modeling_exaone.py

# coding=utf-8
# Copyright 2021 The LG AI Research EXAONE Lab
# Copyright (c) 2018, NVIDIA CORPORATION.  All rights reserved.
#
# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
# and OPT implementations in this library. It has been modified from its
# original forms to accommodate minor architectural differences compared
# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
#
# 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.
""" LG AI Research EXAONE Lab"""
import sys
import os
from typing import List, Optional, Tuple, Union
from packaging import version

import torch
import torch.utils.checkpoint
from torch import nn
from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss
import torch.nn.functional as F

from transformers.activations import ACT2FN
from transformers.cache_utils import Cache, DynamicCache, StaticCache
from transformers.pytorch_utils import ALL_LAYERNORM_LAYERS
from transformers.configuration_utils import PretrainedConfig
from transformers.modeling_attn_mask_utils import AttentionMaskConverter

from transformers.modeling_outputs import (
    BaseModelOutputWithPast,
    BaseModelOutputWithPastAndCrossAttentions,
    CausalLMOutputWithCrossAttentions,
    CausalLMOutputWithPast,
    SequenceClassifierOutputWithPast,
    QuestionAnsweringModelOutput,
)
from transformers.modeling_utils import PreTrainedModel
from transformers.utils import (
    add_code_sample_docstrings,
    add_start_docstrings,
    add_start_docstrings_to_model_forward,
    is_flash_attn_2_available,
    logging,
)
from .configuration_exaone import ExaoneConfig
from torch.nn.utils import skip_init
import math
import numpy as np
from typing import List, Optional, Tuple, Union


if is_flash_attn_2_available():
    try:
        import inspect
        from flash_attn.bert_padding import index_first_axis, pad_input, unpad_input
        from flash_attn import flash_attn_func, flash_attn_varlen_func

        _flash_supports_window_size = "window_size" in list(inspect.signature(flash_attn_func).parameters)

        import flash_attn
        if version.parse(flash_attn.__version__) > version.parse('2.4.2'):
            from flash_attn.ops.triton.layer_norm import rms_norm_fn
        else:
            from flash_attn.ops.triton.layernorm import rms_norm_fn
    except:
        pass


logger = logging.get_logger(__name__)

_CHECKPOINT_FOR_DOC = "exaone"
_CONFIG_FOR_DOC = "ExaoneConfig"

EXAONE_PRETRAINED_MODEL_ARCHIVE_LIST = [
    "exaone",
]


@torch.jit.script
def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
    """
    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
    """
    batch, num_key_value_heads, slen, head_dim = hidden_states.shape
    if n_rep == 1:
        return hidden_states
    hidden_states = hidden_states[:, :, None, :, :].expand(batch, num_key_value_heads, n_rep, slen, head_dim)
    return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)


def apply_rotary_pos_emb(q, k, cos, sin, unsqueeze_dim=1):
    """Applies Rotary Position Embedding to the query and key tensors.

    Args:
        q (`torch.Tensor`): The query tensor.
        k (`torch.Tensor`): The key tensor.
        cos (`torch.Tensor`): The cosine part of the rotary embedding.
        sin (`torch.Tensor`): The sine part of the rotary embedding.
        unsqueeze_dim (`int`, *optional*, defaults to 1):
            The 'unsqueeze_dim' argument specifies the dimension along which to unsqueeze cos[position_ids] and
            sin[position_ids] so that they can be properly broadcasted to the dimensions of q and k. For example, note
            that cos[position_ids] and sin[position_ids] have the shape [batch_size, seq_len, head_dim]. Then, if q and
            k have the shape [batch_size, heads, seq_len, head_dim], then setting unsqueeze_dim=1 makes
            cos[position_ids] and sin[position_ids] broadcastable to the shapes of q and k. Similarly, if q and k have
            the shape [batch_size, seq_len, heads, head_dim], then set unsqueeze_dim=2.
    Returns:
        `tuple(torch.Tensor)` comprising of the query and key tensors rotated using the Rotary Position Embedding.
    """
    cos = cos.unsqueeze(unsqueeze_dim)
    sin = sin.unsqueeze(unsqueeze_dim)
    q_embed = (q * cos) + (rotate_half(q) * sin)
    k_embed = (k * cos) + (rotate_half(k) * sin)
    return q_embed, k_embed


def rotate_half(x):
    """ Rotates half the hidden dims of the input. """
    x1 = x[..., : x.shape[-1] // 2]
    x2 = x[..., x.shape[-1] // 2 :]
    return torch.cat((-x2, x1), dim=-1)


# copied from llama
def _prepare_4d_causal_attention_mask_with_cache_position(
    attention_mask: torch.Tensor,
    sequence_length: int,
    target_length: int,
    dtype: torch.dtype,
    device: torch.device,
    min_dtype: float,
    cache_position: torch.Tensor,
    batch_size: int,
):
    """
    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.
        device (`torch.device`):
            The device to plcae the 4D attention mask on.
        min_dtype (`float`):
            The minimum value representable with the dtype `dtype`.
        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:
        causal_mask = torch.full((sequence_length, target_length), fill_value=min_dtype, dtype=dtype, device=device)
        if sequence_length != 1:
            causal_mask = torch.triu(causal_mask, diagonal=1)
        causal_mask *= torch.arange(target_length, device=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, :]
            padding_mask = padding_mask == 0
            causal_mask[:, :, :, :mask_length] = causal_mask[:, :, :, :mask_length].masked_fill(
                padding_mask, min_dtype
            )

    return causal_mask


class ExaoneRMSNorm(torch.nn.Module):
    def __init__(self, hidden_size, eps=1e-6):
        super().__init__()
        self.eps = eps
        self.weight = torch.nn.Parameter(torch.ones(hidden_size))

    def forward(self, hidden_states):
        input_dtype = hidden_states.dtype
        hidden_states = hidden_states.to(torch.float32)
        variance = hidden_states.pow(2).mean(-1, keepdim=True)
        hidden_states = hidden_states * torch.rsqrt(variance + self.eps)
        return self.weight * hidden_states.to(input_dtype)


class ExaoneTritonRMSNorm(torch.nn.Module):
    def __init__(
        self,
        hidden_size: int = 0,
        eps: float = 1e-5,
    ):
        super().__init__()
        self.eps = eps
        self.drop = None
        self.weight = torch.nn.Parameter(torch.empty(hidden_size))
        self.register_parameter("bias", None)
        self.reset_parameters()

    def reset_parameters(self):
        torch.nn.init.ones_(self.weight)

    def forward(self, x, residual=None, prenorm=False, residual_in_fp32=False):
        return rms_norm_fn(
            x,
            self.weight,
            self.bias,
            residual=residual,
            eps=self.eps,
            dropout_p=self.drop.p if self.drop is not None and self.training else 0.0,
            prenorm=prenorm,
            residual_in_fp32=residual_in_fp32,
        )


ALL_LAYERNORM_LAYERS.append(ExaoneRMSNorm)
ALL_LAYERNORM_LAYERS.append(ExaoneTritonRMSNorm)


class ExaoneRotaryEmbedding(nn.Module):
    """
    Common description for the functions named `_compute_XXX_rope_parameters()`
        - Copied from `transformers.modeling_rope_utils` in v4.43, with some modifications.

