mask
        if attention_mask is not None:
            # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len]
            attention_mask = _expand_mask(attention_mask, inputs_embeds.dtype)

        encoder_states = () if output_hidden_states else None
        all_attentions = () if output_attentions else None

        # check if head_mask has a correct number of layers specified if desired
        if head_mask is not None:
            if head_mask.size()[0] != (len(self.layers)):
                raise ValueError(
                    f"The head_mask should be specified for {len(self.layers)} layers, but it is for {head_mask.size()[0]}."
                )

        # check if prefix has a correct number of layers specified if desired
        if encoder_prefix is not None:
            if encoder_prefix[0].size()[1] != (len(self.layers)):
                raise ValueError(
                    f"The encoder prefix key should be specified for {len(self.layers)} layers, but it is for {encoder_prefix[0].size()[1]}."
                )
            if encoder_prefix[1].size()[1] != (len(self.layers)):
                raise ValueError(
                    f"The encoder prefix value should be specified for {len(self.layers)} layers, but it is for {encoder_prefix[1].size()[1]}."
                )

        for idx, encoder_layer in enumerate(self.layers):
            if output_hidden_states:
                encoder_states = encoder_states + (hidden_states,)
            # add LayerDrop (see https://arxiv.org/abs/1909.11556 for description)
            dropout_probability = random.uniform(0, 1)
            if self.training and (dropout_probability < self.layerdrop):  # skip the layer
                layer_outputs = (None, None)
            else:
                if self.gradient_checkpointing and self.training:

                    def create_custom_forward(module):
                        def custom_forward(*inputs):
                            return module(*inputs, output_attentions)

                        return custom_forward

                    layer_outputs = torch.utils.checkpoint.checkpoint(
                        create_custom_forward(encoder_layer),
                        hidden_states,
                        attention_mask,
                        head_mask[idx] if head_mask is not None else None,
                        (encoder_prefix[0][:,idx,:,:,:], encoder_prefix[1][:,idx,:,:,:]) if encoder_prefix is not None else None,
                    )
                else:
                    layer_outputs = encoder_layer(
                        hidden_states,
                        attention_mask,
                        layer_head_mask=head_mask[idx] if head_mask is not None else None,
                        prefix=(encoder_prefix[0][:,idx,:,:,:], encoder_prefix[1][:,idx,:,:,:]) if encoder_prefix is not None else None,
                        output_attentions=output_attentions,
                    )

                hidden_states = layer_outputs[0]

            if output_attentions:
                all_attentions = all_attentions + (layer_outputs[1],)

        if output_hidden_states:
            encoder_states = encoder_states + (hidden_states,)

        if not return_dict:
            return tuple(v for v in [hidden_states, encoder_states, all_attentions] if v is not None)
        return BaseModelOutput(
            last_hidden_state=hidden_states, hidden_states=encoder_states, attentions=all_attentions
        )


class PrefixGenBartDecoder(BartPretrainedModel):
    """
    Transformer decoder consisting of *config.decoder_layers* layers. Each layer is a [`BartDecoderLayer`]
    Args:
        config: BartConfig
        embed_tokens (nn.Embedding): output embedding
    """

    def __init__(self, config: BartConfig, embed_tokens: Optional[nn.Embedding] = None):
        super().__init__(config)
        self.dropout = config.dropout
        self.layerdrop = config.decoder_layerdrop
        self.padding_idx = config.pad_token_id
        self.max_target_positions = config.max_position_embeddings
        self.embed_scale = math.sqrt(config.d_model) if config.scale_embedding else 1.0

        if embed_tokens is not None:
            self.embed_tokens = embed_tokens
        else:
            self.embed_tokens = nn.Embedding(config.vocab_size, config.d_model, self.padding_idx)

        self.embed_positions = BartLearnedPositionalEmbedding(
            config.max_position_embeddings,
            config.d_model,
        )
        self.layers = nn.ModuleList([PrefixGenBartDecoderLayer(config) for _ in range(config.decoder_layers)])
        self.layernorm_embedding = nn.LayerNorm(config.d_model)

        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 _prepare_decoder_attention_mask(self, attention_mask, input_shape, inputs_embeds, past_key_values_length):
        # create causal mask
        # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len]
        combined_attention_mask = None
        if input_shape[-1] > 1:
            combined_attention_mask = _make_causal_mask(
                input_shape, inputs_embeds.dtype, past_key_values_length=past_key_values_length
            ).to(self.device)

        if attention_mask is not None:
            # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len]
            expanded_attn_mask = _expand_mask(attention_mask, inputs_embeds.dtype, tgt_len=input_shape[-1])
            combined_attention_mask = (
                expanded_attn_mask if combined_attention_mask is None else expanded_attn_mask + combined_attention_mask
            )

        return combined_attention_mask

    def forward(
        self,
        input_ids: torch.LongTensor = None,
        cross_prefix: Optional[Tuple[torch.Tensor]] = None,
        decoder_prefix: Optional[Tuple[torch.Tensor]] = None,
        attention_mask: Optional[torch.Tensor] = None,
        encoder_hidden_states: Optional[torch.FloatTensor] = None,
        encoder_attention_mask: Optional[torch.LongTensor] = None,
        head_mask: Optional[torch.Tensor] = None,
        cross_attn_head_mask: Optional[torch.Tensor] = None,
        past_key_values: Optional[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,
    ) -> Union[Tuple, BaseModelOutputWithPastAndCrossAttentions]:
        r"""
        Args:
            input_ids (`torch.LongTensor` of shape `(batch_size, sequence_length)`):
                Indices of input sequence tokens in the vocabulary. Padding will be ignored by default should you
                provide it.
                Indices can be obtained using [`BartTokenizer`]. See [`PreTrainedTokenizer.encode`] and
                [`PreTrainedTokenizer.__call__`] for details.
                [What are input IDs?](../glossary#input-ids)
            attention_mask (`torch.Tensor` of shape `(batch_size, sequence_length)`, *optional*):
                Mask to avoid performing attention on padding token indices. Mask values selected in `[0, 1]`:
                - 1 for tokens that are **not masked**,
                - 0 for tokens that are **masked**.
                [What are attention masks?](../glossary#attention-mask)
            encoder_hidden_states (`torch.FloatTensor` of shape `(batch_size, encoder_sequence_length, hidden_size)`, *optional*):
                Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention
                of the decoder.
            encoder_attention_mask (`torch.LongTensor` of shape `(batch_size, encoder_sequence_length)`, *optional*):
                Mask to avoid performing cross-attention on padding tokens indices of encoder input_ids. Mask values
                selected in `[0, 1]`:
                - 1 for tokens that are **not masked**,
                - 0 for tokens that are **masked**.
                [What are attention masks?](../glossary#attention-mask)
            head_mask (`torch.Tensor` of shape `(decoder_layers, decoder_attention_heads)`, *optional*):
                Mask to nullify selected heads of the attention modules. Mask values selected in `[0, 1]`:
                - 1 indicates the head is **not masked**,
                - 0 indicates the head is **masked**.
            cross_attn_head_mask (`torch.Tensor` of shape `(decoder_layers, decoder_attention_heads)`, *optional*):
                Mask to nullify selected heads of the cross-attention modules in the decoder to avoid performing
                cross-attention on hidden heads. Mask values selected in `[0, 1]`:
                - 1 indicates the head is **not masked**,
                - 0 indicates the head is **masked**.
            past_key_values (`tuple(tuple(torch.FloatTensor))`, *optional*, returned when `use_cache=True` is passed or when `config.use_cache=True`):
                Tuple of `tuple(torch.FloatTensor)` of length `config.n_layers`, with each tuple having 2 tensors of
                shape `(batch_size, num_heads, sequence_length, embed_size_per_head)`) and 2 additional tensors of
                shape `(batch_size, num_heads, encoder_sequence_length, embed_size_per_head)`.
                Contains pre-computed hidden-states (key and values in the self-attention blocks and in the
                cross-attention blocks) that can be used (see `past_key_values` input) to speed up sequential decoding.
                If `past_key_values` are used, the user can optionally input only the last `decoder_input_ids` (those
                that don't have their past key value states given to this model) of shape `(batch_size, 1)` instead of
                all `decoder_input_ids` of shape `(batch_size, sequence_length)`. inputs_embeds (`torch.FloatTensor` of
                shape `(batch_size, sequence_length, hidden_size)`, *optional*): Optionally, instead of passing
                `input_ids` you can choose to directly pass an embedded representation. This is useful if you want more
                control over how to convert `input_ids` indices into associated vectors than the model's internal
                embedding lookup matrix.
            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.
        """
        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 not None and inputs_embeds is not None:
            raise ValueError("You cannot specify both decoder_input_ids and decoder_inputs_embeds at the same time")
        elif input_ids is not None:
            input_shape = input_ids.size()
            input_ids = input_ids.view(-1, input_shape[-1])
        elif inputs_embeds is not None:
            input_shape = inputs_embeds.size()[:-1]
        else:
            raise ValueError("You have to specify either decoder_input_ids or decoder_inputs_embeds")

