AutoAndroidController/py/venv/Lib/site-packages/transformers/models/biogpt/modeling_biogpt.py

#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
#           This file was automatically generated from src/transformers/models/biogpt/modular_biogpt.py.
#               Do NOT edit this file manually as any edits will be overwritten by the generation of
#             the file from the modular. If any change should be done, please apply the change to the
#                          modular_biogpt.py file directly. One of our CI enforces this.
#                🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨🚨
# coding=utf-8
# Copyright 2022 The HuggingFace Team and Microsoft Research AI4Science All rights reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import math
from typing import Callable, Optional, Union

import torch
import torch.nn as nn
from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss

from ...activations import ACT2FN
from ...cache_utils import Cache, DynamicCache, EncoderDecoderCache
from ...generation import GenerationMixin
from ...modeling_attn_mask_utils import AttentionMaskConverter
from ...modeling_flash_attention_utils import FlashAttentionKwargs
from ...modeling_layers import GradientCheckpointingLayer
from ...modeling_outputs import (
    BaseModelOutputWithPastAndCrossAttentions,
    CausalLMOutputWithCrossAttentions,
    SequenceClassifierOutputWithPast,
    TokenClassifierOutput,
)
from ...modeling_utils import ALL_ATTENTION_FUNCTIONS, PreTrainedModel
from ...processing_utils import Unpack
from ...utils import TransformersKwargs, auto_docstring, is_torch_flex_attn_available, logging
from ...utils.deprecation import deprecate_kwarg
from .configuration_biogpt import BioGptConfig


if is_torch_flex_attn_available():
    from ...integrations.flex_attention import BlockMask, make_flex_block_causal_mask


logger = logging.get_logger(__name__)


class BioGptLearnedPositionalEmbedding(nn.Embedding):
    """
    This module learns positional embeddings up to a fixed maximum size.
    """

    def __init__(self, num_embeddings: int, embedding_dim: int):
        # BIOGPT is set up so that if padding_idx is specified then offset the embedding ids by 2
        # and adjust num_embeddings appropriately. Other models don't have this hack
        self.offset = 2
        super().__init__(num_embeddings + self.offset, embedding_dim)

    def forward(
        self,
        attention_mask: torch.LongTensor,
        past_key_values_length: int = 0,
        position_ids: Optional[torch.LongTensor] = None,
    ):
        """`input_ids_shape` is expected to be [bsz x seqlen]."""

        if position_ids is None:
            position_ids = torch.cumsum(attention_mask, dim=1)
            position_ids = (position_ids * attention_mask - 1).long()
            # cut positions if `past_key_values_length` is > 0
            position_ids = position_ids[:, past_key_values_length:]

        return super().forward(position_ids + self.offset)


class BioGptScaledWordEmbedding(nn.Embedding):
    """
    This module overrides nn.Embeddings' forward by multiplying with embeddings scale.
    """

    def __init__(self, num_embeddings: int, embedding_dim: int, padding_idx: int, embed_scale: Optional[float] = 1.0):
        super().__init__(num_embeddings, embedding_dim, padding_idx)
        self.embed_scale = embed_scale

    def forward(self, input_ids: torch.Tensor):
        return super().forward(input_ids) * self.embed_scale


def eager_attention_forward(
    module: nn.Module,
    query: torch.Tensor,
    key: torch.Tensor,
    value: torch.Tensor,
    attention_mask: Optional[torch.Tensor],
    scaling: Optional[float] = None,
    dropout: float = 0.0,
    head_mask: Optional[torch.Tensor] = None,
    **kwargs,
):
    if scaling is None:
        scaling = query.size(-1) ** -0.5

    attn_weights = torch.matmul(query, key.transpose(2, 3)) * scaling
    if attention_mask is not None:
        attn_weights = attn_weights + attention_mask

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

    if head_mask is not None:
        attn_weights = attn_weights * head_mask.view(1, -1, 1, 1)

    attn_weights = nn.functional.dropout(attn_weights, p=dropout, training=module.training)
    attn_output = torch.matmul(attn_weights, value)
    attn_output = attn_output.transpose(1, 2).contiguous()

    return attn_output, attn_weights


class BioGptAttention(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,
        is_causal: bool = False,
        config: Optional[BioGptConfig] = None,
        layer_idx: Optional[int] = None,
    ):
        super().__init__()
        self.embed_dim = embed_dim
        self.num_heads = num_heads
        self.dropout = dropout
        self.head_dim = embed_dim // num_heads
        self.config = config