        Computes the inverse frequencies with linear scaling. 
        The EXAONE model supports 'default', 'linear', 'dynamic', and 'yarn'.
        
        Args:
            config (:obj:`~transformers.PretrainedConfig`):
                The model configuration.
            device (:obj:`torch.device`):
                The device to use for initialization of the inverse frequencies.
            seq_len (:obj:`int`, `optional`):
                The current sequence length. Unused for this type of RoPE.
        Returns:
            Tuple of (:obj:`torch.Tensor`, :obj:`float`), containing the inverse frequencies for the RoPE embeddings and the
            post-processing scaling factor applied to the computed cos/sin (unused in some types of RoPE).
    """

    def _compute_default_rope_parameters(
        self,
        config: Optional[PretrainedConfig],
        device: Optional["torch.device"] = None,
        seq_len: Optional[int] = None,
    ) -> Tuple["torch.Tensor", float]:
        base = config.rope_theta
        partial_rotary_factor = config.partial_rotary_factor if hasattr(config, "partial_rotary_factor") else 1.0
        dim = int((config.hidden_size // config.num_attention_heads) * partial_rotary_factor)

        attention_factor = 1.0  # Unused in this type of RoPE

        inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2, dtype=torch.int64).float().to(device) / dim))
        return inv_freq, attention_factor

    def _compute_linear_scaling_rope_parameters(
        self,
        config: Optional[PretrainedConfig],
        device: Optional["torch.device"] = None,
        seq_len: Optional[int] = None,
    ) -> Tuple["torch.Tensor", float]:
        factor = config.rope_scaling["factor"]
        if factor < 1.0:
            logger.warning_once(f"`rope_scaling`'s factor field must be a float >= 1, got {factor}")

        inv_freq, attention_factor = self._compute_default_rope_parameters(config, device, seq_len)
        inv_freq /= factor
        return inv_freq, attention_factor

    def _compute_dynamic_ntk_parameters(
        self,
        config: Optional[PretrainedConfig],
        device: Optional["torch.device"] = None,
        seq_len: Optional[int] = None,
    ) -> Tuple["torch.Tensor", float]:
        base = config.rope_theta
        partial_rotary_factor = config.partial_rotary_factor if hasattr(config, "partial_rotary_factor") else 1.0
        dim = int((config.hidden_size // config.num_attention_heads) * partial_rotary_factor)
        max_position_embeddings = config.max_position_embeddings
        factor = config.rope_scaling["factor"]
        if factor < 1.0:
            logger.warning_once(f"`rope_scaling`'s factor field must be a float >= 1, got {factor}")

        attention_factor = 1.0  # Unused in this type of RoPE
        seq_len = seq_len if seq_len is not None else max_position_embeddings

        base = base * ((factor * seq_len / max_position_embeddings) - (factor - 1)) ** (dim / (dim - 2))
        inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2, dtype=torch.int64).float().to(device) / dim))
        return inv_freq, attention_factor
    
    def _compute_yarn_parameters(
        self,
        config: PretrainedConfig, 
        device: "torch.device", 
        seq_len: Optional[int] = None, 
    ) -> Tuple["torch.Tensor", float]:
        base = config.rope_theta
        partial_rotary_factor = config.partial_rotary_factor if hasattr(config, "partial_rotary_factor") else 1.0
        dim = int((config.hidden_size // config.num_attention_heads) * partial_rotary_factor)
        max_position_embeddings = config.max_position_embeddings
        factor = config.rope_scaling["factor"]
        if factor < 1.0:
            logger.warning_once(f"`rope_scaling`'s factor field must be a float >= 1, got {factor}")

        # Sets the attention factor as suggested in the paper
        attention_factor = config.rope_scaling.get("attention_factor")
        if attention_factor is None:
            attention_factor = 0.1 * math.log(factor) + 1.0
        if attention_factor < 0:
            logger.warning_once(
                f"`rope_scaling`'s attention_factor field must be a float greater than 0, got {attention_factor}"
            )

        # Optional config options
        # beta_fast/beta_slow: as suggested in the paper, default to 32/1 (correspondingly)
        beta_fast = config.rope_scaling.get("beta_fast") or 32
        beta_slow = config.rope_scaling.get("beta_slow") or 1
        if not isinstance(beta_fast, float):
            logger.warning_once(f"`rope_scaling`'s beta_fast field must be a float, got {beta_fast}")
        if not isinstance(beta_slow, float):
            logger.warning_once(f"`rope_scaling`'s beta_slow field must be a float, got {beta_fast}")
        if beta_fast < beta_slow:
            logger.warning_once(
                f"`rope_scaling`'s beta_fast field must be greater than beta_slow, got beta_fast={beta_fast} "
                f"(defaults to 32 if None) and beta_slow={beta_slow} (defaults to 1 if None)"
            )

        # Compute the inverse frequencies
        def find_correction_dim(num_rotations, dim, base, max_position_embeddings):
            """Inverse dimension formula to find the dimension based on the number of rotations"""
            return (dim * math.log(max_position_embeddings / (num_rotations * 2 * math.pi))) / (2 * math.log(base))

        def find_correction_range(low_rot, high_rot, dim, base, max_position_embeddings):
            """Find dimension range bounds based on rotations"""
            low = math.floor(find_correction_dim(low_rot, dim, base, max_position_embeddings))
            high = math.ceil(find_correction_dim(high_rot, dim, base, max_position_embeddings))
            return max(low, 0), min(high, dim - 1)

        def linear_ramp_mask(min, max, dim):
            if min == max:
                max += 0.001  # Prevent singularity

            linear_func = (torch.arange(dim, dtype=torch.float32) - min) / (max - min)
            ramp_func = torch.clamp(linear_func, 0, 1)
            return ramp_func

        pos_freqs = base ** (torch.arange(0, dim, 2).float().to(device) / dim)
        inv_freq_extrapolation = 1.0 / pos_freqs
        inv_freq_interpolation = 1.0 / (factor * pos_freqs)

        low, high = find_correction_range(beta_fast, beta_slow, dim, base, max_position_embeddings)

