modeling_baichuan.py

import math
import os
from typing import List, Optional, Tuple, Union, Dict, Any

import torch
import torch.nn.functional as F
import torch.utils.checkpoint
from einops import rearrange
from torch import nn
from torch.nn import CrossEntropyLoss
from transformers import add_start_docstrings, PreTrainedModel, DynamicCache, \
    GenerationMixin, StaticCache, GenerationConfig
from transformers.activations import ACT2FN
from transformers.modeling_attn_mask_utils import AttentionMaskConverter
from transformers.modeling_flash_attention_utils import _flash_supports_window_size, \
    _upad_input
from transformers.modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast
from transformers.utils import is_flash_attn_2_available, is_flash_attn_greater_or_equal_2_10, \
    add_start_docstrings_to_model_forward, is_torchdynamo_compiling, logging, \
    is_flash_attn_greater_or_equal

if is_flash_attn_2_available():
    from flash_attn.bert_padding import  pad_input
    from flash_attn import flash_attn_func, flash_attn_varlen_func
    from flash_attn.layers.rotary import apply_rotary_emb_func
from .configuration_baichuan import BaichuanM1Config

logger = logging.get_logger(__name__)


class CustomCache(DynamicCache):
    def __init__(self):
        super().__init__()
        self.past_len = []

    def get_past_len(self, layer_idx: Optional[int] = 0) -> int:
        if len(self.past_len) <= layer_idx:
            return 0
        return self.past_len[layer_idx]

    def get_seq_length(self, layer_idx: Optional[int] = 0) -> int:
        """Returns the sequence length of the cached states. A layer index can be optionally passed."""
        # TODO: deprecate this function in favor of `cache_position`
        if len(self.key_cache) <= layer_idx:
            return 0
        return self.key_cache[layer_idx].shape[1]

    def update(
            self,
            key_states: torch.Tensor,
            value_states: torch.Tensor,
            layer_idx: int,
            cache_kwargs: Optional[Dict[str, Any]] = None,
    ) -> Tuple[torch.Tensor, torch.Tensor]:
        """
        Updates the cache with the new `key_states` and `value_states` for the layer `layer_idx`.

        Parameters:
            key_states (`torch.Tensor`):
                The new key states to cache.
            value_states (`torch.Tensor`):
                The new value states to cache.
            layer_idx (`int`):
                The index of the layer to cache the states for.
            cache_kwargs (`Dict[str, Any]`, `optional`):
                Additional arguments for the cache subclass. No additional arguments are used in `DynamicCache`.

        Return:
            A tuple containing the updated key and value states.
        """
        # Update the number of seen tokens
        if layer_idx == 0:
            self._seen_tokens += key_states.shape[1]

        # Update the cache
        if len(self.key_cache) <= layer_idx:
            self.key_cache.append(key_states)
            self.value_cache.append(value_states)
        else:
            self.key_cache[layer_idx] = torch.cat([self.key_cache[layer_idx], key_states], dim=1)
            self.value_cache[layer_idx] = torch.cat([self.value_cache[layer_idx], value_states], dim=1)

        if len(self.past_len) <= layer_idx:
            self.past_len.append(key_states.shape[1])
        else:
            self.past_len[layer_idx] += key_states.shape[1]

        return self.key_cache[layer_idx], self.value_cache[layer_idx]


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 BaichuanRMSNorm(nn.Module):
    def __init__(self, hidden_size, eps=1e-6):
        """
        RMSNorm is equivalent to T5LayerNorm
        """
        super().__init__()
        self.weight = nn.Parameter(torch.ones(hidden_size))
        self.variance_epsilon = eps

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

        # convert into half-precision if necessary
        if self.weight.dtype in [torch.float16, torch.bfloat16]:
            hidden_states = hidden_states.to(self.weight.dtype)

        return self.weight * hidden_states


class RotaryEmbedding(torch.nn.Module):
    def __init__(self, dim, max_position_embeddings=2048, base=1e5, device=None, interleaved=False):
        super().__init__()
        self.inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2).float() / dim))
        self.base = base
        self.dim = dim
        # Build here to make `torch.jit.trace` work.
        self.max_seq_len_cached = 0
        self.interleaved = interleaved