        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.is_causal = is_causal
        self.layer_idx = layer_idx
        if layer_idx is None and self.is_decoder:
            logger.warning_once(
                f"Instantiating a decoder {self.__class__.__name__} without passing `layer_idx` is not recommended and "
                "will lead to errors during the forward call, if caching is used. Please make sure to provide a `layer_idx` "
                "when creating this class."
            )

        self.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)

    @deprecate_kwarg("past_key_value", new_name="past_key_values", version="4.58")
    def forward(
        self,
        hidden_states: torch.Tensor,
        key_value_states: Optional[torch.Tensor] = None,
        past_key_values: Optional[Cache] = None,
        attention_mask: Optional[torch.Tensor] = None,
        layer_head_mask: Optional[torch.Tensor] = None,
        output_attentions: bool = False,
        cache_position: Optional[torch.Tensor] = None,
        # TODO: we need a refactor so that the different attention modules can get their specific kwargs
        # ATM, we have mixed things encoder, decoder, and encoder-decoder attn
        **kwargs: Unpack[FlashAttentionKwargs],
    ) -> 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

        # determine input shapes
        bsz, tgt_len = hidden_states.shape[:-1]
        src_len = key_value_states.shape[1] if is_cross_attention else tgt_len

        q_input_shape = (bsz, tgt_len, -1, self.head_dim)
        kv_input_shape = (bsz, src_len, -1, self.head_dim)

        # get query proj
        query_states = self.q_proj(hidden_states).view(*q_input_shape).transpose(1, 2)

        is_updated = False
        if past_key_values is not None:
            if isinstance(past_key_values, EncoderDecoderCache):
                is_updated = past_key_values.is_updated.get(self.layer_idx)
                if is_cross_attention:
                    # after the first generated id, we can subsequently re-use all key/value_states from cache
                    curr_past_key_value = past_key_values.cross_attention_cache
                else:
                    curr_past_key_value = past_key_values.self_attention_cache
            else:
                curr_past_key_value = past_key_values

        current_states = key_value_states if is_cross_attention else hidden_states
        if is_cross_attention and past_key_values is not None and is_updated:
            # reuse k,v, cross_attentions
            key_states = curr_past_key_value.layers[self.layer_idx].keys
            value_states = curr_past_key_value.layers[self.layer_idx].values
        else:
            key_states = self.k_proj(current_states)
            value_states = self.v_proj(current_states)
            key_states = key_states.view(*kv_input_shape).transpose(1, 2)
            value_states = value_states.view(*kv_input_shape).transpose(1, 2)

            if past_key_values is not None:
                # save all key/value_states to cache to be re-used for fast auto-regressive generation
                cache_position = cache_position if not is_cross_attention else None
                key_states, value_states = curr_past_key_value.update(
                    key_states, value_states, self.layer_idx, {"cache_position": cache_position}
                )
                # set flag that curr layer for cross-attn is already updated so we can re-use in subsequent calls
                if is_cross_attention and isinstance(past_key_values, EncoderDecoderCache):
                    past_key_values.is_updated[self.layer_idx] = True

        attention_interface: Callable = eager_attention_forward
        if self.config._attn_implementation != "eager":
            attention_interface = ALL_ATTENTION_FUNCTIONS[self.config._attn_implementation]

        attn_output, attn_weights = attention_interface(
            self,
            query_states,
            key_states,
            value_states,
            attention_mask,
            dropout=0.0 if not self.training else self.dropout,
            scaling=self.scaling,
            output_attentions=output_attentions,
            head_mask=layer_head_mask,
            **kwargs,
        )

        attn_output = attn_output.reshape(bsz, tgt_len, -1).contiguous()
        attn_output = self.out_proj(attn_output)

        return attn_output, attn_weights


class BioGptDecoderLayer(GradientCheckpointingLayer):
    def __init__(self, config: BioGptConfig, layer_idx: Optional[int] = None):
        super().__init__()
        self.embed_dim = config.hidden_size

        self.self_attn = BioGptAttention(
            embed_dim=self.embed_dim,
            num_heads=config.num_attention_heads,
            dropout=config.attention_probs_dropout_prob,
            is_decoder=True,
            is_causal=True,
            config=config,
            layer_idx=layer_idx,
        )
        self.dropout = config.hidden_dropout_prob
        self.activation_fn = ACT2FN[config.hidden_act]
        self.activation_dropout = config.activation_dropout

        self.self_attn_layer_norm = nn.LayerNorm(self.embed_dim)

        self.fc1 = nn.Linear(self.embed_dim, config.intermediate_size)
        self.fc2 = nn.Linear(config.intermediate_size, self.embed_dim)
        self.final_layer_norm = nn.LayerNorm(self.embed_dim)