        # Get n-dimensional rotational scaling corrected for extrapolation
        inv_freq_mask = 1 - linear_ramp_mask(low, high, dim // 2).float().to(device)
        inv_freq = inv_freq_interpolation * (1 - inv_freq_mask) + inv_freq_extrapolation * inv_freq_mask

        return inv_freq, attention_factor

    def __init__(self, config: ExaoneConfig, device=None):
        ROPE_INIT_FUNCTIONS = {
            "default": self._compute_default_rope_parameters,
            "linear": self._compute_linear_scaling_rope_parameters,
            "dynamic": self._compute_dynamic_ntk_parameters,
            "yarn": self._compute_yarn_parameters,
        }

        super().__init__()
        if config.rope_scaling is not None:
            self.rope_type = config.rope_scaling.get("rope_type", config.rope_scaling.get("type"))
        else:
            self.rope_type = "default"
        self.max_seq_len = config.max_position_embeddings
        self.original_max_seq_len = config.max_position_embeddings

        self.config = config
        if self.rope_type not in ROPE_INIT_FUNCTIONS:
            raise KeyError(f"The EXAONE model does not support RoPE type: {self.rope_type}")
        self.rope_init_fn = ROPE_INIT_FUNCTIONS[self.rope_type]

        inv_freq, self.attention_scaling = self.rope_init_fn(self.config, device)
        self.register_buffer("inv_freq", inv_freq, persistent=False)
        self.original_inv_freq = self.inv_freq

    def _update_freq(self, position_ids, device):
        """
        dynamic RoPE layers should recompute `inv_freq` in the following situations:
        1 - growing beyond the cached sequence length (allow scaling)
        2 - the current sequence length is in the original scale (avoid losing precision with small sequences)
        """
        seq_len = torch.max(position_ids) + 1
        if seq_len > self.max_seq_len:  # expand to seq_len
            inv_freq, self.attention_scaling = self.rope_init_fn(self.config, device, seq_len=seq_len)
            self.register_buffer("inv_freq", inv_freq, persistent=False)
            self.max_seq_len = seq_len
        
        if seq_len < self.original_max_seq_len and self.max_seq_len > self.original_max_seq_len:    # reset to original
            self.register_buffer("inv_freq", self.original_inv_freq, persistent=False)
            self.max_seq_len = self.original_max_seq_len

    @torch.no_grad()
    def forward(self, x, position_ids):
        if "dynamic" in self.rope_type:
            self._update_freq(position_ids, device=x.device)

        inv_freq_expanded = self.inv_freq[None, :, None].float().expand(position_ids.shape[0], -1, 1)
        position_ids_expanded = position_ids[:, None, :].float()
        
        device_type = x.device.type
        device_type = device_type if isinstance(device_type, str) and device_type != "mps" else "cpu"
        with torch.autocast(device_type=device_type, enabled=False):
            freqs = (inv_freq_expanded @ position_ids_expanded).transpose(1, 2)
            emb = torch.cat((freqs, freqs), dim=-1)
            cos, sin = emb.cos(), emb.sin()
        
        cos, sin = cos * self.attention_scaling, sin * self.attention_scaling
        return cos.to(x.dtype), sin.to(x.dtype)


class ExaoneSelfAttention(nn.Module):
    def __init__(self, config: ExaoneConfig, layer_idx: Optional[int] = None):
        super().__init__()
        self.config = config
        self.layer_idx = layer_idx
        self.embed_dim = config.hidden_size
        self.num_heads = config.num_attention_heads
        self.head_dim = self.embed_dim // self.num_heads
        self.num_key_value_heads = config.num_key_value_heads
        self.num_key_value_groups = self.num_heads // self.num_key_value_heads
        self.attention_dropout_rate = config.attention_dropout

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

        self.rotary = ExaoneRotaryEmbedding(config)
        
        self.k_proj = nn.Linear(self.embed_dim, self.num_key_value_heads * self.head_dim, bias=False)
        self.v_proj = nn.Linear(self.embed_dim, self.num_key_value_heads * self.head_dim, bias=False)
        self.q_proj = nn.Linear(self.embed_dim, self.num_heads * self.head_dim, bias=False)
        self.out_proj = nn.Linear(self.embed_dim, self.embed_dim, bias=False)

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_value: Optional[Cache] = None,
        output_attentions: Optional[bool] = False,
        use_cache: Optional[bool] = False,
        cache_position: Optional[torch.LongTensor] = None,
        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
        **kwargs,
    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:

        bsz, q_len, _ = hidden_states.size()
        query_states = self.q_proj(hidden_states)
        key_states = self.k_proj(hidden_states)
        value_states = self.v_proj(hidden_states)

        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)

        if position_embeddings is None:
            cos, sin = self.rotary(value_states, position_ids=position_ids)
        else:
            cos, sin = position_embeddings
        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)

        if past_key_value is not None:
            # sin and cos are specific to RoPE models; cache_position needed for the static cache
            cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)

        key_states = repeat_kv(key_states, self.num_key_value_groups)
        value_states = repeat_kv(value_states, self.num_key_value_groups)

        attn_weights = torch.matmul(query_states, key_states.transpose(2, 3)) / math.sqrt(self.head_dim)

        if attention_mask is not None:
            causal_mask = attention_mask[:, :, :, :key_states.shape[-2]]
            attn_weights = attn_weights + causal_mask

        attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype)
        attn_weights = nn.functional.dropout(attn_weights, p=self.attention_dropout_rate, training=self.training)
        attn_output = torch.matmul(attn_weights, value_states)

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

        attn_output = attn_output.transpose(1, 2).contiguous()
        attn_output = attn_output.reshape(bsz, q_len, self.embed_dim).contiguous()

        attn_output = self.out_proj(attn_output)

        if not output_attentions:
            attn_weights = None

        return attn_output, attn_weights, past_key_value


class ExaoneFlashAttention(ExaoneSelfAttention):
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)

    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()

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_value: Optional[Cache] = None,
        output_attentions: Optional[bool] = False,
        use_cache: Optional[bool] = False,
        cache_position: Optional[torch.LongTensor] = None,
        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
        **kwargs,
    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
        if isinstance(past_key_value, StaticCache):
            raise ValueError(
                "`static` cache implementation is not compatible with `attn_implementation==flash_attention_2` "
                "make sure to use `sdpa` in the mean time, and open an issue at https://github.com/huggingface/transformers"
            )

        output_attentions = False

        bsz, q_len, h_size = hidden_states.size()

        query_states = self.q_proj(hidden_states)
        key_states = self.k_proj(hidden_states)
        value_states = self.v_proj(hidden_states)

        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)

        if position_embeddings is None:
            cos, sin = self.rotary(value_states, position_ids=position_ids)
        else:
            cos, sin = position_embeddings
        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)

        if past_key_value is not None:
            # sin and cos are specific to RoPE models; cache_position needed for the static cache
            cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
            # Only update cache as shape of [bsz, n_head, q_len, head_dim]
            # TODO: need to be fixed when transformers' KV cache layout is changed
            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)

        query_states = query_states.transpose(1, 2)
        key_states = key_states.transpose(1, 2)
        value_states = value_states.transpose(1, 2)