    def forward(self, q, k, seqlen_offset=None, cu_seqlens=None, max_seqlen=None):
        # x: [bs, num_attention_heads, seq_len, head_size]
        # This `if` block is unlikely to be run after we build sin/cos in `__init__`. Keep the logic here just in case.
        seq_len_dim = 1
        seq_len = q.shape[seq_len_dim] + seqlen_offset
        if seq_len > self.max_seq_len_cached:
            self.max_seq_len_cached = seq_len
            self.inv_freq = 1.0 / (
                    self.base ** (torch.arange(0, self.dim, 2).float().to(self.inv_freq.device) / self.dim))
            t = torch.arange(self.max_seq_len_cached, device=self.inv_freq.device, dtype=self.inv_freq.dtype)
            # freqs = torch.einsum("i,j->ij", t, self.inv_freq) # dont use this, bug in fp16
            freqs = torch.outer(t, self.inv_freq)
            self.cos_cached = freqs.cos().to(q.device)
            self.sin_cached = freqs.sin().to(k.device)
        q_ori_size = q.size()
        k_ori_size = k.size()
        if cu_seqlens is not None:
            q = flatten_one_dim(q)
            k = flatten_one_dim(k)
        q_new = apply_rotary_emb_func(
            q.float(), self.cos_cached[seqlen_offset:], self.sin_cached[seqlen_offset:],
            self.interleaved, True,  # inplace=True
            cu_seqlens=cu_seqlens,
            max_seqlen=max_seqlen
        ).to(q.dtype)
        k_new = apply_rotary_emb_func(
            k.float(), self.cos_cached[seqlen_offset:], self.sin_cached[seqlen_offset:],
            self.interleaved, True,
            cu_seqlens=cu_seqlens,
            max_seqlen=max_seqlen
        ).to(k.dtype)
        if cu_seqlens is not None:
            q_new = q_new.reshape(*q_ori_size)
            k_new = k_new.reshape(*k_ori_size)
        return q_new, k_new


class BaichuanMLP(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.hidden_size = config.hidden_size
        self.intermediate_size = config.intermediate_size
        self.gate_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
        self.up_proj = nn.Linear(self.hidden_size, self.intermediate_size, bias=False)
        self.down_proj = nn.Linear(self.intermediate_size, self.hidden_size, bias=False)
        self.act_fn = ACT2FN[config.hidden_act]

    def forward(self, hidden_state):
        return self.down_proj(self.act_fn(self.gate_proj(hidden_state)) * self.up_proj(hidden_state))


class BaichuanAttention(nn.Module):
    """
    Multi-headed attention from 'Attention Is All You Need' paper. Modified to use sliding window attention: Longformer
    and "Generating Long Sequences with Sparse Transformers".
    """

    def __init__(self, config: BaichuanM1Config, layer_idx: Optional[int] = None):
        super().__init__()
        self.config = config
        self.layer_idx = layer_idx
        if layer_idx is None:
            raise ValueError(
                f"Instantiating {self.__class__.__name__} without passing `layer_idx` is not recommended and will "
                "to errors during the forward call, if caching is used. Please make sure to provide a `layer_idx` "
                "when creating this class."
            )

        self.hidden_size = config.hidden_size
        self.is_swa = layer_idx in self.config.sliding_window_layers
        self.num_heads = config.num_swa_attention_heads if self.is_swa else config.num_attention_heads
        self.head_dim = self.hidden_size // self.num_heads
        self.num_key_value_heads = config.num_swa_key_value_heads if self.is_swa else config.num_key_value_heads
        self.max_position_embeddings = config.max_position_embeddings
        self.rope_theta = config.rope_theta
        self.is_causal = True
        self.attention_dropout = config.attention_dropout

        if (self.head_dim * self.num_heads) != self.hidden_size:
            raise ValueError(
                f"hidden_size must be divisible by num_heads (got `hidden_size`: {self.hidden_size}"
                f" and `num_heads`: {self.num_heads})."
            )
        self.W_pack = nn.Linear(config.hidden_size, self.hidden_size + 2 * self.num_key_value_heads * self.head_dim,
                                bias=False)
        self.o_proj = nn.Linear(self.num_heads * self.head_dim, self.hidden_size, bias=False)