    @deprecate_kwarg("past_key_value", new_name="past_key_values", version="4.58")
    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
        layer_head_mask: Optional[torch.Tensor] = None,
        past_key_values: Optional[Cache] = None,
        output_attentions: Optional[bool] = False,
        use_cache: Optional[bool] = True,
        position_ids: Optional[torch.LongTensor] = None,
        cache_position: Optional[torch.Tensor] = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> 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.
            layer_head_mask (`torch.FloatTensor`): mask for attention heads in a given layer of size
                `(encoder_attention_heads,)`.
            past_key_values (`Cache`): 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.
            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`).
            cache_position (`torch.LongTensor` of shape `(sequence_length)`, *optional*):
                Indices depicting the position of the input sequence tokens in the sequence. It is used to update the
                cache in the correct position and to infer the complete sequence length.
        """
        residual = hidden_states

        hidden_states = self.self_attn_layer_norm(hidden_states)

        # Self Attention
        hidden_states, self_attn_weights = self.self_attn(
            hidden_states=hidden_states,
            past_key_values=past_key_values,
            attention_mask=attention_mask,
            layer_head_mask=layer_head_mask,
            output_attentions=output_attentions,
            position_ids=position_ids,
            cache_position=cache_position,
            **kwargs,
        )
        hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
        hidden_states = residual + hidden_states

        # Fully Connected
        residual = hidden_states
        hidden_states = self.final_layer_norm(hidden_states)
        hidden_states = self.fc1(hidden_states)
        hidden_states = self.activation_fn(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

        outputs = (hidden_states,)

        if output_attentions:
            outputs += (self_attn_weights,)

        return outputs


@auto_docstring
class BioGptPreTrainedModel(PreTrainedModel):
    config: BioGptConfig
    base_model_prefix = "biogpt"
    supports_gradient_checkpointing = True
    _supports_flash_attn = True
    _supports_sdpa = True
    _supports_flex_attn = True

    _can_compile_fullgraph = True

    # Copied from transformers.models.bart.modeling_bart.BartPreTrainedModel._update_causal_mask
    def _update_causal_mask(
        self,
        attention_mask: Optional[Union[torch.Tensor, "BlockMask"]],
        input_tensor: torch.Tensor,
        cache_position: torch.Tensor,
        past_key_values: Cache,
    ):
        if self.config._attn_implementation == "flex_attention":
            if isinstance(attention_mask, torch.Tensor):
                attention_mask = make_flex_block_causal_mask(attention_mask)
            # Other attention flavors support in-built causal (when `mask is None`)
            # while we need to create our specific block mask regardless
            elif attention_mask is None:
                attention_mask = make_flex_block_causal_mask(
                    torch.ones(
                        size=(input_tensor.shape[0], input_tensor.shape[1]),
                        device=attention_mask.device,
                    )
                )
            return attention_mask

        if self.config._attn_implementation == "flash_attention_2":
            if attention_mask is not None and (attention_mask == 0.0).any():
                return attention_mask
            return None

        # For SDPA, when possible, we will rely on its `is_causal` argument instead of its `attn_mask` argument, in
        # order to dispatch on Flash Attention 2. This feature is not compatible with static cache, as SDPA will fail
        # to infer the attention mask.
        past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
        using_compilable_cache = past_key_values.is_compileable if past_key_values is not None else False

        # When output attentions is True, sdpa implementation's forward method calls the eager implementation's forward
        if self.config._attn_implementation == "sdpa" and not using_compilable_cache:
            if AttentionMaskConverter._ignore_causal_mask_sdpa(
                attention_mask,
                inputs_embeds=input_tensor,
                past_key_values_length=past_seen_tokens,
                is_training=self.training,
            ):
                return None

        dtype = input_tensor.dtype
        sequence_length = input_tensor.shape[1]
        if using_compilable_cache:
            target_length = past_key_values.get_max_cache_shape()
        else:
            target_length = (
                attention_mask.shape[-1]
                if isinstance(attention_mask, torch.Tensor)
                else past_seen_tokens + sequence_length + 1
            )