        # past_key_values_length
        past_key_values_length = past_key_values[0][0].shape[2] if past_key_values is not None else 0

        if inputs_embeds is None:
            inputs_embeds = self.embed_tokens(input_ids) * self.embed_scale

        attention_mask = self._prepare_decoder_attention_mask(
            attention_mask, input_shape, inputs_embeds, past_key_values_length
        )

        # expand encoder attention mask
        if encoder_hidden_states is not None and encoder_attention_mask is not None:
            # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len]
            encoder_attention_mask = _expand_mask(encoder_attention_mask, inputs_embeds.dtype, tgt_len=input_shape[-1])

        # embed positions
        positions = self.embed_positions(input_shape, past_key_values_length)

        hidden_states = inputs_embeds + positions
        hidden_states = self.layernorm_embedding(hidden_states)

        hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)

        # decoder layers
        all_hidden_states = () if output_hidden_states else None
        all_self_attns = () if output_attentions else None
        all_cross_attentions = () if (output_attentions and encoder_hidden_states is not None) else None
        next_decoder_cache = () if use_cache else None

        # check if head_mask/cross_attn_head_mask has a correct number of layers specified if desired
        for attn_mask, mask_name in zip([head_mask, cross_attn_head_mask], ["head_mask", "cross_attn_head_mask"]):
            if attn_mask is not None:
                if attn_mask.size()[0] != (len(self.layers)):
                    raise ValueError(
                        "The `{mask_name}` should be specified for {len(self.layers)} layers, but it is for {head_mask.size()[0]}."
                    )

        # check if prefix has a correct number of layers specified if desired
        if cross_prefix is not None:
            if cross_prefix[0].size()[1] != (len(self.layers)):
                raise ValueError(
                    f"The cross prefix key should be specified for {len(self.layers)} layers, but it is for {cross_prefix[0].size()[1]}."
                )
            if cross_prefix[1].size()[1] != (len(self.layers)):
                raise ValueError(
                    f"The cross prefix value should be specified for {len(self.layers)} layers, but it is for {cross_prefix[1].size()[1]}."
                )
        
        if decoder_prefix is not None:
            if decoder_prefix[0].size()[1] != (len(self.layers)):
                raise ValueError(
                    f"The decoder prefix key should be specified for {len(self.layers)} layers, but it is for {decoder_prefix[0].size()[1]}."
                )
            if decoder_prefix[1].size()[1] != (len(self.layers)):
                raise ValueError(
                    f"The decoder prefix value should be specified for {len(self.layers)} layers, but it is for {decoder_prefix[1].size()[1]}."
                )

        for idx, decoder_layer in enumerate(self.layers):
            # add LayerDrop (see https://arxiv.org/abs/1909.11556 for description)
            if output_hidden_states:
                all_hidden_states += (hidden_states,)
            dropout_probability = random.uniform(0, 1)
            if self.training and (dropout_probability < self.layerdrop):
                continue

            past_key_value = past_key_values[idx] if past_key_values is not None else None

            if self.gradient_checkpointing and self.training:

                if use_cache:
                    logger.warning(
                        "`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`..."
                    )
                    use_cache = False

                def create_custom_forward(module):
                    def custom_forward(*inputs):
                        # None for past_key_value
                        return module(*inputs, output_attentions, use_cache)

                    return custom_forward

                layer_outputs = torch.utils.checkpoint.checkpoint(
                    create_custom_forward(decoder_layer),
                    hidden_states,
                    attention_mask,
                    encoder_hidden_states,
                    encoder_attention_mask,
                    head_mask[idx] if head_mask is not None else None,
                    (cross_prefix[0][:,idx,:,:,:], cross_prefix[1][:,idx,:,:,:]) if cross_prefix is not None else None,
                    (decoder_prefix[0][:,idx,:,:,:], decoder_prefix[1][:,idx,:,:,:]) if decoder_prefix is not None else None,
                    cross_attn_head_mask[idx] if cross_attn_head_mask is not None else None,
                    None,
                )
            else:

                layer_outputs = decoder_layer(
                    hidden_states,
                    attention_mask=attention_mask,
                    encoder_hidden_states=encoder_hidden_states,
                    encoder_attention_mask=encoder_attention_mask,
                    layer_head_mask=(head_mask[idx] if head_mask is not None else None),
                    cross_prefix=(cross_prefix[0][:,idx,:,:,:], cross_prefix[1][:,idx,:,:,:]) if cross_prefix is not None else None,
                    decoder_prefix=(decoder_prefix[0][:,idx,:,:,:], decoder_prefix[1][:,idx,:,:,:]) if decoder_prefix is not None else None,
                    cross_attn_layer_head_mask=(
                        cross_attn_head_mask[idx] if cross_attn_head_mask is not None else None
                    ),
                    past_key_value=past_key_value,
                    output_attentions=output_attentions,
                    use_cache=use_cache,
                )
            hidden_states = layer_outputs[0]

            if use_cache:
                next_decoder_cache += (layer_outputs[3 if output_attentions else 1],)

            if output_attentions:
                all_self_attns += (layer_outputs[1],)

                if encoder_hidden_states is not None:
                    all_cross_attentions += (layer_outputs[2],)