        self.rotary_emb = RotaryEmbedding(dim=self.head_dim, base=self.config.rope_theta,
                                          max_position_embeddings=self.config.max_position_embeddings)
        self.conv_window = config.conv_window
        assert self.conv_window == 2 #%% Currently, only supported window=2 when inference
        self.conv_k = nn.Parameter(torch.softmax(torch.randn((1, 1, self.num_key_value_heads, 1, self.conv_window)), dim=-1))
        self.conv_v = nn.Parameter(torch.softmax(torch.randn((1, 1, self.num_key_value_heads, 1, self.conv_window)), dim=-1))
        self.last_k, self.last_v = None, None


def get_max_seqlen(cu_seqlens):
    max_seqlen = (cu_seqlens[1:] - cu_seqlens[:-1]).max().item()
    return max_seqlen


def flatten_one_dim(tensor):
    tensor = tensor.view(-1, tensor.size(-2), tensor.size(-1))
    return tensor


def prepare_for_flash_attention_varlen(query, key, value, cu_seqlens):
    query = query.view(-1, query.size(-2), query.size(-1))
    key = key.view(-1, key.size(-2), key.size(-1))
    value = value.view(-1, value.size(-2), value.size(-1))
    return query, key, value, get_max_seqlen(cu_seqlens)


def flash_attention_forward(
        query_states: torch.Tensor,
        key_states: torch.Tensor,
        value_states: torch.Tensor,
        query_length: int,
        is_causal: bool,
        dropout: float = 0.0,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.Tensor] = None,
        seqlens: Optional[torch.LongTensor] = None,
        softmax_scale: Optional[float] = None,
        sliding_window: Optional[int] = None,
        use_top_left_mask: bool = False,
        softcap: Optional[float] = None,
        deterministic: bool = None,
):
    """
    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. .
        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 - 1, 0)} if use_sliding_windows else {}

    if is_flash_attn_greater_or_equal("2.4.1"):
        if deterministic is None:
            deterministic = os.environ.get("FLASH_ATTENTION_DETERMINISTIC", "0") == "1"
        flash_kwargs["deterministic"] = deterministic

    if softcap is not None:
        flash_kwargs["softcap"] = softcap
    # Contains at least one padding token in the sequence
    if seqlens is not None:
        batch_size = query_states.shape[0]
        query_states, key_states, value_states, max_seqlen = prepare_for_flash_attention_varlen(query_states,
                                                                                                key_states,
                                                                                                value_states, seqlens)
        attn_output = flash_attn_varlen_func(
            query_states,
            key_states,
            value_states,
            cu_seqlens_q=seqlens,
            cu_seqlens_k=seqlens,
            max_seqlen_q=max_seqlen,
            max_seqlen_k=max_seqlen,
            dropout_p=dropout,
            softmax_scale=softmax_scale,
            causal=causal,
            **flash_kwargs,
        )

        attn_output = attn_output.reshape(batch_size, -1, attn_output.size(-2), attn_output.size(-1))

    elif 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 = _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


def custom_convolution(U, K): 
    """
    U: Input matrix, shape (bs, seq, h, d)
    K: Convolution kernel, shape (w, h)
    Returns: Output matrix V, shape (bs, seq, h, d)
    """
    # h, w = K.shape
    w = K.size(-1)
    padding = (w - 1, 0)
    U_padded = F.pad(U, (0, 0, 0, 0, *padding))  # Shape becomes (bs, seq+w-1, h, d)
    U_unfolded = U_padded.unfold(1, w, 1)  # Shape becomes (bs, seq+w-1, h, d, w)
    V_unfolded = U_unfolded * K  # Shape remains (bs, seq, h, d, w)
    V = V_unfolded.sum(dim=-1)  # Shape becomes (bs, seq, h, d)
    return V


def custom_convolution_with_splits(U, K, cu_seqlens):
    """
    U: Input matrix, shape (bs, seq, h, d)
    K: Convolution kernel, shape (w, h)
    cu_seqlens: Cumulative sequence lengths, indicating how to split the input.
    Returns: Output matrix, shape (bs, seq, h, d)
    """
    ori_shape = U.size()  # Save the original shape of U
    # Flatten U to handle variable-length sequences
    U_flatten = U.reshape(1, -1, ori_shape[-2], ori_shape[-1])  # Shape: (1, total_seq, h, d)