        # In PEFT, usually we cast the layer norms in float32 for training stability reasons
        # therefore the input hidden states gets silently casted in float32. Hence, we need
        # cast them back in the correct dtype just to be sure everything works as expected.
        input_dtype = query_states.dtype
        if input_dtype == torch.float32:
            if torch.is_autocast_enabled():
                target_dtype = torch.get_autocast_gpu_dtype()
            # Handle the case where the model is quantized
            elif hasattr(self.config, "_pre_quantization_dtype"):
                target_dtype = self.config._pre_quantization_dtype
            else:
                target_dtype = self.q_proj.weight.dtype

            logger.warning_once(
                f"The input hidden states seems to be silently casted in float32, this might be related to"
                f" the fact you have upcasted embedding or layer norm layers in float32. We will cast back the input in"
                f" {target_dtype}."
            )

            query_states = query_states.to(target_dtype)
            key_states = key_states.to(target_dtype)
            value_states = value_states.to(target_dtype)

        dropout_rate = self.attention_dropout_rate if self.training else 0.0

        attn_output = self._flash_attention_forward(
            query_states, key_states, value_states, attention_mask, q_len, dropout=dropout_rate, is_causal=True
        )

        attn_output = attn_output.reshape(bsz, q_len, self.embed_dim).contiguous()
        attn_output = self.out_proj(attn_output)

        if not output_attentions:
            attn_weights = None

        return attn_output, attn_weights, past_key_value
    
    @staticmethod
    def _flash_attention_forward(
        query_states: torch.Tensor,
        key_states: torch.Tensor,
        value_states: torch.Tensor,
        attention_mask: torch.Tensor,
        query_length: int,
        is_causal: bool,
        dropout: float = 0.0,
        softmax_scale: Optional[float] = None,
        sliding_window: Optional[int] = None,
        use_top_left_mask: bool = False,
        softcap: Optional[float] = None,
        deterministic: bool = os.environ.get("FLASH_ATTENTION_DETERMINISTIC", "0") == "1",
    ):
        """
        Calls the forward method of Flash Attention - if the input hidden states contain at least one padding token
        first unpad the input, then computes the attention scores and pad the final attention scores.

        Args:
            query_states (`torch.Tensor`):
                Input query states to be passed to Flash Attention API
            key_states (`torch.Tensor`):
                Input key states to be passed to Flash Attention API
            value_states (`torch.Tensor`):
                Input value states to be passed to Flash Attention API
            attention_mask (`torch.Tensor`):
                The padding mask - corresponds to a tensor of size `(batch_size, seq_len)` where 0 stands for the
                position of padding tokens and 1 for the position of non-padding tokens.
            dropout (`float`):
                Attention dropout
            softmax_scale (`float`, *optional*):
                The scaling of QK^T before applying softmax. Default to 1 / sqrt(head_dim)
            use_top_left_mask (`bool`, defaults to `False`):
                flash_attn<2.1 generates top-left aligned causal mask, while what is needed here is bottom-right alignement, that was made default for flash_attn>=2.1. This attribute is used to handle this difference.
            softcap (`float`, *optional*):
                Softcap for the attention logits, used e.g. in gemma2.
            deterministic (`bool`, *optional*):
                Determines if the deterministic option introduced in flash_attn>=2.4.1 is enabled.
        """
        if not use_top_left_mask:
            causal = is_causal
        else:
            # TODO: Remove the `query_length != 1` check once Flash Attention for RoCm is bumped to 2.1. For details, please see the comment in transformers.models.llama.modeling_llama.LlamaFlashAttention2.__init__.
            causal = is_causal and query_length != 1

        # Assuming 4D tensors, key_states.shape[1] is the key/value sequence length (source length).
        use_sliding_windows = (
            _flash_supports_window_size and sliding_window is not None and key_states.shape[1] > sliding_window
        )
        flash_kwargs = {"window_size": (sliding_window, sliding_window)} if use_sliding_windows else {}

        if softcap is not None:
            flash_kwargs["softcap"] = softcap

        # Contains at least one padding token in the sequence
        if attention_mask is not None:
            batch_size = query_states.shape[0]
            query_states, key_states, value_states, indices_q, cu_seq_lens, max_seq_lens = ExaoneFlashAttention._upad_input(
                query_states, key_states, value_states, attention_mask, query_length
            )
            cu_seqlens_q, cu_seqlens_k = cu_seq_lens
            max_seqlen_in_batch_q, max_seqlen_in_batch_k = max_seq_lens

            attn_output_unpad = flash_attn_varlen_func(
                query_states,
                key_states,
                value_states,
                cu_seqlens_q=cu_seqlens_q,
                cu_seqlens_k=cu_seqlens_k,
                max_seqlen_q=max_seqlen_in_batch_q,
                max_seqlen_k=max_seqlen_in_batch_k,
                dropout_p=dropout,
                softmax_scale=softmax_scale,
                causal=causal,
                **flash_kwargs,
            )
            attn_output = pad_input(attn_output_unpad, indices_q, batch_size, query_length)
        else:
            attn_output = flash_attn_func(
                query_states, key_states, value_states, dropout, softmax_scale=softmax_scale, causal=causal, **flash_kwargs
            )

        return attn_output
    
    @staticmethod
    def _upad_input(
        query_layer: torch.Tensor,
        key_layer: torch.Tensor,
        value_layer: torch.Tensor,
        attention_mask: torch.Tensor,
        query_length: int,
    ):
        """
        Unpads query, key, and values tensors, using a single dimension for all tokens even though they belong to different batches.

        This function is used instead of `flash_attn.bert_padding.unpad_input` in order to avoid the recomputation of the same intermediary
        tensors for query, key, value tensors.

        Arguments:
            query_layer (`torch.Tensor`):
                Query state with padding. Shape: (batch_size, query_length, num_heads, head_dim).
            key_layer (`torch.Tensor`):
                Key state with padding. Shape: (batch_size, kv_seq_len, num_key_value_heads, head_dim).
            value_layer (`torch.Tensor`):
                Value state with padding. Shape: (batch_size, kv_seq_len, num_key_value_heads, head_dim).
            attention_mask (`torch.Tensor`):
                Boolean or int tensor of shape (batch_size, sequence_length), 1 means valid and 0 means not valid.
            query_length (`int`):
                Target length.