        # 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 not return_dict:
            return tuple(
                v
                for v in [hidden_states, next_cache, all_hidden_states, all_self_attns, all_cross_attentions]
                if v is not None
            )
        return BaseModelOutputWithPastAndCrossAttentions(
            last_hidden_state=hidden_states,
            past_key_values=next_cache,
            hidden_states=all_hidden_states,
            attentions=all_self_attns,
            cross_attentions=all_cross_attentions,
        )


@add_start_docstrings(
    "The bare BART Model outputting raw hidden-states without any specific head on top.",
    BART_START_DOCSTRING,
)
class PrefixGenBartModel(BartPretrainedModel):
    def __init__(self, config: BartConfig):
        super().__init__(config)

        padding_idx, vocab_size = config.pad_token_id, config.vocab_size
        self.shared = nn.Embedding(vocab_size, config.d_model, padding_idx)

        self.encoder = PrefixGenBartEncoder(config, self.shared)
        self.decoder = PrefixGenBartDecoder(config, self.shared)

        self.use_encoder_prefix = config.use_encoder_prefix
        self.use_cross_prefix = config.use_cross_prefix
        self.use_decoder_prefix = config.use_decoder_prefix

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

    def get_input_embeddings(self):
        return self.shared

    def set_input_embeddings(self, value):
        self.shared = value
        self.encoder.embed_tokens = self.shared
        self.decoder.embed_tokens = self.shared

    def get_encoder(self):
        return self.encoder

    def get_decoder(self):
        return self.decoder

    @add_start_docstrings_to_model_forward(BART_INPUTS_DOCSTRING)
    @add_code_sample_docstrings(
        processor_class=_TOKENIZER_FOR_DOC,
        checkpoint=_CHECKPOINT_FOR_DOC,
        output_type=Seq2SeqModelOutput,
        config_class=_CONFIG_FOR_DOC,
        expected_output=_EXPECTED_OUTPUT_SHAPE,
    )
    def forward(
        self,
        input_ids: torch.LongTensor = None,
        prefix: Optional[Dict] = None,   # Tag: Changed
        attention_mask: Optional[torch.Tensor] = None,
        decoder_input_ids: Optional[torch.LongTensor] = None,
        decoder_attention_mask: Optional[torch.LongTensor] = None,
        head_mask: Optional[torch.Tensor] = None,
        decoder_head_mask: Optional[torch.Tensor] = None,
        cross_attn_head_mask: Optional[torch.Tensor] = None,
        encoder_outputs: Optional[List[torch.FloatTensor]] = None,
        past_key_values: Optional[List[torch.FloatTensor]] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        decoder_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,
    ) -> Union[Tuple, Seq2SeqModelOutput]:

        # different to other models, Bart automatically creates decoder_input_ids from
        # input_ids if no decoder_input_ids are provided
        if decoder_input_ids is None and decoder_inputs_embeds is None:
            if input_ids is None:
                raise ValueError(
                    "If no `decoder_input_ids` or `decoder_inputs_embeds` are "
                    "passed, `input_ids` cannot be `None`. Please pass either "
                    "`input_ids` or `decoder_input_ids` or `decoder_inputs_embeds`."
                )

            decoder_input_ids = shift_tokens_right(
                input_ids, self.config.pad_token_id, self.config.decoder_start_token_id
            )

        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 encoder_outputs is None:
            encoder_outputs = self.encoder(
                input_ids=input_ids,
                encoder_prefix=prefix['encoder_prefix'] if self.use_encoder_prefix else None,
                attention_mask=attention_mask,
                head_mask=head_mask,
                inputs_embeds=inputs_embeds,
                output_attentions=output_attentions,
                output_hidden_states=output_hidden_states,
                return_dict=return_dict,
            )
        # If the user passed a tuple for encoder_outputs, we wrap it in a BaseModelOutput when return_dict=True
        elif return_dict and not isinstance(encoder_outputs, BaseModelOutput):
            encoder_outputs = BaseModelOutput(
                last_hidden_state=encoder_outputs[0],
                hidden_states=encoder_outputs[1] if len(encoder_outputs) > 1 else None,
                attentions=encoder_outputs[2] if len(encoder_outputs) > 2 else None,
            )

        # decoder outputs consists of (dec_features, past_key_value, dec_hidden, dec_attn)
        decoder_outputs = self.decoder(
            input_ids=decoder_input_ids,
            cross_prefix=prefix['cross_prefix'] if self.use_cross_prefix else None,
            decoder_prefix=prefix['decoder_prefix'] if self.use_decoder_prefix else None,
            attention_mask=decoder_attention_mask,
            encoder_hidden_states=encoder_outputs[0],
            encoder_attention_mask=attention_mask,
            head_mask=decoder_head_mask,
            cross_attn_head_mask=cross_attn_head_mask,
            past_key_values=past_key_values,
            inputs_embeds=decoder_inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        if not return_dict:
            return decoder_outputs + encoder_outputs

        return Seq2SeqModelOutput(
            last_hidden_state=decoder_outputs.last_hidden_state,
            past_key_values=decoder_outputs.past_key_values,
            decoder_hidden_states=decoder_outputs.hidden_states,
            decoder_attentions=decoder_outputs.attentions,
            cross_attentions=decoder_outputs.cross_attentions,
            encoder_last_hidden_state=encoder_outputs.last_hidden_state,
            encoder_hidden_states=encoder_outputs.hidden_states,
            encoder_attentions=encoder_outputs.attentions,
        )


@add_start_docstrings(
    "The BART Model with a language modeling head. Can be used for summarization.", BART_START_DOCSTRING
)
class PrefixGenBartForConditionalGeneration(BartPretrainedModel):
    base_model_prefix = "model"
    _keys_to_ignore_on_load_missing = [r"final_logits_bias", r"lm_head\.weight"]

    def __init__(self, config: BartConfig):
        super().__init__(config)
        self.model = PrefixGenBartModel(config)
        self.register_buffer("final_logits_bias", torch.zeros((1, self.model.shared.num_embeddings)))
        self.lm_head = nn.Linear(config.d_model, self.model.shared.num_embeddings, bias=False)