    # Perform convolution on each subsequence separately
    V_parts = []  # Store the results of each subsequence
    start = 0  # Start index of the current subsequence
    for end in cu_seqlens[1:]:
        end = end.item()  # Convert scalar tensor to int
        U_part = U_flatten[:, start:end, :, :]  # Slice the subsequence (1, seq_sub, h, d)
        V_part = custom_convolution(U_part, K)  # Apply custom convolution
        V_parts.append(V_part)  # Append the result
        start = end  # Update the start index for the next subsequence

    # Concatenate the results along the sequence dimension
    V = torch.cat(V_parts, dim=1).to(U)  # Shape: (1, total_seq, h, d)

    # Reshape the output to match the original input shape
    return V.reshape(ori_shape)


class BaichuanFlashAttention2(BaichuanAttention):
    """
    Baichuan flash attention module, following Baichuan attention module. This module inherits from `BaichuanAttention`
    as the weights of the module stays untouched. The only required change would be on the forward pass
    where it needs to correctly call the public API of flash attention and deal with padding tokens
    in case the input contains any of them. 
    """

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

        # TODO: Should be removed once Flash Attention for RoCm is bumped to 2.1.
        # 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. Reference: https://github.com/Dao-AILab/flash-attention/releases/tag/v2.1.0.
        # Beware that with flash_attn<2.1, using q_seqlen != k_seqlen (except for the case q_seqlen == 1) produces a wrong mask (top-left).
        self._flash_attn_uses_top_left_mask = not is_flash_attn_greater_or_equal_2_10()

    def forward(
            self,
            hidden_states: torch.Tensor,
            attention_mask: Optional[torch.Tensor] = None,
            position_ids: Optional[torch.LongTensor] = None,
            seqlens: Optional[torch.LongTensor] = None,
            past_key_value: Optional[CustomCache] = None,
            output_attentions: bool = False,
            use_cache: bool = False,
            cache_position: Optional[torch.LongTensor] = None,
            position_embeddings: Optional[Tuple[torch.Tensor, torch.Tensor]] = None,  # will become mandatory in v4.46
    ):

        bsz, q_len, _ = hidden_states.size()
        proj = self.W_pack(hidden_states)
        proj = rearrange(proj, 'bs seq_len (n_head head_dim) -> n_head bs seq_len head_dim', head_dim=self.head_dim)
        query_states = rearrange(proj[:self.num_heads], 'n_head bs seq_len head_dim -> bs seq_len n_head head_dim')
        key_states = rearrange(proj[self.num_heads:self.num_heads + self.num_key_value_heads],
                               'n_head bs seq_len head_dim -> bs seq_len n_head head_dim')
        value_states = rearrange(proj[self.num_heads + self.num_key_value_heads:],
                                 'n_head bs seq_len head_dim -> bs seq_len n_head head_dim')


        if past_key_value is None or past_key_value.get_seq_length(self.layer_idx) == 0:# prefill
            if not self.training:
                self.last_k = key_states[:, -1:]
                self.last_v = value_states[:, -1:]
            if seqlens is None:
                key_states = custom_convolution(key_states, self.conv_k)
                value_states = custom_convolution(value_states, self.conv_v)
            else:
                assert seqlens.ndim==1
                key_states=custom_convolution_with_splits(key_states,self.conv_k,seqlens)
                value_states=custom_convolution_with_splits(value_states,self.conv_v,seqlens)
        else: # decode
            self.last_k, key_states = key_states, self.conv_k[0, 0, :, 0, :1] * self.last_k + self.conv_k[0, 0, :, 0, 1:] * key_states
            self.last_v, value_states = value_states, self.conv_v[0, 0, :, 0, :1] * self.last_v + self.conv_v[0, 0, :, 0, 1:] * value_states
        if seqlens is not None:
            max_seqlen = get_max_seqlen(seqlens)
        else:
            max_seqlen = None

        past_len = past_key_value.get_past_len(self.layer_idx) if past_key_value is not None else 0
        query_states, key_states = self.rotary_emb(
            query_states,
            key_states,
            seqlen_offset=past_len,
            cu_seqlens=seqlens,
            max_seqlen=max_seqlen
        )