        Return:
            query_layer (`torch.Tensor):
                Query state without padding. Shape: (total_target_length, num_heads, head_dim).
            key_layer (`torch.Tensor`):
                Key state with padding. Shape: (total_source_length, num_key_value_heads, head_dim).
            value_layer (`torch.Tensor`):
                Value state with padding. Shape: (total_source_length, num_key_value_heads, head_dim).
            indices_q (`torch.Tensor`):
                The indices of non-masked tokens from the flattened input target sequence.
            (cu_seqlens_q, cu_seqlens_k) (`Tuple[int]`):
                The cumulative sequence lengths for the target (query) and source (key, value), used to index into ragged (unpadded) tensors. `cu_seqlens` shape is (batch_size + 1,).
            (max_seqlen_in_batch_q, max_seqlen_in_batch_k) (`Tuple[int]`):
                Maximum sequence length in batch (`max_seqlen_in_batch_q` for the target sequence i.e. query, `max_seqlen_in_batch_k` for the source sequence i.e. key/value).
        """
        indices_k, cu_seqlens_k, max_seqlen_in_batch_k = ExaoneFlashAttention._get_unpad_data(attention_mask)
        batch_size, kv_seq_len, num_key_value_heads, head_dim = key_layer.shape

        key_layer = index_first_axis(key_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k)
        value_layer = index_first_axis(
            value_layer.reshape(batch_size * kv_seq_len, num_key_value_heads, head_dim), indices_k
        )
        if query_length == kv_seq_len:
            query_layer = index_first_axis(query_layer.reshape(batch_size * kv_seq_len, -1, head_dim), indices_k)
            cu_seqlens_q = cu_seqlens_k
            max_seqlen_in_batch_q = max_seqlen_in_batch_k
            indices_q = indices_k
        elif query_length == 1:
            max_seqlen_in_batch_q = 1
            cu_seqlens_q = torch.arange(
                batch_size + 1, dtype=torch.int32, device=query_layer.device
            )  # There is a memcpy here, that is very bad.
            indices_q = cu_seqlens_q[:-1]
            query_layer = query_layer.squeeze(1)
        else:
            # The -q_len: slice assumes left padding.
            attention_mask = attention_mask[:, -query_length:]
            query_layer, indices_q, cu_seqlens_q, max_seqlen_in_batch_q = unpad_input(query_layer, attention_mask)

        return (
            query_layer,
            key_layer,
            value_layer,
            indices_q,
            (cu_seqlens_q, cu_seqlens_k),
            (max_seqlen_in_batch_q, max_seqlen_in_batch_k),
        )
    
    @staticmethod
    def _get_unpad_data(attention_mask: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor, int]:
        """
        Retrieves indexing data required to repad unpadded (ragged) tensors.

        Arguments:
            attention_mask (`torch.Tensor`):
                Boolean or int tensor of shape (batch_size, sequence_length), 1 means valid and 0 means not valid.

        Return:
            indices (`torch.Tensor):
                The indices of non-masked tokens from the flattened input sequence.
            cu_seqlens (`torch.Tensor`):
                The cumulative sequence lengths, used to index into ragged (unpadded) tensors. `cu_seqlens` shape is (batch_size + 1,).
            max_seqlen_in_batch (`int`):
                Maximum sequence length in batch.
        """
        seqlens_in_batch = attention_mask.sum(dim=-1, dtype=torch.int32)
        indices = torch.nonzero(attention_mask.flatten(), as_tuple=False).flatten()
        max_seqlen_in_batch = seqlens_in_batch.max().item()
        cu_seqlens = F.pad(torch.cumsum(seqlens_in_batch, dim=0, dtype=torch.int32), (1, 0))
        return (
            indices,
            cu_seqlens,
            max_seqlen_in_batch,
        )


class ExaoneSdpaAttention(ExaoneSelfAttention):
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
    
    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_value: Optional[Cache] = None,
        output_attentions: Optional[bool] = False,
        use_cache: Optional[bool] = False,
        cache_position: Optional[torch.LongTensor] = None,
        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
        **kwargs,
    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:

        if output_attentions:
            logger.warning_once(
                "ExaoneModel is using ExaoneSdpaAttention, but `torch.nn.functional.scaled_dot_product_attention` does not support `output_attentions=True`. Falling back to the manual attention implementation, "
                'but specifying the manual implementation will be required from Transformers version v5.0.0 onwards. This warning can be removed using the argument `attn_implementation="eager"` when loading the model.'
            )
            return super().forward(
                hidden_states=hidden_states,
                attention_mask=attention_mask,
                position_ids=position_ids,
                past_key_value=past_key_value,
                output_attentions=output_attentions,
                use_cache=use_cache,
                cache_position=cache_position,
                position_embeddings=position_embeddings,
                **kwargs,
            )

        bsz, q_len, _ = hidden_states.size()

        query_states = self.q_proj(hidden_states)
        key_states = self.k_proj(hidden_states)
        value_states = self.v_proj(hidden_states)

        query_states = query_states.view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
        key_states = key_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)
        value_states = value_states.view(bsz, q_len, self.num_key_value_heads, self.head_dim).transpose(1, 2)

        if position_embeddings is None:
            cos, sin = self.rotary(value_states, position_ids=position_ids)
        else:
            cos, sin = position_embeddings
        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin)

        if past_key_value is not None:
            # sin and cos are specific to RoPE models; cache_position needed for the static cache
            cache_kwargs = {"sin": sin, "cos": cos, "cache_position": cache_position}
            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx, cache_kwargs)

        key_states = repeat_kv(key_states, self.num_key_value_groups)
        value_states = repeat_kv(value_states, self.num_key_value_groups)

        causal_mask = attention_mask
        if attention_mask is not None:
            causal_mask = causal_mask[:, :, :, :key_states.shape[-2]]        

        # SDPA with memory-efficient backend is currently (torch==2.1.2) bugged with non-contiguous inputs with custom attn_mask,
        # Reference: https://github.com/pytorch/pytorch/issues/112577.
        if query_states.device.type == "cuda" and causal_mask is not None:
            query_states = query_states.contiguous()
            key_states = key_states.contiguous()
            value_states = value_states.contiguous()

        # We dispatch to SDPA's Flash Attention or Efficient kernels via this `is_causal` if statement instead of an inline conditional assignment
        # in SDPA to support both torch.compile's dynamic shapes and full graph options. An inline conditional prevents dynamic shapes from compiling.
        is_causal = True if causal_mask is None and q_len > 1 else False

        attn_output = torch.nn.functional.scaled_dot_product_attention(
            query_states,
            key_states,
            value_states,
            attn_mask=causal_mask,
            dropout_p=self.attention_dropout_rate if self.training else 0.0,
            is_causal=is_causal,
        )

        attn_output = attn_output.transpose(1, 2).contiguous()
        attn_output = attn_output.reshape(bsz, q_len, self.embed_dim).contiguous()

        attn_output = self.out_proj(attn_output)

        return attn_output, None, past_key_value


class ExaoneAttention(nn.Module):
    def __init__(self, config, layer_id=0):
        super().__init__()
        self.layer_id = layer_id
        if 'flash' in config._attn_implementation:
            self.attention = ExaoneFlashAttention(config, self.layer_id)
        elif 'sdpa' in config._attn_implementation:
            self.attention = ExaoneSdpaAttention(config, self.layer_id)
        else:
            self.attention = ExaoneSelfAttention(config, self.layer_id)