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

    def get_encoder(self):
        return self.model.get_encoder()

    def get_decoder(self):
        return self.model.get_decoder()

    def resize_token_embeddings(self, new_num_tokens: int) -> nn.Embedding:
        new_embeddings = super().resize_token_embeddings(new_num_tokens)
        self._resize_final_logits_bias(new_num_tokens)
        return new_embeddings

    def _resize_final_logits_bias(self, new_num_tokens: int) -> None:
        old_num_tokens = self.final_logits_bias.shape[-1]
        if new_num_tokens <= old_num_tokens:
            new_bias = self.final_logits_bias[:, :new_num_tokens]
        else:
            extra_bias = torch.zeros((1, new_num_tokens - old_num_tokens), device=self.final_logits_bias.device)
            new_bias = torch.cat([self.final_logits_bias, extra_bias], dim=1)
        self.register_buffer("final_logits_bias", new_bias)

    def get_output_embeddings(self):
        return self.lm_head

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

    @add_start_docstrings_to_model_forward(BART_INPUTS_DOCSTRING)
    @replace_return_docstrings(output_type=Seq2SeqLMOutput, config_class=_CONFIG_FOR_DOC)
    @add_end_docstrings(BART_GENERATION_EXAMPLE)
    def forward(
        self,
        input_ids: torch.LongTensor = None,
        prefix: Optional[Dict] = None,   # Tag: Changed
        attention_mask: Optional[torch.Tensor] = None,
        decoder_input_ids: Optional[torch.LongTensor] = None,
        decoder_attention_mask: Optional[torch.LongTensor] = None,
        head_mask: Optional[torch.Tensor] = None,
        decoder_head_mask: Optional[torch.Tensor] = None,
        cross_attn_head_mask: Optional[torch.Tensor] = None,
        encoder_outputs: Optional[List[torch.FloatTensor]] = None,
        past_key_values: Optional[List[torch.FloatTensor]] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        decoder_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, Seq2SeqLMOutput]:
        r"""
        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:
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        if labels is not None:
            if use_cache:
                logger.warning("The `use_cache` argument is changed to `False` since `labels` is provided.")
            use_cache = False
            if decoder_input_ids is None and decoder_inputs_embeds is None:
                decoder_input_ids = shift_tokens_right(
                    labels, self.config.pad_token_id, self.config.decoder_start_token_id
                )

        outputs = self.model(
            input_ids,
            prefix=prefix,
            attention_mask=attention_mask,
            decoder_input_ids=decoder_input_ids,
            encoder_outputs=encoder_outputs,
            decoder_attention_mask=decoder_attention_mask,
            head_mask=head_mask,
            decoder_head_mask=decoder_head_mask,
            cross_attn_head_mask=cross_attn_head_mask,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            decoder_inputs_embeds=decoder_inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )
        lm_logits = self.lm_head(outputs[0]) + self.final_logits_bias

        masked_lm_loss = None
        if labels is not None:
            loss_fct = CrossEntropyLoss()
            masked_lm_loss = loss_fct(lm_logits.view(-1, self.config.vocab_size), labels.view(-1))

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

        return Seq2SeqLMOutput(
            loss=masked_lm_loss,
            logits=lm_logits,
            past_key_values=outputs.past_key_values,
            decoder_hidden_states=outputs.decoder_hidden_states,
            decoder_attentions=outputs.decoder_attentions,
            cross_attentions=outputs.cross_attentions,
            encoder_last_hidden_state=outputs.encoder_last_hidden_state,
            encoder_hidden_states=outputs.encoder_hidden_states,
            encoder_attentions=outputs.encoder_attentions,
        )

    # Tag: Changed
    def _prepare_encoder_decoder_kwargs_for_generation(
        self, inputs_tensor: torch.Tensor, model_kwargs, model_input_name: Optional[str] = None
    ) -> Dict[str, Any]:
        # 1. get encoder
        encoder = self.get_encoder()

        # 2. prepare encoder args and encoder kwargs from model kwargs
        irrelevant_prefix = ["decoder_", "cross_attn", "use_cache", "prefix"]
        encoder_kwargs = {
            argument: value
            for argument, value in model_kwargs.items()
            if not any(argument.startswith(p) for p in irrelevant_prefix)
        }

        # 3. make sure that encoder returns `ModelOutput`
        model_input_name = model_input_name if model_input_name is not None else self.main_input_name
        encoder_kwargs["return_dict"] = True
        encoder_kwargs[model_input_name] = inputs_tensor
        model_kwargs["encoder_outputs"]: ModelOutput = encoder(**encoder_kwargs)

        return model_kwargs

    # Tag: Changed
    def prepare_inputs_for_generation(
        self,
        decoder_input_ids,
        past=None,
        attention_mask=None,
        head_mask=None,
        decoder_head_mask=None,
        cross_attn_head_mask=None,
        use_cache=None,
        encoder_outputs=None,
        **kwargs
    ):
        # cut decoder_input_ids if past is used
        if past is not None:
            decoder_input_ids = decoder_input_ids[:, -1:]
        return {
            "input_ids": None,  # encoder_outputs is defined. input_ids not needed
            "encoder_outputs": encoder_outputs,
            "prefix": kwargs.get('prefix', None),
            "past_key_values": past,
            "decoder_input_ids": decoder_input_ids,
            "attention_mask": attention_mask,
            "head_mask": head_mask,
            "decoder_head_mask": decoder_head_mask,
            "cross_attn_head_mask": cross_attn_head_mask,
            "use_cache": use_cache,  # change this to avoid caching (presumably for debugging)
        }

    def prepare_decoder_input_ids_from_labels(self, labels: torch.Tensor):
        return shift_tokens_right(labels, self.config.pad_token_id, self.config.decoder_start_token_id)

    @staticmethod
    def _reorder_cache(past, beam_idx):
        reordered_past = ()
        for layer_past in past:
            # cached cross_attention states don't have to be reordered -> they are always the same
            reordered_past += (
                tuple(past_state.index_select(0, beam_idx) for past_state in layer_past[:2]) + layer_past[2:],
            )
        return reordered_past

# for prefix controlled generation

from transformers.adapters import PrefixTuningConfig
from transformers import BertModel

class PrefixGenBartForPrefixControledGeneration(BartPretrainedModel):
    base_model_prefix = "model"
    _keys_to_ignore_on_load_missing = [r"final_logits_bias", r"lm_head\.weight"]

    def __init__(self, config: BartConfig, n_layers=12, #12, 6 for encoder and 6 for decoder
        n_heads=12, #16
        input_size=768,
        prefix_config=PrefixTuningConfig):
        super().__init__(config)
        self.model = PrefixGenBartModel(config)
        self.register_buffer("final_logits_bias", torch.zeros((1, self.model.shared.num_embeddings)))
        self.lm_head = nn.Linear(config.d_model, self.model.shared.num_embeddings, bias=False)