        if past_key_value is not None:
            cache_has_contents = past_key_value.get_seq_length(self.layer_idx) > 0
            kv_seq_len = key_states.shape[1] + past_key_value.get_seq_length(self.layer_idx)
            if (
                    self.is_swa
                    and kv_seq_len > self.config.sliding_window
                    and cache_has_contents
            ):
                slicing_tokens = 1 - self.config.sliding_window
                past_key = past_key_value[self.layer_idx][0]
                past_value = past_key_value[self.layer_idx][1]

                past_key_value.key_cache[self.layer_idx] = past_key[:, slicing_tokens:, :, :].contiguous()
                past_key_value.value_cache[self.layer_idx] = past_value[:, slicing_tokens:, :, :].contiguous()

                if past_key_value[self.layer_idx][0].shape[1] != self.config.sliding_window - 1:
                    raise ValueError(
                        f"past key must have a shape of (`batch_size, num_heads, self.config.sliding_window-1, head_dim`), got"
                        f" {past_key.shape}"
                    )

                # if attention_mask is not None:
                #     # TODO: not check!!
                #     attention_mask = attention_mask[:, slicing_tokens:]
                #     attention_mask = torch.cat([attention_mask, torch.ones_like(attention_mask[:, -1:])], dim=-1)
            key_states, value_states = past_key_value.update(key_states, value_states, self.layer_idx)

        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)

        if self.is_swa:
            sliding_window = self.config.sliding_window
        else:
            sliding_window = None
        attn_output = flash_attention_forward(
            query_states,
            key_states,
            value_states,
            query_length=q_len,
            position_ids=position_ids,
            seqlens=seqlens,
            sliding_window=sliding_window,
            is_causal=self.is_causal,
            use_top_left_mask=self._flash_attn_uses_top_left_mask,
        )
        attn_output = attn_output.reshape(bsz, q_len, self.hidden_size).contiguous()
        attn_output = self.o_proj(attn_output)

        if not output_attentions:
            attn_weights = None

        return attn_output, attn_weights, past_key_value


Baichuan_ATTENTION_CLASSES = {
    "eager": BaichuanAttention,
    "flash_attention_2": BaichuanFlashAttention2,
}


class BaichuanDecoderLayer(nn.Module):
    def __init__(self, config: BaichuanM1Config, layer_idx: int):
        super().__init__()
        self.hidden_size = config.hidden_size
        self.layer_idx = layer_idx
        self.self_attn = Baichuan_ATTENTION_CLASSES['flash_attention_2'](config, layer_idx)

        self.mlp = BaichuanMLP(config)
        self.input_layernorm = BaichuanRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
        self.post_attention_layernorm = BaichuanRMSNorm(config.hidden_size, eps=config.rms_norm_eps)

    def forward(
            self,
            hidden_states: torch.Tensor,
            attention_mask: Optional[torch.Tensor] = None,
            position_ids: Optional[torch.LongTensor] = None,
            seqlens: Optional[torch.LongTensor] = None,
            past_key_value: Optional[Tuple[torch.Tensor]] = 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]]]:
        """
        Args:
            hidden_states (`torch.FloatTensor`): input to the layer of shape `(batch, seq_len, embed_dim)`
            attention_mask (`torch.FloatTensor`, *optional*): attention mask of size
                `(batch, sequence_length)` where padding elements are indicated by 0.
            output_attentions (`bool`, *optional*):
                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
                returned tensors for more detail.
            use_cache (`bool`, *optional*):
                If set to `True`, `past_key_values` key value states are returned and can be used to speed up decoding
                (see `past_key_values`).
            past_key_value (`Tuple(torch.FloatTensor)`, *optional*): cached past key and value projection states
            cache_position (`torch.LongTensor` of shape `(sequence_length)`, *optional*):
                Indices depicting the position of the input sequence tokens in the sequence.
            position_embeddings (`Tuple[torch.FloatTensor, torch.FloatTensor]`, *optional*):
                Tuple containing the cosine and sine positional embeddings of shape `(batch_size, seq_len, head_dim)`,
                with `head_dim` being the embedding dimension of each attention head.
            kwargs (`dict`, *optional*):
                Arbitrary kwargs to be ignored, used for FSDP and other methods that injects code
                into the model
        """

        residual = hidden_states

        hidden_states = self.input_layernorm(hidden_states)

        hidden_states, self_attn_weights, present_key_value = self.self_attn(
            hidden_states=hidden_states,
            attention_mask=attention_mask,
            position_ids=position_ids,
            seqlens=seqlens,
            past_key_value=past_key_value,
            output_attentions=output_attentions,
            use_cache=use_cache,
            cache_position=cache_position,
            position_embeddings=position_embeddings,
        )
        hidden_states = residual + hidden_states