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_value: Optional[Cache] = None,
        output_attentions: Optional[bool] = False,
        use_cache: Optional[bool] = False,
        cache_position: Optional[torch.LongTensor] = None,
        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
        **kwargs,
    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:

        return self.attention(
            hidden_states=hidden_states,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_value=past_key_value,
            output_attentions=output_attentions,
            use_cache=use_cache,
            cache_position=cache_position,
            position_embeddings=position_embeddings,
            **kwargs,
        )


class ExaoneGatedMLP(nn.Module):
    def __init__(self, intermediate_size, config):
        super().__init__()
        self.config = config
        embed_dim = config.hidden_size
        self.c_fc_0 = nn.Linear(embed_dim, intermediate_size, bias=False)
        self.c_fc_1 = nn.Linear(embed_dim, intermediate_size, bias=False)
        self.c_proj = nn.Linear(intermediate_size, embed_dim, bias=False)
        self.act = ACT2FN[config.activation_function]

    def forward(self, hidden_states):
        output_proj = self.c_proj(self.act(self.c_fc_0(hidden_states)) * self.c_fc_1(hidden_states))
        return output_proj


class ExaoneBlock(nn.Module):
    def __init__(self, config, layer_id):
        super().__init__()
        self.config = config
        hidden_size = config.hidden_size
        inner_dim = config.intermediate_size if config.intermediate_size is not None else 4 * hidden_size
        self.ln_1 = ExaoneRMSNorm(hidden_size = hidden_size, eps=config.layer_norm_epsilon)
        self.attn = ExaoneAttention(config, layer_id)
        self.ln_2 = ExaoneRMSNorm(hidden_size = hidden_size, eps=config.layer_norm_epsilon)
        self.mlp = ExaoneGatedMLP(inner_dim, config)

    def forward(
        self,
        hidden_states: torch.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_value: Optional[Cache] = None,
        output_attentions: Optional[bool] = False,
        use_cache: Optional[bool] = False,
        cache_position: Optional[torch.LongTensor] = None,
        position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,
        **kwargs,
    ) -> Tuple[torch.FloatTensor, Optional[Tuple[torch.FloatTensor, torch.FloatTensor]]]:

        residual = hidden_states
        hidden_states = self.ln_1(hidden_states)

        hidden_states, self_attn_weights, present_key_value = self.attn(
            hidden_states=hidden_states,
            attention_mask=attention_mask,
            position_ids=position_ids,
            past_key_value=past_key_value,
            output_attentions=output_attentions,
            use_cache=use_cache,
            cache_position=cache_position,
            position_embeddings=position_embeddings,
            **kwargs,
        )
        # residual connection
        hidden_states = residual + hidden_states

        residual = hidden_states
        hidden_states = self.ln_2(hidden_states)
        hidden_states = self.mlp(hidden_states)

        hidden_states = residual + hidden_states

        outputs = (hidden_states,)

        if output_attentions:
            outputs += (self_attn_weights,)

        if use_cache:
            outputs += (present_key_value,)

        return outputs


class ExaonePreTrainedModel(PreTrainedModel):
    """
    An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained
    models.
    """

    config_class = ExaoneConfig
    base_model_prefix = "transformer"
    supports_gradient_checkpointing = True
    _no_split_modules = ["ExaoneBlock"]
    _skip_keys_device_placement = "past_key_values"
    _supports_flash_attn_2 = True
    _supports_sdpa = True
    _supports_cache_class = True

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

    def _init_weights(self, module):
        """Initialize the weights."""
        if isinstance(module, (nn.Linear,)):
            # Slightly different from the TF version which uses truncated_normal for initialization
            # cf https://github.com/pytorch/pytorch/pull/5617
            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
            if module.bias is not None:
                module.bias.data.zero_()
        elif isinstance(module, nn.Embedding):
            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
            if module.padding_idx is not None:
                module.weight.data[module.padding_idx].zero_()
        elif isinstance(module, ExaoneRMSNorm):
            module.weight.data.fill_(1.0)


EXAONE_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 (:class:`~transformers.ExaoneConfig`): 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 :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights.
"""

EXAONE_INPUTS_DOCSTRING = r"""
    Args:
        input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, input_ids_length)`):
            :obj:`input_ids_length` = ``sequence_length`` if :obj:`past_key_values` is ``None`` else
            ``past_key_values.get_seq_length()`` (``sequence_length`` of input past key value states). Indices of input
            sequence tokens in the vocabulary.

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

            `What are input IDs? <../glossary.html#input-ids>`__
        attention_mask (:obj:`torch.FloatTensor` of shape :obj:`(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.html#attention-mask>`__
        position_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
            Indices of positions of each input sequence tokens in the position embeddings. Selected in the range ``[0,
            config.max_position_embeddings - 1]``.

            `What are position IDs? <../glossary.html#position-ids>`_
        past_key_values (:obj:`Cache`, `optional`):
            Contains precomputed hidden-states (key and values in the attention blocks) as computed by the model (see
            :obj:`past_key_values` output below). Can be used to speed up sequential decoding. This typically consists 
            in the `past_key_values` returned by the model at a previous stage of decoding, when `use_cache=True` or 
            `config.use_cache=True`.
        inputs_embeds (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`, `optional`):
            Optionally, instead of passing :obj:`input_ids` you can choose to directly pass an embedded representation.
            This is useful if you want more control over how to convert :obj:`input_ids` indices into associated
            vectors than the model's internal embedding lookup matrix.

            If :obj:`past_key_values` is used, optionally only the last :obj:`inputs_embeds` have to be input (see
            :obj:`past_key_values`).
        use_cache (:obj:`bool`, `optional`):
            If set to :obj:`True`, :obj:`past_key_values` key value states are returned and can be used to speed up
            decoding (see :obj:`past_key_values`).
        output_attentions (:obj:`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 (:obj:`bool`, `optional`):
            Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors for
            more detail.
        return_dict (:obj:`bool`, `optional`):
            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple.
        cache_position (:obj:`torch.LongTensor` of shape :obj:`(sequence_length)`, `optional`):
            Indices depicting the position of the input sequence tokens in the sequence. Contrarily to `position_ids`,
            this tensor is not affected by padding. It is used to update the cache in the correct position and to infer
            the complete sequence length.
"""