        # Initialize weights and apply final processing
        #self.prefix_model = BertModel.from_pretrained("bert-base-cased").to(self.model.device)
        self.prefix_wte = nn.Embedding(self.prefix_length, self.input_size)
        self.prefix_config = prefix_config

        self.n_layers = n_layers
        self.n_heads = n_heads
        self.input_size = input_size
        self.n_embd_per_head = self.input_size // self.n_heads
        self.prefix_length = 30

        self.linear = nn.Linear(self.input_size, self.prefix_length)
        self.wte = nn.Embedding(self.prefix_length, self.input_size)
        self.control_trans = nn.Sequential(
            nn.Linear(self.input_size, self.prefix_config.bottleneck_size),
            nn.Tanh(),
            nn.Linear(self.prefix_config.bottleneck_size, self.n_layers * 2 * self.input_size),
        )
        self.dropout = nn.Dropout(self.prefix_config.dropout)
        # config.num_beams = 4
        self.num_beams = config.num_beams
        # self.n_gpus = 1 #model_args.n_gpu
    
        self.post_init()

    def get_encoder(self):
        return self.model.get_encoder()

    def get_decoder(self):
        return self.model.get_decoder()

    def resize_token_embeddings(self, new_num_tokens: int) -> nn.Embedding:
        new_embeddings = super().resize_token_embeddings(new_num_tokens)
        self._resize_final_logits_bias(new_num_tokens)
        return new_embeddings

    def _resize_final_logits_bias(self, new_num_tokens: int) -> None:
        old_num_tokens = self.final_logits_bias.shape[-1]
        if new_num_tokens <= old_num_tokens:
            new_bias = self.final_logits_bias[:, :new_num_tokens]
        else:
            extra_bias = torch.zeros((1, new_num_tokens - old_num_tokens), device=self.final_logits_bias.device)
            new_bias = torch.cat([self.final_logits_bias, extra_bias], dim=1)
        self.register_buffer("final_logits_bias", new_bias)

    def get_output_embeddings(self):
        return self.lm_head

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

    @add_start_docstrings_to_model_forward(BART_INPUTS_DOCSTRING)
    @replace_return_docstrings(output_type=Seq2SeqLMOutput, config_class=_CONFIG_FOR_DOC)
    @add_end_docstrings(BART_GENERATION_EXAMPLE)
    def forward(
        self,
        input_ids: torch.LongTensor = None,
        prefix: Optional[Dict] = None,   # Tag: Changed
        attention_mask: Optional[torch.Tensor] = None,
        decoder_input_ids: Optional[torch.LongTensor] = None,
        decoder_attention_mask: Optional[torch.LongTensor] = None,
        head_mask: Optional[torch.Tensor] = None,
        decoder_head_mask: Optional[torch.Tensor] = None,
        cross_attn_head_mask: Optional[torch.Tensor] = None,
        encoder_outputs: Optional[List[torch.FloatTensor]] = None,
        past_key_values: Optional[List[torch.FloatTensor]] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        decoder_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, Seq2SeqLMOutput]:
        r"""
        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:
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        if labels is not None:
            if use_cache:
                logger.warning("The `use_cache` argument is changed to `False` since `labels` is provided.")
            use_cache = False
            if decoder_input_ids is None and decoder_inputs_embeds is None:
                decoder_input_ids = shift_tokens_right(
                    labels, self.config.pad_token_id, self.config.decoder_start_token_id
                )
        
        if input_ids:
            prefix_ids, input_ids = torch.split(input_ids, 1, dim = 1)
            prefix_ids = prefix_ids.squeeze(1)
            input_ids = input_ids.squeeze(1)
            # print("prefix ids", prefix_ids.size())
            # print('prefix ids', prefix_ids.device)

            #prefix_inputs = self.prefix_model(prefix_ids.to(self.prefix_model.device)).last_hidden_state # size [batch_size, seq_len, 768]
            prefix_inputs = self.prefix_wte(prefix_ids)

            #print('prefix inputs', prefix_inputs.device)
            prefix = prefix_inputs.mean(dim=1).to(torch.long) #.cuda(non_blocking=True) # size [batch_size, 768]
            # prefix = self.prefixtuning_model(prefix_input = prefix.to(self.prefixtuning_model.device))

            # self.batch_size = prefix.size()[0]
            input_tokens = prefix.to(torch.float32)
            input_tokens = self.linear(input_tokens)
            embs = self.wte(input_tokens.to(torch.long))
            key_values = self.control_trans(embs)  # batch_size x prefix_length x n_layers*2*input_size
            key_values = key_values.view(
                math.ceil(prefix.size()[0] / self.n_gpus), self.prefix_length, self.n_layers * 2, self.n_heads, self.n_embd_per_head
            )  # *2 for key and value
            key_values = self.dropout(key_values)
            # n_layers * (2 x batch_size x n_heads x prefix_length x n_embd_per_head)
            prefix_dict = {} # need: (bz, n_encoder_layers, prefix_len, head_num, dim)
            key_values = key_values.permute(0, 2, 1, 3, 4).split(self.n_layers, dim = 1) # before split: (batch_size, prefix_length, n_layers * 2, n_heads, n_embd)
            prefix_dict['encoder_prefix'] = key_values[0].split(self.n_layers // 2, dim = 1)
            # print(prefix_dict['encoder_prefix'][0].size())
            prefix_dict['cross_prefix'] = key_values[0].split(self.n_layers // 2, dim = 1)
            prefix_dict['decoder_prefix'] = key_values[1].split(self.n_layers // 2, dim = 1)

        outputs = self.model(
            input_ids,
            prefix=self.prefix_dict,
            attention_mask=attention_mask,
            decoder_input_ids=decoder_input_ids,
            encoder_outputs=encoder_outputs,
            decoder_attention_mask=decoder_attention_mask,
            head_mask=head_mask,
            decoder_head_mask=decoder_head_mask,
            cross_attn_head_mask=cross_attn_head_mask,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            decoder_inputs_embeds=decoder_inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        lm_logits = self.lm_head(outputs[0]) + self.final_logits_bias

        masked_lm_loss = None
        if labels is not None:
            loss_fct = CrossEntropyLoss()
            masked_lm_loss = loss_fct(lm_logits.view(-1, self.config.vocab_size), labels.view(-1).to(lm_logits.device))

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

        return Seq2SeqLMOutput(
            loss=masked_lm_loss,
            logits=lm_logits,
            past_key_values=outputs.past_key_values,
            decoder_hidden_states=outputs.decoder_hidden_states,
            decoder_attentions=outputs.decoder_attentions,
            cross_attentions=outputs.cross_attentions,
            encoder_last_hidden_state=outputs.encoder_last_hidden_state,
            encoder_hidden_states=outputs.encoder_hidden_states,
            encoder_attentions=outputs.encoder_attentions,
        )

    # Tag: Changed
    def _prepare_encoder_decoder_kwargs_for_generation(
        self, inputs_tensor: torch.Tensor, model_kwargs, model_input_name: Optional[str] = None
    ) -> Dict[str, Any]:
        
        # 1. get encoder
        encoder = self.get_encoder()

        # 2. prepare encoder args and encoder kwargs from model kwargs
        irrelevant_prefix = ["decoder_", "cross_attn", "use_cache"] #, "prefix"]
        encoder_kwargs = {
            argument: value
            for argument, value in model_kwargs.items()
            if not any(argument.startswith(p) for p in irrelevant_prefix)
        }

        # 3. make sure that encoder returns `ModelOutput`
        model_input_name = model_input_name if model_input_name is not None else self.main_input_name
        encoder_kwargs["return_dict"] = True

        prefix_ids, input_ids = torch.split(inputs_tensor, 1, dim = 1)
        prefix_ids = prefix_ids.squeeze(1)
        input_ids = input_ids.squeeze(1)
        # print("prefix ids", prefix_ids.size())
        # print('prefix ids', prefix_ids.device)
        prefix_inputs = self.prefix_model(prefix_ids.to(self.prefix_model.device)).last_hidden_state # size [batch_size, seq_len, 768]
        #print('prefix inputs', prefix_inputs.device)
        prefix = prefix_inputs.mean(dim=1).to(torch.long) #.cuda(non_blocking=True) # size [batch_size, 768]