        # Fully Connected
        residual = hidden_states
        hidden_states = self.post_attention_layernorm(hidden_states)
        hidden_states = self.mlp(hidden_states)
        hidden_states = residual + hidden_states

        outputs = (hidden_states,)

        if output_attentions:
            outputs += (self_attn_weights,)

        if use_cache:
            outputs += (present_key_value,)
        return outputs


Baichuan_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 ([`BaichuanM1Config`]):
            Model configuration class with all the parameters of the model. Initializing with a config file does not
            load the weights associated with the model, only the configuration. Check out the
            [`~PreTrainedModel.from_pretrained`] method to load the model weights.
"""


@add_start_docstrings(
    "The bare Bai chuan Model outputting raw hidden-states without any specific head on top.",
    Baichuan_START_DOCSTRING,
)
class BaichuanPreTrainedModel(PreTrainedModel):
    config_class = BaichuanM1Config
    base_model_prefix = "model"
    supports_gradient_checkpointing = True
    _no_split_modules = ["BaichuanDecoderLayer"]
    _skip_keys_device_placement = "past_key_values"
    _supports_flash_attn_2 = True
    _supports_sdpa = True
    _supports_cache_class = True
    _supports_quantized_cache = True
    _supports_static_cache = True

    def _init_weights(self, module):
        std = self.config.initializer_range
        if isinstance(module, nn.Linear):
            module.weight.data.normal_(mean=0.0, std=std)
            if module.bias is not None:
                module.bias.data.zero_()
        elif isinstance(module, nn.Embedding):
            module.weight.data.normal_(mean=0.0, std=std)
            if module.padding_idx is not None:
                module.weight.data[module.padding_idx].zero_()


Baichuan_INPUTS_DOCSTRING = r"""

"""


@add_start_docstrings(
    "The bare Baichuan Model outputting raw hidden-states without any specific head on top.",
    Baichuan_START_DOCSTRING,
)
class BaichuanModel(BaichuanPreTrainedModel):
    """
    Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`BaichuanDecoderLayer`]

    Args:
        config: BaichuanM1Config
    """

    def __init__(self, config: BaichuanM1Config):
        super().__init__(config)
        self.padding_idx = config.pad_token_id
        self.vocab_size = config.vocab_size

        self.embed_tokens = nn.Embedding(config.vocab_size, config.hidden_size, self.padding_idx)
        self.layers = nn.ModuleList(
            [BaichuanDecoderLayer(config, layer_idx) for layer_idx in range(config.num_hidden_layers)]
        )
        self._attn_implementation = config._attn_implementation
        self.norm = BaichuanRMSNorm(config.hidden_size, eps=config.rms_norm_eps)

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

    def get_input_embeddings(self):
        return self.embed_tokens

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

    @add_start_docstrings_to_model_forward(Baichuan_INPUTS_DOCSTRING)
    def forward(
            self,
            input_ids: torch.LongTensor = None,
            attention_mask: Optional[torch.Tensor] = None,
            position_ids: Optional[torch.LongTensor] = None,
            seqlens: Optional[torch.LongTensor] = None,
            past_key_values: Optional[CustomCache] = None,
            inputs_embeds: Optional[torch.FloatTensor] = None,
            use_cache: Optional[bool] = None,
            output_attentions: Optional[bool] = None,
            output_hidden_states: Optional[bool] = None,
            return_dict: Optional[bool] = None,
            cache_position: Optional[torch.LongTensor] = None,
    ) -> Union[Tuple, BaseModelOutputWithPast]:
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        use_cache = use_cache if use_cache is not None else self.config.use_cache
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
        if (input_ids is None) ^ (inputs_embeds is not None):
            raise ValueError(
                "You cannot specify both input_ids and inputs_embeds at the same time, and must specify either one"
            )