@add_start_docstrings(
    "The bare EXAONE Model transformer outputting raw hidden-states without any specific head on top.",
    EXAONE_START_DOCSTRING,
)
class ExaoneModel(ExaonePreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.config = config
        self.embed_dim = config.hidden_size
        self.wte = nn.Embedding(config.vocab_size, self.embed_dim, self.config.pad_token_id)
        self.drop = nn.Dropout(float(config.embed_dropout))
        self.h = nn.ModuleList([ExaoneBlock(config, layer_id=i) for i in range(config.num_layers)])
        self.ln_f = ExaoneRMSNorm(hidden_size=self.embed_dim, eps=config.layer_norm_epsilon)
        self.rotary = ExaoneRotaryEmbedding(config)
        self.gradient_checkpointing = False
        # Initialize weights and apply final processing
        self.post_init()

    def get_input_embeddings(self):
        return self.wte

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

    @add_start_docstrings_to_model_forward(EXAONE_INPUTS_DOCSTRING)
    @add_code_sample_docstrings(
        checkpoint=_CHECKPOINT_FOR_DOC,
        output_type=BaseModelOutputWithPastAndCrossAttentions,
        config_class=_CONFIG_FOR_DOC,
    )
    def forward(
        self,
        input_ids: Optional[torch.Tensor] = None,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.Tensor] = None,
        past_key_values: Optional[Cache] = None,
        inputs_embeds: Optional[torch.Tensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        cache_position: Optional[torch.LongTensor] = None,
    ) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPast]:
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        use_cache = use_cache if use_cache is not None else self.config.use_cache
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        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

        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:
            batch_size, seq_length = input_ids.shape[:2]
        elif inputs_embeds is not None:
            batch_size, seq_length = inputs_embeds.shape[:2]
        else:
            raise ValueError("You have to specify either input_ids or inputs_embeds")

        return_legacy_cache = False
        if (
            use_cache and not isinstance(past_key_values, Cache) and not self.training
        ):  # kept for BC (non `Cache` `past_key_values` inputs)
            return_legacy_cache = True
            past_key_values = DynamicCache.from_legacy_cache(past_key_values)
            logger.warning_once(
                "We detected that you are passing `past_key_values` as a tuple and this is deprecated and will be removed in v4.43. "
                "Please use an appropriate `Cache` class (https://huggingface.co/docs/transformers/v4.41.3/en/internal/generation_utils#transformers.Cache)"
            )

        if inputs_embeds is None:
            inputs_embeds = self.wte(input_ids)

        if cache_position is None:
            past_seen_tokens = past_key_values.get_seq_length() if past_key_values is not None else 0
            cache_position = torch.arange(
                past_seen_tokens, past_seen_tokens + inputs_embeds.shape[1], device=inputs_embeds.device
            )
        if position_ids is None:
            position_ids = cache_position.unsqueeze(0)

        causal_mask = self._update_causal_mask(
            attention_mask, inputs_embeds, cache_position, past_key_values, output_attentions
        )

        hidden_states = inputs_embeds
        hidden_states = self.drop(hidden_states)

        position_embeddings = self.rotary(hidden_states, position_ids)

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

        for block in self.h:
            if output_hidden_states:
                all_hidden_states = all_hidden_states + (hidden_states,)

            if self.gradient_checkpointing and self.training:
                outputs = self._gradient_checkpointing_func(
                    block.__call__,
                    hidden_states,
                    causal_mask,
                    position_ids,
                    past_key_values,
                    output_attentions,
                    use_cache,
                    cache_position,
                    position_embeddings,
                )
            else:
                outputs = block(
                    hidden_states,
                    attention_mask=causal_mask,
                    position_ids=position_ids,
                    past_key_value=past_key_values,
                    output_attentions=output_attentions,
                    use_cache=use_cache,
                    cache_position=cache_position,
                    position_embeddings=position_embeddings,
                )

            hidden_states = outputs[0]
            if use_cache:
                next_decoder_cache = outputs[2 if output_attentions else 1]

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

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

        next_cache = None
        if use_cache:
            next_cache = next_decoder_cache.to_legacy_cache() if return_legacy_cache else next_decoder_cache
        if not return_dict:
            return tuple(v for v in [hidden_states, next_cache, all_hidden_states, all_self_attns] if v is not None)

        return BaseModelOutputWithPast(
            last_hidden_state=hidden_states,
            past_key_values=next_cache,
            hidden_states=all_hidden_states,
            attentions=all_self_attns,
        )
    
    # copied from llama
    def _update_causal_mask(
        self,
        attention_mask: torch.Tensor,
        input_tensor: torch.Tensor,
        cache_position: torch.Tensor,
        past_key_values: Cache,
        output_attentions: bool,
    ):
        # TODO: As of torch==2.2.0, the `attention_mask` passed to the model in `generate` is 2D and of dynamic length even when the static
        # KV cache is used. This is an issue for torch.compile which then recaptures cudagraphs at each decode steps due to the dynamic shapes.
        # (`recording cudagraph tree for symint key 13`, etc.), which is VERY slow. A workaround is `@torch.compiler.disable`, but this prevents using
        # `fullgraph=True`. See more context in https://github.com/huggingface/transformers/pull/29114

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

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

        # 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_static_cache and not output_attentions:
            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, device = input_tensor.dtype, input_tensor.device
        min_dtype = torch.finfo(dtype).min
        sequence_length = input_tensor.shape[1]
        if using_static_cache:
            target_length = past_key_values.get_max_length()
        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 = _prepare_4d_causal_attention_mask_with_cache_position(
            attention_mask,
            sequence_length=sequence_length,
            target_length=target_length,
            dtype=dtype,
            device=device,
            min_dtype=min_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 == "cuda"
            and not output_attentions
        ):
            # Attend to all tokens in fully masked rows in the causal_mask, for example the relevant first rows when
            # using left padding. This is required by F.scaled_dot_product_attention memory-efficient attention path.
            # Details: https://github.com/pytorch/pytorch/issues/110213
            causal_mask = AttentionMaskConverter._unmask_unattended(causal_mask, min_dtype)

        return causal_mask


@add_start_docstrings(
    """
    The EXAONE Model transformer with a language modeling head on top (linear layer with weights tied to the input
    embeddings).
    """,
    EXAONE_START_DOCSTRING,
)
class ExaoneForCausalLM(ExaonePreTrainedModel):

    def __init__(self, config):
        super().__init__(config)
        self.transformer = ExaoneModel(config)
        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)
        self.config = config
        # Initialize weights and apply final processing
        self.post_init()

    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(EXAONE_INPUTS_DOCSTRING)
    @add_code_sample_docstrings(
        checkpoint=_CHECKPOINT_FOR_DOC,
        output_type=BaseModelOutputWithPast,
        config_class=_CONFIG_FOR_DOC,
    )
    def forward(
        self,
        input_ids: Optional[torch.Tensor] = None,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.Tensor] = None,
        past_key_values: Optional[Cache] = None,
        inputs_embeds: Optional[torch.Tensor] = None,
        labels: Optional[torch.Tensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        cache_position: Optional[torch.LongTensor] = None,
    ) -> Union[Tuple[torch.Tensor], BaseModelOutputWithPast]:
        r"""
        Args:
            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]`