        # self.batch_size = prefix.size()[0]
        input_tokens = prefix.to(torch.float32)
        input_tokens = self.linear(input_tokens)
        embs = self.wte(input_tokens.to(torch.long))
        key_values = self.control_trans(embs)  # batch_size x prefix_length x n_layers*2*input_size
        key_values = key_values.view(
            math.ceil(prefix.size()[0] / self.n_gpus), self.prefix_length, self.n_layers * 2, self.n_heads, self.n_embd_per_head
        )  # *2 for key and value
        key_values = self.dropout(key_values)
        # n_layers * (2 x batch_size x n_heads x prefix_length x n_embd_per_head)
        prefix_dict = {} # need: (bz, n_encoder_layers, prefix_len, head_num, dim)
        key_values = key_values.permute(0, 2, 1, 3, 4).split(self.n_layers, dim = 1) # before split: (batch_size, prefix_length, n_layers * 2, n_heads, n_embd)
        
        #print(self.num_beams)

        if self.num_beams > 1:
            encoder_prefix = key_values[0]
            expanded_enc_return_idx = (torch.arange(encoder_prefix.shape[0]).view(-1, 1).repeat(1, self.num_beams).view(-1))
            # print(self.num_beams)
            encoder_prefix = encoder_prefix.index_select(0, expanded_enc_return_idx)
            #print('encoder prefix 2', encoder_prefix.size())
            encoder_prefix = encoder_prefix.split(self.n_layers // 2, dim = 1)
            prefix_dict['encoder_prefix'] = encoder_prefix
            prefix_dict['cross_prefix'] = encoder_prefix

            decoder_prefix = key_values[1]
            expanded_dec_return_idx = (torch.arange(decoder_prefix.shape[0]).view(-1, 1).repeat(1, self.num_beams).view(-1))
            decoder_prefix = decoder_prefix.index_select(0, expanded_dec_return_idx)
            #print('decoder prefix 2', decoder_prefix.size())
            decoder_prefix = decoder_prefix.split(self.n_layers // 2, dim = 1)
            prefix_dict['decoder_prefix'] = decoder_prefix
            #prefix_dict['decoder_prefix'] = key_values[1].split(self.n_layers // 2, dim = 1)

        else:
            prefix_dict['encoder_prefix'] = key_values[0].split(self.n_layers // 2, dim = 1)
            #print('encoder prefix 2', prefix_dict['encoder_prefix'][0].size())
            prefix_dict['cross_prefix'] = key_values[0].split(self.n_layers // 2, dim = 1)
            prefix_dict['decoder_prefix'] = key_values[1].split(self.n_layers // 2, dim = 1)

        encoder_kwargs[model_input_name] = input_ids
        self.prefix_dict = prefix_dict
        encoder_kwargs["encoder_prefix"] = prefix_dict['encoder_prefix']
        model_kwargs["encoder_outputs"]: ModelOutput = encoder(**encoder_kwargs)
        self.encoder_outputs = model_kwargs["encoder_outputs"]
        # print('encoder outputss', self.encoder_outputs[0].size())

        return model_kwargs

    # Tag: Changed
    def prepare_inputs_for_generation(
        self,
        decoder_input_ids,
        past=None,
        attention_mask=None,
        head_mask=None,
        decoder_head_mask=None,
        cross_attn_head_mask=None,
        use_cache=None,
        encoder_outputs=None,
        **kwargs
    ):
        # cut decoder_input_ids if past is used
        # print('decoder idss', decoder_input_ids.size())
        if past is not None:
            decoder_input_ids = decoder_input_ids[:, -1:]
        # print('encoder outputss222', self.encoder_outputs[0].size())
        
        return {
            "input_ids": None,  # encoder_outputs is defined. input_ids not needed
            "encoder_outputs": encoder_outputs,
            "prefix": kwargs.get('prefix', None),
            "past_key_values": past,
            "decoder_input_ids": decoder_input_ids,
            "attention_mask": attention_mask,
            "head_mask": head_mask,
            "decoder_head_mask": decoder_head_mask,
            "cross_attn_head_mask": cross_attn_head_mask,
            "use_cache": use_cache,  # change this to avoid caching (presumably for debugging)
        }

    def prepare_decoder_input_ids_from_labels(self, labels: torch.Tensor):
        return shift_tokens_right(labels, self.config.pad_token_id, self.config.decoder_start_token_id)

    @staticmethod
    def _reorder_cache(past, beam_idx):
        reordered_past = ()
        for layer_past in past:
            # cached cross_attention states don't have to be reordered -> they are always the same
            reordered_past += (
                tuple(past_state.index_select(0, beam_idx) for past_state in layer_past[:2]) + layer_past[2:],
            )
        return reordered_past

from transformers.adapters import PrefixTuningConfig
from transformers import BertModel

class PrefixGenBartForGraphPrefixControledGeneration(BartPretrainedModel):
    base_model_prefix = "model"
    _keys_to_ignore_on_load_missing = [r"final_logits_bias", r"lm_head\.weight"]

    def __init__(self, config: BartConfig, n_layers=12, #12, 6 for encoder and 6 for decoder
        n_heads=12, #16
        input_size=768,
        prefix_config=PrefixTuningConfig):
        super().__init__(config)
        self.model = PrefixGenBartModel(config)
        self.register_buffer("final_logits_bias", torch.zeros((1, self.model.shared.num_embeddings)))
        self.lm_head = nn.Linear(config.d_model, self.model.shared.num_embeddings, bias=False)

        # Initialize weights and apply final processing
        self.prefix_model = BertModel.from_pretrained("bert-base-cased").to(self.model.device)
        self.prefix_config = prefix_config

        self.n_layers = n_layers
        self.n_heads = n_heads
        self.input_size = input_size
        self.n_embd_per_head = self.input_size // self.n_heads
        self.prefix_length = 30

        self.prefix_control_trans = nn.Sequential(
            nn.Linear(self.input_size, self.prefix_config.bottleneck_size),
            nn.Tanh(),
            nn.Linear(self.prefix_config.bottleneck_size, 3 * self.input_size),
        )
        self.softmax = nn.Softmax(dim = 0)

        self.linear = nn.Linear(self.input_size, self.prefix_length)
        self.wte = nn.Embedding(self.prefix_length, self.input_size)
        self.control_trans = nn.Sequential(
            nn.Linear(self.input_size, self.prefix_config.bottleneck_size),
            nn.Tanh(),
            nn.Linear(self.prefix_config.bottleneck_size, self.n_layers * 2 * self.input_size),
        )
        self.dropout = nn.Dropout(self.prefix_config.dropout)
        # config.num_beams = 4
        self.num_beams = config.num_beams
        # self.n_gpus = model_args.n_gpu
    
        self.post_init()

