modeling.py

from dataclasses import dataclass
from typing import Dict
from typing import Iterable, Optional

import numpy as np
import torch
import torch.nn.functional as F
from torch import Tensor
from torch import nn
from transformers import GenerationConfig
from collections import OrderedDict
from typing import Tuple, Union
from typing import List, Optional, Tuple, Union
import numpy as np
import torch
import torch.nn.functional as F
from torch import nn
from torch.nn import CrossEntropyLoss
from transformers import CLIPProcessor, CLIPModel
from torch import Tensor
from transformers.utils import logging
import math
import random
from transformers.activations import ACT2FN
from transformers.modeling_utils import PreTrainedModel, PretrainedConfig
from transformers.modeling_outputs import BaseModelOutput
import copy
logger = logging.get_logger(__name__)
# Copyright (c) Meta Platforms, Inc. and affiliates.
# This software may be used and distributed according to the terms of the GNU General Public License version 3.

from typing import Optional, Tuple
from dataclasses import dataclass
import math

from transformers import CLIPConfig, WhisperConfig, LlamaConfig
from transformers.modeling_outputs import BaseModelOutputWithPast, CausalLMOutputWithPast, SequenceClassifierOutputWithPast
from transformers import LlamaConfig, WhisperConfig, WhisperPreTrainedModel, WhisperModel
from transformers.models.clip.modeling_clip import CLIPVisionTransformer
from transformers.models.whisper.modeling_whisper import WhisperEncoderLayer


# Copied from transformers.models.bart.modeling_bart._make_causal_mask
def _make_causal_mask(
    input_ids_shape: torch.Size, dtype: torch.dtype, device: torch.device, past_key_values_length: int = 0
):
    """
    Make causal mask used for bi-directional self-attention.
    """
    bsz, tgt_len = input_ids_shape
    mask = torch.full((tgt_len, tgt_len), torch.tensor(torch.finfo(dtype).min, device=device), device=device)
    mask_cond = torch.arange(mask.size(-1), device=device)
    mask.masked_fill_(mask_cond < (mask_cond + 1).view(mask.size(-1), 1), 0)
    mask = mask.to(dtype)

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


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

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

    inverted_mask = 1.0 - expanded_mask

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


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)


def apply_rotary_pos_emb(q, k, cos, sin, position_ids):
    # The first two dimensions of cos and sin are always 1, so we can `squeeze` them.
    cos = cos.squeeze(1).squeeze(0)  # [seq_len, dim]
    sin = sin.squeeze(1).squeeze(0)  # [seq_len, dim]
    cos = cos[position_ids].unsqueeze(1)  # [bs, 1, seq_len, dim]
    sin = sin[position_ids].unsqueeze(1)  # [bs, 1, seq_len, dim]
    q_embed = (q * cos) + (rotate_half(q) * sin)
    k_embed = (k * cos) + (rotate_half(k) * sin)
    return q_embed, k_embed


class LlamaRotaryEmbedding(torch.nn.Module):
    def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None):
        super().__init__()
        inv_freq = 1.0 / (base ** (torch.arange(0, dim, 2).float().to(device) / dim))
        self.register_buffer("inv_freq", inv_freq)

        # Build here to make `torch.jit.trace` work.
        self.max_seq_len_cached = max_position_embeddings
        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)
        # Different from paper, but it uses a different permutation in order to obtain the same calculation
        emb = torch.cat((freqs, freqs), dim=-1)
        self.register_buffer("cos_cached", emb.cos()[None, None, :, :], persistent=False)
        self.register_buffer("sin_cached", emb.sin()[None, None, :, :], persistent=False)

    def forward(self, x, seq_len=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.
        if seq_len > self.max_seq_len_cached:
            self.max_seq_len_cached = seq_len
            t = torch.arange(self.max_seq_len_cached, device=x.device, dtype=self.inv_freq.dtype)
            freqs = torch.einsum("i,j->ij", t, self.inv_freq)
            # Different from paper, but it uses a different permutation in order to obtain the same calculation
            emb = torch.cat((freqs, freqs), dim=-1).to(x.device)
            self.register_buffer("cos_cached", emb.cos()[None, None, :, :], persistent=False)
            self.register_buffer("sin_cached", emb.sin()[None, None, :, :], persistent=False)
        return (
            self.cos_cached[:, :, :seq_len, ...].to(dtype=x.dtype),
            self.sin_cached[:, :, :seq_len, ...].to(dtype=x.dtype),
        )


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

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


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

    def __init__(self, config: LlamaConfig):
        super().__init__()
        self.config = config
        self.hidden_size = config.hidden_size
        self.num_heads = config.num_attention_heads
        self.head_dim = self.hidden_size // self.num_heads
        self.max_position_embeddings = config.max_position_embeddings