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

        if seqlens is not None:
            assert seqlens.ndim == 2
                # batch multi-pack 样本拉平
            cu_seqlens = []
            offset, seqlen = 0, seqlens.size(1)
            for lens in seqlens:
                cu_seqlens.append(offset)
                cu_seqlens.extend((lens[(lens > 0) & (lens < seqlen)] + offset).tolist())
                offset += seqlen
            cu_seqlens.append(offset)
            seqlens = torch.tensor(cu_seqlens, dtype=torch.int32, device=input_ids.device)
            # unset attention_mask to save memory
            attention_mask = None
        # kept for BC (non `Cache` `past_key_values` inputs)
        return_legacy_cache = False
        if use_cache and not isinstance(past_key_values, CustomCache):
            return_legacy_cache = False
            if past_key_values is None:
                past_key_values = CustomCache()
            else:
                past_key_values = CustomCache.from_legacy_cache(past_key_values)
                logger.warning_once(
                    "We detected that you are passing `past_key_values` as a tuple of tuples. This is deprecated and "
                    "will be removed in v4.47. Please convert your cache or use an appropriate `Cache` class "
                    "(https://huggingface.co/docs/transformers/kv_cache#legacy-cache-format)"
                )
        if inputs_embeds is None:
            inputs_embeds = self.embed_tokens(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

        # create position embeddings to be shared across the decoder layers
        # position_embeddings = self.rotary_emb(hidden_states, position_ids)
        position_embeddings = None

        # decoder layers
        all_hidden_states = () if output_hidden_states else None
        all_self_attns = () if output_attentions else None
        next_decoder_cache = None

        for decoder_layer in self.layers:
            if output_hidden_states:
                all_hidden_states += (hidden_states,)
            if self.gradient_checkpointing and self.training:
                layer_outputs = torch.utils.checkpoint.checkpoint(
                    decoder_layer,
                    hidden_states,
                    causal_mask,
                    position_ids,
                    seqlens,
                    past_key_values,
                    output_attentions,
                    use_cache,
                    cache_position,
                    position_embeddings,
                )
            else:
                layer_outputs = decoder_layer(
                    hidden_states,
                    attention_mask=causal_mask,
                    position_ids=position_ids,
                    seqlens=seqlens,
                    past_key_value=past_key_values,
                    output_attentions=output_attentions,
                    use_cache=use_cache,
                    cache_position=cache_position,
                    position_embeddings=position_embeddings,
                )

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

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

        hidden_states = self.norm(hidden_states)

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

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


    def _update_causal_mask(
            self,
            attention_mask: torch.Tensor,
            input_tensor: torch.Tensor,
            cache_position: torch.Tensor,
            past_key_values: CustomCache,
            output_attentions: bool,
    ):
        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


class NormHead(nn.Module):
    def __init__(self, hidden_size, vocab_size, bias=False):
        super().__init__()
        self.weight = nn.Parameter(torch.empty((vocab_size, hidden_size)))
        nn.init.kaiming_uniform_(self.weight, a=math.sqrt(5))

    def forward(self, hidden_states):
        norm_weight = nn.functional.normalize(self.weight)
        return nn.functional.linear(hidden_states, norm_weight)


class BaichuanM1ForCausalLM(BaichuanPreTrainedModel, GenerationMixin):
    _tied_weights_keys = ["lm_head.weight"]

    def __init__(self, config):
        super().__init__(config)
        self.model = BaichuanModel(config)
        self.vocab_size = config.vocab_size
        self.lm_head = NormHead(config.hidden_size, config.vocab_size, bias=False)