    def get_encoder(self):
        return self.model.get_encoder()

    def get_decoder(self):
        return self.model.get_decoder()

    def resize_token_embeddings(self, new_num_tokens: int) -> nn.Embedding:
        new_embeddings = super().resize_token_embeddings(new_num_tokens)
        self._resize_final_logits_bias(new_num_tokens)
        return new_embeddings

    def _resize_final_logits_bias(self, new_num_tokens: int) -> None:
        old_num_tokens = self.final_logits_bias.shape[-1]
        if new_num_tokens <= old_num_tokens:
            new_bias = self.final_logits_bias[:, :new_num_tokens]
        else:
            extra_bias = torch.zeros((1, new_num_tokens - old_num_tokens), device=self.final_logits_bias.device)
            new_bias = torch.cat([self.final_logits_bias, extra_bias], dim=1)
        self.register_buffer("final_logits_bias", new_bias)

    def get_output_embeddings(self):
        return self.lm_head

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

    @add_start_docstrings_to_model_forward(BART_INPUTS_DOCSTRING)
    @replace_return_docstrings(output_type=Seq2SeqLMOutput, config_class=_CONFIG_FOR_DOC)
    @add_end_docstrings(BART_GENERATION_EXAMPLE)
    def forward(
        self,
        input_ids: torch.LongTensor = None,
        prefix: Optional[Dict] = None,   # Tag: Changed
        attention_mask: Optional[torch.Tensor] = None,
        decoder_input_ids: Optional[torch.LongTensor] = None,
        decoder_attention_mask: Optional[torch.LongTensor] = None,
        head_mask: Optional[torch.Tensor] = None,
        decoder_head_mask: Optional[torch.Tensor] = None,
        cross_attn_head_mask: Optional[torch.Tensor] = None,
        encoder_outputs: Optional[List[torch.FloatTensor]] = None,
        past_key_values: Optional[List[torch.FloatTensor]] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        decoder_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, Seq2SeqLMOutput]:
        r"""
        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:
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        if labels is not None:
            if use_cache:
                logger.warning("The `use_cache` argument is changed to `False` since `labels` is provided.")
            use_cache = False
            if decoder_input_ids is None and decoder_inputs_embeds is None:
                decoder_input_ids = shift_tokens_right(
                    labels, self.config.pad_token_id, self.config.decoder_start_token_id
                )
        batch_size = input_ids.size()[0]
        max_len = input_ids.size()[2]
        
        #if input_ids != None:
        attn_matrices, d_matrices, prefix_ids, input_ids = torch.split(input_ids, [max_len, max_len, 1, 1], dim = 1)
        attn_matrices = attn_matrices.to(torch.float32).squeeze(1).unsqueeze(-1).expand(attn_matrices.size()[0],attn_matrices.size()[1],attn_matrices.size()[2], self.input_size)
        d_matrices = d_matrices.to(torch.float32).squeeze(1).unsqueeze(-1).expand(d_matrices.size()[0],d_matrices.size()[1],d_matrices.size()[2], self.input_size)
        input_ids = input_ids.squeeze(1)
        #print('input id', input_ids.size())
        prefix_ids = prefix_ids.squeeze(1)
        # print(input_ids)
        # print('input_ids', input_ids.size(), type(input_ids)) #batch size, seq_len
        prefix_inputs = self.prefix_model(prefix_ids.to(self.prefix_model.device)).last_hidden_state # size [batch_size, 1, seq_len, 768]

        prefix_inputs = prefix_inputs.unsqueeze(1)
        prefix_keys = self.prefix_control_trans(prefix_inputs).permute(0,2,1,3)  # batch_size x seq_len x 1 x 3 * input_size
        prefix_query, prefix_key, prefix_value = prefix_keys.split(self.input_size, dim = -1)
        prefix_key = torch.transpose(prefix_key, 1, 2)
        # batch_size x seq_len x seq_len x 3 * input_size

        dk = self.input_size # math.sqrt(prefix_value.shape[0])
        p = self.softmax(torch.einsum('bwnh, bwlh -> bwnh', torch.div(torch.einsum('bwlh, blnh-> bwnh', prefix_query, prefix_key.to(prefix_query.device)), dk) + attn_matrices.to(prefix_query.device), prefix_value.to(prefix_query.device)))
        # p = softmax(torch.matmul(torch.div(torch.matmul(query, key), dk) + torch.tensor(attn_matrix).float(), value), dim = 0)

        z = torch.sum(torch.div(p, d_matrices.to(p.device)), dim = 2).unsqueeze(2).expand(batch_size, max_len, max_len, self.input_size)
        #revised self attention matrix
        new_attention_matrices = torch.div(p, torch.mul(z, d_matrices.to(z.device)).to(p.device))

        # prefix_inputs = self.prefix_model(prefix_ids.to(self.prefix_model.device)).last_hidden_state # size [batch_size, seq_len, 768]
        prefix_inputs = torch.einsum('blwh, bwnh -> blnh', prefix_inputs, new_attention_matrices.to(prefix_inputs.device))
        prefix = prefix_inputs.mean(dim=2).mean(dim=1).to(torch.long) #.cuda(non_blocking=True) # size [batch_size, 768]

        # self.batch_size = prefix.size()[0]
        input_tokens = prefix.to(torch.float32)
        input_tokens = self.linear(input_tokens)

        embs = self.wte(input_tokens.to(torch.long))
        
        key_values = self.control_trans(embs)  # batch_size x prefix_length x n_layers*2*input_size
        key_values = key_values.view(
            #math.ceil(prefix.size()[0] / self.n_gpus), self.prefix_length, self.n_layers * 2, self.n_heads, self.n_embd_per_head
            math.ceil(prefix.size()[0]), self.prefix_length, self.n_layers * 2, self.n_heads, self.n_embd_per_head
        )  # *2 for key and value
        key_values = self.dropout(key_values)
        # n_layers * (2 x batch_size x n_heads x prefix_length x n_embd_per_head)
        prefix_dict = {} # need: (bz, n_encoder_layers, prefix_len, head_num, dim)
        key_values = key_values.permute(0, 2, 1, 3, 4).to(self.model.device).split(self.n_layers, dim = 1) # before split: (batch_size, prefix_length, n_layers * 2, n_heads, n_embd)

        prefix_dict['encoder_prefix'] = key_values[0].split(self.n_layers // 2, dim = 1)
        # print(prefix_dict['encoder_prefix'][0].size())
        prefix_dict['cross_prefix'] = key_values[0].split(self.n_layers // 2, dim = 1)
        prefix_dict['decoder_prefix'] = key_values[1].split(self.n_layers // 2, dim = 1)

        outputs = self.model(
            input_ids.to(self.model.device),
            prefix=prefix_dict,
            attention_mask=attention_mask.to(self.model.device),
            decoder_input_ids=decoder_input_ids.to(self.model.device),
            encoder_outputs=encoder_outputs,
            decoder_attention_mask=decoder_attention_mask,
            head_mask=head_mask,
            decoder_head_mask=decoder_head_mask,
            cross_attn_head_mask=cross_attn_head_mask,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            decoder_inputs_embeds=decoder_inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        lm_logits = self.lm_head(outputs[0]) + self.final_logits_bias

        masked_lm_loss = None
        if labels is not None:
            loss_fct = CrossEntropyLoss()
            masked_lm_loss = loss_fct(lm_logits.view(-1, self.config.vocab_size), labels.view(-1).to(lm_logits.device))