        # In case the provided `attention` mask is 2D, we generate a causal mask here (4D).
        causal_mask = self._prepare_4d_causal_attention_mask_with_cache_position(
            attention_mask,
            sequence_length=sequence_length,
            target_length=target_length,
            dtype=dtype,
            cache_position=cache_position,
            batch_size=input_tensor.shape[0],
        )

        if (
            self.config._attn_implementation == "sdpa"
            and attention_mask is not None
            and attention_mask.device.type in ["cuda", "xpu", "npu"]
        ):
            # Attend to all tokens in fully masked rows in the causal_mask, for example the relevant first rows when
            # using left padding. This is required by F.scaled_dot_product_attention memory-efficient attention path.
            # Details: https://github.com/pytorch/pytorch/issues/110213
            min_dtype = torch.finfo(dtype).min
            causal_mask = AttentionMaskConverter._unmask_unattended(causal_mask, min_dtype)

        return causal_mask

    @staticmethod
    # Copied from transformers.models.gptj.modeling_gptj.GPTJModel._prepare_4d_causal_attention_mask_with_cache_position
    def _prepare_4d_causal_attention_mask_with_cache_position(
        attention_mask: torch.Tensor,
        sequence_length: int,
        target_length: int,
        dtype: torch.dtype,
        cache_position: torch.Tensor,
        batch_size: int,
        **kwargs,
    ):
        """
        Creates a causal 4D mask of shape `(batch_size, 1, query_length, key_value_length)` from a 2D mask of shape
        `(batch_size, key_value_length)`, or if the input `attention_mask` is already 4D, do nothing.

        Args:
            attention_mask (`torch.Tensor`):
                A 2D attention mask of shape `(batch_size, key_value_length)` or a 4D attention mask of shape
                `(batch_size, 1, query_length, key_value_length)`.
            sequence_length (`int`):
                The sequence length being processed.
            target_length (`int`):
                The target length: when generating with static cache, the mask should be as long as the static cache,
                to account for the 0 padding, the part of the cache that is not filled yet.
            dtype (`torch.dtype`):
                The dtype to use for the 4D attention mask.
            cache_position (`torch.Tensor`):
                Indices depicting the position of the input sequence tokens in the sequence.
            batch_size (`torch.Tensor`):
                Batch size.
        """
        if attention_mask is not None and attention_mask.dim() == 4:
            # In this case we assume that the mask comes already in inverted form and requires no inversion or slicing.
            causal_mask = attention_mask
        else:
            min_dtype = torch.finfo(dtype).min
            causal_mask = torch.full(
                (sequence_length, target_length), fill_value=min_dtype, dtype=dtype, device=cache_position.device
            )
            if sequence_length != 1:
                causal_mask = torch.triu(causal_mask, diagonal=1)
            causal_mask *= torch.arange(target_length, device=cache_position.device) > cache_position.reshape(-1, 1)
            causal_mask = causal_mask[None, None, :, :].expand(batch_size, 1, -1, -1)
            if attention_mask is not None:
                causal_mask = causal_mask.clone()  # copy to contiguous memory for in-place edit
                mask_length = attention_mask.shape[-1]
                padding_mask = causal_mask[:, :, :, :mask_length] + attention_mask[:, None, None, :].to(
                    causal_mask.device
                )
                padding_mask = padding_mask == 0
                causal_mask[:, :, :, :mask_length] = causal_mask[:, :, :, :mask_length].masked_fill(
                    padding_mask, min_dtype
                )

        return causal_mask


@auto_docstring
class BioGptModel(BioGptPreTrainedModel):
    def __init__(self, config: BioGptConfig):
        super().__init__(config)
        self.config = config
        self.layerdrop = config.layerdrop
        self.dropout = config.hidden_dropout_prob
        self.embed_dim = config.hidden_size
        self.padding_idx = config.pad_token_id
        embed_scale = math.sqrt(config.hidden_size) if config.scale_embedding else 1.0

        self.embed_tokens = BioGptScaledWordEmbedding(
            config.vocab_size, self.embed_dim, self.padding_idx, embed_scale=embed_scale
        )
        self.embed_positions = BioGptLearnedPositionalEmbedding(config.max_position_embeddings, self.embed_dim)

        self.layers = nn.ModuleList([BioGptDecoderLayer(config, layer_idx=i) for i in range(config.num_hidden_layers)])
        self.layer_norm = nn.LayerNorm(self.embed_dim)