        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.q_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=False)
        self.k_proj = nn.Linear(self.hidden_size, self.num_heads * self.head_dim, bias=False)
        self.v_proj = nn.Linear(self.hidden_size, self.num_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 = LlamaRotaryEmbedding(self.head_dim, max_position_embeddings=self.max_position_embeddings)

    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[Tuple[torch.Tensor]] = None,
        output_attentions: bool = False,
        use_cache: bool = False,
    ) -> Tuple[torch.Tensor, Optional[torch.Tensor], Optional[Tuple[torch.Tensor]]]:
        bsz, q_len, _ = hidden_states.size()

        query_states = self.q_proj(hidden_states).view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
        key_states = self.k_proj(hidden_states).view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)
        value_states = self.v_proj(hidden_states).view(bsz, q_len, self.num_heads, self.head_dim).transpose(1, 2)

        kv_seq_len = key_states.shape[-2]
        if past_key_value is not None:
            kv_seq_len += past_key_value[0].shape[-2]
        cos, sin = self.rotary_emb(value_states, seq_len=kv_seq_len)
        query_states, key_states = apply_rotary_pos_emb(query_states, key_states, cos, sin, position_ids)
        # [bsz, nh, t, hd]

        if past_key_value is not None:
            # reuse k, v, self_attention
            key_states = torch.cat([past_key_value[0], key_states], dim=2)
            value_states = torch.cat([past_key_value[1], value_states], dim=2)

        past_key_value = (key_states, value_states) if use_cache else None

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

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

        if attention_mask is not None:
            if attention_mask.size() != (bsz, 1, q_len, kv_seq_len):
                raise ValueError(
                    f"Attention mask should be of size {(bsz, 1, q_len, kv_seq_len)}, but is {attention_mask.size()}"
                )
            attn_weights = attn_weights + attention_mask
            attn_weights = torch.max(attn_weights, torch.tensor(torch.finfo(attn_weights.dtype).min))

        # upcast attention to fp32
        attn_weights = nn.functional.softmax(attn_weights, dim=-1, dtype=torch.float32).to(query_states.dtype)
        attn_output = torch.matmul(attn_weights, value_states)

        if attn_output.size() != (bsz, self.num_heads, q_len, self.head_dim):
            raise ValueError(
                f"`attn_output` 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)
        attn_output = attn_output.reshape(bsz, q_len, self.hidden_size)

        attn_output = self.o_proj(attn_output)

        if not output_attentions:
            attn_weights = None

        return attn_output, attn_weights, past_key_value


class LlamaDecoderLayer(nn.Module):
    def __init__(self, config: LlamaConfig):
        super().__init__()
        self.hidden_size = config.hidden_size
        self.self_attn = LlamaAttention(config=config)
        self.mlp = LlamaMLP(
            hidden_size=self.hidden_size,
            intermediate_size=config.intermediate_size,
            hidden_act=config.hidden_act,
        )
        self.input_layernorm = LlamaRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
        self.post_attention_layernorm = LlamaRMSNorm(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,
        past_key_value: Optional[Tuple[torch.Tensor]] = None,
        output_attentions: Optional[bool] = False,
        use_cache: Optional[bool] = False,
    ) -> 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, 1, tgt_len, src_len)` where padding elements are indicated by very large negative values.
            output_attentions (`bool`, *optional*):
                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
                returned tensors for more detail.
            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
        """

        residual = hidden_states

        hidden_states = self.input_layernorm(hidden_states)

        # Self Attention
        hidden_states, self_attn_weights, present_key_value = self.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,
        )
        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


class LlamaRMSNorm(nn.Module):
    def __init__(self, hidden_size, eps=1e-6):
        """
        LlamaRMSNorm 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 LlamaPreTrainedModel(PreTrainedModel):
    config_class = LlamaConfig
    base_model_prefix = "model"
    supports_gradient_checkpointing = True
    _no_split_modules = ["LlamaDecoderLayer"]
    _keys_to_ignore_on_load_unexpected = [r"decoder\.version"]

    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_()

    def _set_gradient_checkpointing(self, module, value=False):
        if isinstance(module, LlamaModel):
            module.gradient_checkpointing = value


class LlamaModel(LlamaPreTrainedModel):
    """
    Transformer decoder consisting of *config.num_hidden_layers* layers. Each layer is a [`LlamaDecoderLayer`]

    Args:
        config: LlamaConfig
    """

    def __init__(self, config: LlamaConfig):
        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([LlamaDecoderLayer(config) for _ in range(config.num_hidden_layers)])
        self.norm = LlamaRMSNorm(config.hidden_size, eps=config.rms_norm_eps)