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

        return Seq2SeqLMOutput(
            loss=masked_lm_loss,
            logits=lm_logits,
            past_key_values=outputs.past_key_values,
            decoder_hidden_states=outputs.decoder_hidden_states,
            decoder_attentions=outputs.decoder_attentions,
            cross_attentions=outputs.cross_attentions,
            encoder_last_hidden_state=outputs.encoder_last_hidden_state,
            encoder_hidden_states=outputs.encoder_hidden_states,
            encoder_attentions=outputs.encoder_attentions,
        )

    # Tag: Changed
    def _prepare_encoder_decoder_kwargs_for_generation(
        self, inputs_tensor: torch.Tensor, model_kwargs, model_input_name: Optional[str] = None
    ) -> Dict[str, Any]:
        
        # 1. get encoder
        encoder = self.get_encoder()

        # 2. prepare encoder args and encoder kwargs from model kwargs
        irrelevant_prefix = ["decoder_", "cross_attn", "use_cache"] #, "prefix"]
        encoder_kwargs = {
            argument: value
            for argument, value in model_kwargs.items()
            if not any(argument.startswith(p) for p in irrelevant_prefix)
        }

        # 3. make sure that encoder returns `ModelOutput`
        model_input_name = model_input_name if model_input_name is not None else self.main_input_name
        encoder_kwargs["return_dict"] = True

        prefix_ids, input_ids = torch.split(inputs_tensor, 1, dim = 1)
        prefix_ids = prefix_ids.squeeze(1)
        input_ids = input_ids.squeeze(1)
        # print("prefix ids", prefix_ids.size())
        # print('prefix ids', prefix_ids.device)
        prefix_inputs = self.prefix_model(prefix_ids.to(self.prefix_model.device)).last_hidden_state # size [batch_size, seq_len, 768]
        #print('prefix inputs', prefix_inputs.device)
        prefix = prefix_inputs.mean(dim=1).to(torch.long) #.cuda(non_blocking=True) # size [batch_size, 768]

        # self.batch_size = prefix.size()[0]
        input_tokens = prefix.to(torch.float32)
        input_tokens = self.linear(input_tokens)
        embs = self.wte(input_tokens.to(torch.long))
        key_values = self.control_trans(embs)  # batch_size x prefix_length x n_layers*2*input_size
        key_values = key_values.view(
            math.ceil(prefix.size()[0] / self.n_gpus), self.prefix_length, self.n_layers * 2, self.n_heads, self.n_embd_per_head
        )  # *2 for key and value
        key_values = self.dropout(key_values)
        # n_layers * (2 x batch_size x n_heads x prefix_length x n_embd_per_head)
        prefix_dict = {} # need: (bz, n_encoder_layers, prefix_len, head_num, dim)
        key_values = key_values.permute(0, 2, 1, 3, 4).split(self.n_layers, dim = 1) # before split: (batch_size, prefix_length, n_layers * 2, n_heads, n_embd)
        
        #print(self.num_beams)

        if self.num_beams > 1:
            encoder_prefix = key_values[0]
            expanded_enc_return_idx = (torch.arange(encoder_prefix.shape[0]).view(-1, 1).repeat(1, self.num_beams).view(-1))
            # print(self.num_beams)
            encoder_prefix = encoder_prefix.index_select(0, expanded_enc_return_idx)
            #print('encoder prefix 2', encoder_prefix.size())
            encoder_prefix = encoder_prefix.split(self.n_layers // 2, dim = 1)
            prefix_dict['encoder_prefix'] = encoder_prefix
            prefix_dict['cross_prefix'] = encoder_prefix

            decoder_prefix = key_values[1]
            expanded_dec_return_idx = (torch.arange(decoder_prefix.shape[0]).view(-1, 1).repeat(1, self.num_beams).view(-1))
            decoder_prefix = decoder_prefix.index_select(0, expanded_dec_return_idx)
            #print('decoder prefix 2', decoder_prefix.size())
            decoder_prefix = decoder_prefix.split(self.n_layers // 2, dim = 1)
            prefix_dict['decoder_prefix'] = decoder_prefix
            #prefix_dict['decoder_prefix'] = key_values[1].split(self.n_layers // 2, dim = 1)

        else:
            prefix_dict['encoder_prefix'] = key_values[0].split(self.n_layers // 2, dim = 1)
            #print('encoder prefix 2', prefix_dict['encoder_prefix'][0].size())
            prefix_dict['cross_prefix'] = key_values[0].split(self.n_layers // 2, dim = 1)
            prefix_dict['decoder_prefix'] = key_values[1].split(self.n_layers // 2, dim = 1)

        encoder_kwargs[model_input_name] = input_ids
        self.prefix_dict = prefix_dict
        encoder_kwargs["encoder_prefix"] = prefix_dict['encoder_prefix']
        model_kwargs["encoder_outputs"]: ModelOutput = encoder(**encoder_kwargs)
        self.encoder_outputs = model_kwargs["encoder_outputs"]
        # print('encoder outputss', self.encoder_outputs[0].size())

        return model_kwargs

    # Tag: Changed
    def prepare_inputs_for_generation(
        self,
        decoder_input_ids,
        past=None,
        attention_mask=None,
        head_mask=None,
        decoder_head_mask=None,
        cross_attn_head_mask=None,
        use_cache=None,
        encoder_outputs=None,
        **kwargs
    ):
        # cut decoder_input_ids if past is used
        # print('decoder idss', decoder_input_ids.size())
        if past is not None:
            decoder_input_ids = decoder_input_ids[:, -1:]
        # print('encoder outputss222', self.encoder_outputs[0].size())
        
        return {
            "input_ids": None,  # encoder_outputs is defined. input_ids not needed
            "encoder_outputs": encoder_outputs,
            "prefix": kwargs.get('prefix', None),
            "past_key_values": past,
            "decoder_input_ids": decoder_input_ids,
            "attention_mask": attention_mask,
            "head_mask": head_mask,
            "decoder_head_mask": decoder_head_mask,
            "cross_attn_head_mask": cross_attn_head_mask,
            "use_cache": use_cache,  # change this to avoid caching (presumably for debugging)
        }

    def prepare_decoder_input_ids_from_labels(self, labels: torch.Tensor):
        return shift_tokens_right(labels, self.config.pad_token_id, self.config.decoder_start_token_id)

    @staticmethod
    def _reorder_cache(past, beam_idx):
        reordered_past = ()
        for layer_past in past:
            # cached cross_attention states don't have to be reordered -> they are always the same
            reordered_past += (
                tuple(past_state.index_select(0, beam_idx) for past_state in layer_past[:2]) + layer_past[2:],
            )
        return reordered_past