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

    @auto_docstring
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        past_key_values: Optional[Cache] = None,
        use_cache: Optional[bool] = None,
        position_ids: Optional[torch.LongTensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        cache_position: Optional[torch.Tensor] = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> Union[tuple, BaseModelOutputWithPastAndCrossAttentions]:
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        use_cache = use_cache if use_cache is not None else self.config.use_cache
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

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

        if inputs_embeds is None:
            inputs_embeds = self.embed_tokens(input)

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

        # initialize past_key_values
        if use_cache and past_key_values is None:
            past_key_values = DynamicCache(config=self.config)
        if use_cache and isinstance(past_key_values, tuple):
            logger.warning_once(
                "Passing a tuple of `past_key_values` is deprecated and will be removed in Transformers v4.58.0. "
                "You should pass an instance of `DynamicCache` instead, e.g. "
                "`past_key_values=DynamicCache.from_legacy_cache(past_key_values)`."
            )
            past_key_values = DynamicCache.from_legacy_cache(past_key_values)

        batch_size, seq_length = inputs_embeds.size()[:-1]
        past_key_values_length = past_key_values.get_seq_length() if past_key_values is not None else 0
        if cache_position is None:
            cache_position = torch.arange(
                past_key_values_length, past_key_values_length + seq_length, device=inputs_embeds.device
            )

        if attention_mask is None:
            # required mask seq length can be calculated via length of past cache
            mask_seq_length = past_key_values_length + seq_length
            attention_mask = torch.ones(batch_size, mask_seq_length, device=inputs_embeds.device)

        self_attn_cache = past_key_values

        causal_mask = self._update_causal_mask(
            attention_mask,
            inputs_embeds,
            cache_position,
            self_attn_cache,
        )

        # embed positions
        if position_ids is None:
            # position_ids = cache_position.unsqueeze(0)
            position_ids = torch.cumsum(attention_mask, dim=1)
            position_ids = (position_ids * attention_mask - 1).long()
            # cut positions if `past_seen_tokens` is > 0
            position_ids = position_ids[:, past_key_values_length:]

        positions = self.embed_positions(attention_mask, past_key_values_length, position_ids=position_ids)
        hidden_states = inputs_embeds + positions
        hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)

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

        all_hidden_states = () if output_hidden_states else None
        all_self_attns = () if output_attentions else None
        all_cross_attentions = None

        for idx, decoder_layer in enumerate(self.layers):
            # add LayerDrop (see https://huggingface.co/papers/1909.11556 for description)
            if output_hidden_states:
                all_hidden_states += (hidden_states,)
            if self.training:
                dropout_probability = torch.rand([])
                if dropout_probability < self.layerdrop:
                    continue

            layer_outputs = decoder_layer(
                hidden_states,
                attention_mask=causal_mask,
                layer_head_mask=(head_mask[idx] if head_mask is not None else None),
                past_key_values=past_key_values,
                output_attentions=output_attentions,
                use_cache=use_cache,
                position_ids=position_ids,
                cache_position=cache_position,
                **kwargs,
            )

            hidden_states = layer_outputs[0]

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

        # add hidden states from the last decoder layer
        if output_hidden_states:
            all_hidden_states += (hidden_states,)

        hidden_states = self.layer_norm(hidden_states)

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


@auto_docstring(
    custom_intro="""
    BioGPT Model with a `language modeling` head on top for CLM fine-tuning.
    """
)
class BioGptForCausalLM(BioGptPreTrainedModel, GenerationMixin):
    _tied_weights_keys = ["output_projection.weight"]

    def __init__(self, config):
        super().__init__(config)

        self.biogpt = BioGptModel(config)
        self.output_projection = nn.Linear(config.hidden_size, config.vocab_size, bias=False)

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

    def get_output_embeddings(self):
        return self.output_projection

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

    @auto_docstring
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        past_key_values: Optional[Cache] = None,
        labels: Optional[torch.LongTensor] = None,
        use_cache: Optional[bool] = None,
        position_ids: Optional[torch.LongTensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        cache_position: Optional[torch.Tensor] = None,
        **kwargs: Unpack[TransformersKwargs],
    ) -> Union[tuple, CausalLMOutputWithCrossAttentions]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
            Labels for language modeling. Note that the labels **are shifted** inside the model, i.e. you can set
            `labels = input_ids` Indices are selected in `[-100, 0, ..., config.vocab_size]` All labels set to `-100`
            are ignored (masked), the loss is only computed for labels in `[0, ..., config.vocab_size]`
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        outputs = self.biogpt(
            input_ids,
            attention_mask=attention_mask,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            past_key_values=past_key_values,
            use_cache=use_cache,
            position_ids=position_ids,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
            cache_position=cache_position,
            **kwargs,
        )

        sequence_output = outputs[0]
        prediction_scores = self.output_projection(sequence_output)

        lm_loss = None
        if labels is not None:
            lm_loss = self.loss_function(
                prediction_scores,
                labels,
                vocab_size=self.config.vocab_size,
                **kwargs,
            )