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

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

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

    def get_output_embeddings(self):
        return self.lm_head

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

    def set_decoder(self, decoder):
        self.model = decoder

    def get_decoder(self):
        return self.model

    @add_start_docstrings_to_model_forward(Baichuan_INPUTS_DOCSTRING)
    def forward(
            self,
            input_ids: torch.LongTensor = None,
            attention_mask: Optional[torch.Tensor] = None,
            position_ids: Optional[torch.LongTensor] = None,
            seqlens: Optional[torch.LongTensor] = None,
            past_key_values: Optional[List[torch.FloatTensor]] = None,
            inputs_embeds: Optional[torch.FloatTensor] = None,
            labels: Optional[torch.LongTensor] = None,
            use_cache: Optional[bool] = None,
            output_attentions: Optional[bool] = None,
            output_hidden_states: Optional[bool] = None,
            return_dict: Optional[bool] = None,
            cache_position: Optional[torch.LongTensor] = None,
            num_logits_to_keep: int = 0,
    ) -> Union[Tuple, CausalLMOutputWithPast]:
        r"""
        Args:
            labels (`torch.LongTensor` of shape `(batch_size, sequence_length)`, *optional*):
                Labels for computing the masked language modeling loss. Indices should either be in `[0, ...,
                config.vocab_size]` or -100 (see `input_ids` docstring). Tokens with indices set to `-100` are ignored
                (masked), the loss is only computed for the tokens with labels in `[0, ..., config.vocab_size]`.

            num_logits_to_keep (`int`, *optional*):
                Calculate logits for the last `num_logits_to_keep` tokens. If `0`, calculate logits for all
                `input_ids` (special case). Only last token logits are needed for generation, and calculating them only for that
                token can save memory, which becomes pretty significant for long sequences or large vocabulary size.

        Returns:

        Example:

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

        >>> model = BaichuanForCausalLM.from_pretrained(PATH_TO_CONVERTED_WEIGHTS)
        >>> tokenizer = AutoTokenizer.from_pretrained(PATH_TO_CONVERTED_TOKENIZER)

        >>> prompt = "Hey, are you conscious? Can you talk to me?"
        >>> inputs = tokenizer(prompt, return_tensors="pt")

        >>> # Generate
        >>> generate_ids = model.generate(inputs.input_ids, max_length=30)
        >>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
        "Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
        ```"""
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
        if input_ids is not None:
            input_ids[input_ids == self.config.vocab_size] = 0
        if labels is not None:
            labels[labels == self.config.vocab_size] = 0
        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
        outputs = self.model(
            input_ids=input_ids,
            attention_mask=attention_mask,
            position_ids=position_ids,
            seqlens=seqlens,
            past_key_values=past_key_values,
            inputs_embeds=inputs_embeds,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
            cache_position=cache_position,
        )

        hidden_states = outputs[0]
        if labels is None and not is_torchdynamo_compiling():
            logger.warning_once(
                "Starting from v4.46, the `logits` model output will have the same type as the model (except at train time, where it will always be FP32)"
            )
        # Only compute necessary logits, and do not upcast them to float if we are not computing the loss
        # TODO: remove the float() operation in v4.46
        logits = self.lm_head(hidden_states[:, -num_logits_to_keep:, :])

        loss = None
        if labels is not None:
            # Upcast to float if we need to compute the loss to avoid potential precision issues
            # logits = logits.float()
            # Shift so that tokens < n predict n
            shift_logits = logits[..., :-1, :].contiguous()
            shift_labels = labels[..., 1:].contiguous()
            #shift_logits = logits
            #shift_labels = labels
            # Flatten the tokens
            loss_fct = CrossEntropyLoss()
            shift_logits = shift_logits.view(-1, self.config.vocab_size)
            shift_labels = shift_labels.view(-1)
            # Enable model parallelism
            shift_labels = shift_labels.to(shift_logits.device)
            loss = loss_fct(shift_logits, shift_labels)

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

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

    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,
            num_logits_to_keep=None,
            **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, "input_ids": None}
        else:
            # The clone here is for the same reason as for `position_ids`.
            model_inputs = {"input_ids": input_ids.clone(memory_format=torch.contiguous_format), "inputs_embeds": None}

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

        if num_logits_to_keep is not None:
            model_inputs["num_logits_to_keep"] = num_logits_to_keep

        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

    @torch.no_grad()
    def chat(self, tokenizer, messages: List[dict], stream=False,
             generation_config: Optional[GenerationConfig] = None):
        generation_config = generation_config or self.generation_config
        input_ids = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True)
        input_ids = torch.LongTensor([input_ids]).to(self.device)
        if stream:
            streamer = TextIterStreamer(tokenizer, skip_prompt=True, skip_special_tokens=True)
            Thread(target=self.generate, kwargs=dict(
                inputs=input_ids, streamer=streamer,
                generation_config=generation_config,
            )).start()
            return streamer
        else:
            outputs = self.generate(input_ids, generation_config=generation_config)
            response = tokenizer.decode(outputs[0][len(input_ids[0]):], skip_special_tokens=True)
            return response