        Example:

        ```python
        >>> from transformers import AutoModelForCausalLM, AutoTokenizer

        >>> model = AutoModelForCausalLM.from_pretrained("LGAI-EXAONE/EXAONE-3.0-7.8B-Instruct",
                                                         trust_remote_code=True)
        >>> tokenizer = AutoTokenizer.from_pretrained("LGAI-EXAONE/EXAONE-3.0-7.8B-Instruct")

        >>> prompt = "Explain how wonderful you are"
        >>> messages = [
            {"role": "system", "content": "You are a helpful assistant."},
            {"role": "user", "content": prompt}
        ]
        >>> input_ids = tokenizer.apply_chat_template(
            messages,
            tokenize=True,
            add_generation_prompt=True,
            return_tensors="pt"
        )

        >>> output = model.generate(input_ids, max_new_tokens=128)
        >>> tokenizer.decode(output[0], skip_special_tokens=True)
        "[|system|]You are a helpful assistant.\n[|user|]Explain how wonderful you are\n[|assistant|]Thank you for your kind words! I'm here to assist you with information, answer questions, and help you in any way I can. My goal is to provide accurate, helpful, and timely responses. Whether you need help with a specific task, want to learn something new, or just need someone to talk to, I'm here for you. How can I assist you today?"
        ```
        """

        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
        transformer_outputs = self.transformer(
            input_ids,
            attention_mask=attention_mask,
            past_key_values=past_key_values,
            position_ids=position_ids,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
            cache_position=cache_position,
        )
        hidden_states = transformer_outputs[0]
        lm_logits = self.lm_head(hidden_states)
        lm_logits = lm_logits.float()
        loss = None
        if labels is not None:
            lm_logits = lm_logits.to(torch.float32)

            # Shift so that tokens < n predict n
            shift_logits = lm_logits[..., :-1, :].contiguous()
            shift_labels = labels[..., 1:].contiguous()
            # Flatten the tokens
            loss_fct = CrossEntropyLoss()
            loss = loss_fct(shift_logits.view(-1, shift_logits.size(-1)), shift_labels.view(-1))

            lm_logits = lm_logits.to(hidden_states.dtype)
            loss = loss.to(hidden_states.dtype)

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

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

    def prepare_inputs_for_generation(
        self,
        input_ids,
        past_key_values=None,
        attention_mask=None,
        inputs_embeds=None,
        cache_position=None,
        position_ids=None,
        use_cache=True,
        **kwargs,
    ):
        # If we have cache: let's slice `input_ids` through `cache_position`, to keep only the unprocessed tokens
        # Exception 1: when passing input_embeds, input_ids may be missing entries
        # Exception 2: some generation methods do special slicing of input_ids, so we don't need to do it here
        if past_key_values is not None:
            if inputs_embeds is not None:  # Exception 1
                input_ids = input_ids[:, -cache_position.shape[0] :]
            elif input_ids.shape[1] != cache_position.shape[0]:  # Default case (the "else", a no op, is Exception 2)
                input_ids = input_ids[:, cache_position]

        if attention_mask is not None and position_ids is None:
            # create position_ids on the fly for batch generation
            position_ids = attention_mask.long().cumsum(-1) - 1
            position_ids.masked_fill_(attention_mask == 0, 1)
            if past_key_values:
                position_ids = position_ids[:, -input_ids.shape[1] :]

                # This `clone` call is needed to avoid recapturing cuda graphs with `torch.compile`'s  `mode="reduce-overhead`, as otherwise the input `position_ids` would have various stride during the decoding. Here, simply using `.contiguous()` is not sufficient as in the batch size = 1 case, `position_ids` is already contiguous but with varying stride which retriggers a capture.
                position_ids = position_ids.clone(memory_format=torch.contiguous_format)

        # if `inputs_embeds` are passed, we only want to use them in the 1st generation step
        if inputs_embeds is not None and cache_position[0] == 0:
            model_inputs = {"inputs_embeds": inputs_embeds}
        else:
            model_inputs = {"input_ids": input_ids}

        if isinstance(past_key_values, StaticCache) and attention_mask.ndim == 2:
            if inputs_embeds is not None:
                batch_size, sequence_length = inputs_embeds.shape
                device = inputs_embeds.device
            else:
                batch_size, sequence_length = input_ids.shape
                device = input_ids.device

            dtype = self.lm_head.weight.dtype
            min_dtype = torch.finfo(dtype).min

            attention_mask = _prepare_4d_causal_attention_mask_with_cache_position(
                attention_mask,
                sequence_length=sequence_length,
                target_length=past_key_values.get_max_length(),
                dtype=dtype,
                device=device,
                min_dtype=min_dtype,
                cache_position=cache_position,
                batch_size=batch_size,
            )

        model_inputs.update(
            {
                "position_ids": position_ids,
                "cache_position": cache_position,
                "past_key_values": past_key_values,
                "use_cache": use_cache,
                "attention_mask": attention_mask,
            }
        )
        return model_inputs

    @staticmethod
    def _reorder_cache(past_key_values, beam_idx):
        reordered_past = ()
        for layer_past in past_key_values:
            reordered_past += (
                tuple(past_state.index_select(0, beam_idx.to(past_state.device)) for past_state in layer_past),
            )
        return reordered_past


@add_start_docstrings(
    """
    The EXAONE Model transformer with a sequence classification head on top (linear layer).

    :class:`~transformers.ExaoneForSequenceClassification` uses the last token in order to do the classification, as
    other causal models (e.g. GPT-1) do.

    Since it does classification on the last token, it requires to know the position of the last token. If a
    :obj:`pad_token_id` is defined in the configuration, it finds the last token that is not a padding token in each
    row. If no :obj:`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 :obj:`inputs_embeds` are passed instead of :obj:`input_ids`, it does the same (take
    the last value in each row of the batch).
    """,
    EXAONE_START_DOCSTRING,
)
class ExaoneForSequenceClassification(ExaonePreTrainedModel):
    _keys_to_ignore_on_load_missing = ["lm_head.weight"]
    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels
        self.transformer = ExaoneModel(config)
        self.score = nn.Linear(config.hidden_size, self.num_labels, bias=False)

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

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

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

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

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

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

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

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

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


@add_start_docstrings(
    """
    The EXAONE Model transformer with a span classification head on top for extractive question-answering tasks like
    SQuAD (a linear layers on top of the hidden-states output to compute `span start logits` and `span end logits`).
    """,
    EXAONE_START_DOCSTRING,
)
class ExaoneForQuestionAnswering(ExaonePreTrainedModel):
    _keys_to_ignore_on_load_missing = ["lm_head.weight"]

    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels
        self.transformer = ExaoneModel(config)
        self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels)

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

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

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

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

        sequence_output = outputs[0]

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

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

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

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

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