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

        return CausalLMOutputWithCrossAttentions(
            loss=lm_loss,
            logits=prediction_scores,
            past_key_values=outputs.past_key_values,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
            cross_attentions=outputs.cross_attentions,
        )


@auto_docstring
class BioGptForTokenClassification(BioGptPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels

        self.biogpt = BioGptModel(config)
        if hasattr(config, "classifier_dropout") and config.classifier_dropout is not None:
            classifier_dropout = config.classifier_dropout
        else:
            classifier_dropout = config.hidden_dropout_prob
        self.dropout = nn.Dropout(classifier_dropout)
        self.classifier = nn.Linear(config.hidden_size, config.num_labels)

        self.post_init()

    @auto_docstring
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        token_type_ids: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        past_key_values: Optional[Cache] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        labels: Optional[torch.LongTensor] = None,
        use_cache: Optional[bool] = None,
        position_ids: Optional[torch.LongTensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        cache_position: Optional[torch.Tensor] = None,
    ) -> Union[tuple, TokenClassifierOutput]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

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

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

        loss = None
        if labels is not None:
            loss_fct = CrossEntropyLoss()
            # Only keep active parts of the loss
            if attention_mask is not None:
                active_loss = attention_mask.view(-1) == 1
                active_logits = logits.view(-1, self.num_labels)
                active_labels = torch.where(
                    active_loss, labels.view(-1), torch.tensor(loss_fct.ignore_index).type_as(labels)
                )
                loss = loss_fct(active_logits, active_labels)
            else:
                loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))

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

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


@auto_docstring(
    custom_intro="""
    The BioGpt Model transformer with a sequence classification head on top (linear layer).

    [`BioGptForSequenceClassification`] uses the last token in order to do the classification, as other causal models
    (e.g. GPT-2) do.

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

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

    @auto_docstring
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        head_mask: Optional[torch.FloatTensor] = None,
        past_key_values: Optional[Cache] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        labels: Optional[torch.LongTensor] = None,
        use_cache: Optional[bool] = None,
        position_ids: Optional[torch.LongTensor] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        cache_position: Optional[torch.Tensor] = None,
        logits_to_keep: Union[int, torch.Tensor] = 0,
    ) -> Union[tuple, SequenceClassifierOutputWithPast]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for computing the sequence classification/regression loss. Indices should be in `[0, ...,
            config.num_labels - 1]`. If `config.num_labels == 1` a regression loss is computed (Mean-Square loss), If
            `config.num_labels > 1` a classification loss is computed (Cross-Entropy).
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        transformer_outputs = self.biogpt(
            input_ids,
            past_key_values=past_key_values,
            attention_mask=attention_mask,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            position_ids=position_ids,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
            cache_position=cache_position,
        )
        hidden_states = transformer_outputs[0]
        slice_indices = slice(-logits_to_keep, None) if isinstance(logits_to_keep, int) else logits_to_keep
        logits = self.score(hidden_states[:, slice_indices, :])

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

        if self.config.pad_token_id is None:
            sequence_length = -1
        else:
            if input_ids is not None:
                sequence_length = (torch.ne(input_ids, self.config.pad_token_id).sum(-1) - 1).to(logits.device)
            else:
                sequence_length = -1
                logger.warning_once(
                    f"{self.__class__.__name__} will not detect padding tokens in `inputs_embeds`. Results may be "
                    "unexpected if using padding tokens in conjunction with `inputs_embeds.`"
                )

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

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

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

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

    def get_input_embeddings(self):
        return self.biogpt.embed_tokens

    def set_input_embeddings(self, value):
        self.biogpt.embed_tokens = value


__all__ = [
    "BioGptForCausalLM",
    "BioGptForTokenClassification",
    "BioGptForSequenceClassification",
    "BioGptModel",
    "BioGptPreTrainedModel",
]