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

    def get_input_embeddings(self):
        return self.embed_tokens

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

    # Copied from transformers.models.bart.modeling_bart.BartDecoder._prepare_decoder_attention_mask
    def _prepare_decoder_attention_mask(self, attention_mask, input_shape, inputs_embeds, past_key_values_length):
        # create causal mask
        # [bsz, seq_len] -> [bsz, 1, tgt_seq_len, src_seq_len]
        combined_attention_mask = None
        if input_shape[-1] > 1:
            combined_attention_mask = _make_causal_mask(
                input_shape,
                inputs_embeds.dtype,
                device=inputs_embeds.device,
                past_key_values_length=past_key_values_length,
            )

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

        return combined_attention_mask

    # @add_start_docstrings_to_model_forward(LLAMA_INPUTS_DOCSTRING)
    def forward(
        self,
        input_ids: torch.LongTensor = None,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        past_key_values: Optional[List[torch.FloatTensor]] = None,
        inputs_embeds: Optional[torch.FloatTensor] = None,
        use_cache: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        output_hidden_states: Optional[bool] = None,
        return_dict: Optional[bool] = None,
    ) -> Union[Tuple, 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

        # retrieve input_ids and inputs_embeds
        if input_ids is not None and inputs_embeds is not None:
            raise ValueError("You cannot specify both decoder_input_ids and decoder_inputs_embeds at the same time")
        elif input_ids is not None:
            batch_size, seq_length = input_ids.shape
        elif inputs_embeds is not None:
            batch_size, seq_length, _ = inputs_embeds.shape
        else:
            raise ValueError("You have to specify either decoder_input_ids or decoder_inputs_embeds")

        seq_length_with_past = seq_length
        past_key_values_length = 0

        if past_key_values is not None:
            past_key_values_length = past_key_values[0][0].shape[2]
            seq_length_with_past = seq_length_with_past + past_key_values_length

        if position_ids is None:
            device = input_ids.device if input_ids is not None else inputs_embeds.device
            position_ids = torch.arange(
                past_key_values_length, seq_length + past_key_values_length, dtype=torch.long, device=device
            )
            position_ids = position_ids.unsqueeze(0).view(-1, seq_length)
        else:
            position_ids = position_ids.view(-1, seq_length).long()

        if inputs_embeds is None:
            inputs_embeds = self.embed_tokens(input_ids)
        # embed positions
        if attention_mask is None:
            attention_mask = torch.ones(
                (batch_size, seq_length_with_past), dtype=torch.bool, device=inputs_embeds.device
            )
        attention_mask = self._prepare_decoder_attention_mask(
            attention_mask, (batch_size, seq_length), inputs_embeds, past_key_values_length
        )

        hidden_states = inputs_embeds

        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

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

        for idx, decoder_layer in enumerate(self.layers):
            if output_hidden_states:
                all_hidden_states += (hidden_states,)

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

            if self.gradient_checkpointing and self.training:

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

                    return custom_forward

                layer_outputs = torch.utils.checkpoint.checkpoint(
                    create_custom_forward(decoder_layer),
                    hidden_states,
                    attention_mask,
                    position_ids,
                    None,
                )
            else:
                layer_outputs = decoder_layer(
                    hidden_states,
                    attention_mask=attention_mask,
                    position_ids=position_ids,
                    past_key_value=past_key_value,
                    output_attentions=output_attentions,
                    use_cache=use_cache,
                )

            hidden_states = layer_outputs[0]

            if use_cache:
                next_decoder_cache += (layer_outputs[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 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,
        )


class LlamaForCausalLM(LlamaPreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.model = LlamaModel(config)

        self.lm_head = nn.Linear(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(LLAMA_INPUTS_DOCSTRING)
    # @replace_return_docstrings(output_type=CausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC)
    def forward(
        self,
        input_ids: torch.LongTensor = None,
        attention_mask: Optional[torch.Tensor] = None,
        position_ids: 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,
    ) -> Union[Tuple, CausalLMOutputWithPast]:

        # 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]`.
        #
        # Returns:

        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

        # 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,
            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 = outputs[0]
        logits = self.lm_head(hidden_states)

        loss = None
        if labels is not None:
            # Shift so that tokens < n predict n
            shift_logits = logits[..., :-1, :].contiguous()
            shift_labels = labels[..., 1:].contiguous()
            # 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, **kwargs
    ):
        if past_key_values:
            input_ids = input_ids[:, -1:]

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

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

        model_inputs.update(
            {
                "position_ids": position_ids,
                "past_key_values": past_key_values,
                "use_cache": kwargs.get("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) for past_state in layer_past),)
        return reordered_past


class WhisperEncoder(WhisperPreTrainedModel):
    """
    Transformer encoder consisting of *config.encoder_layers* self attention layers. Each layer is a
    [`WhisperEncoderLayer`].

    Args:
        config: WhisperConfig
    """

    def __init__(self, config: WhisperConfig):
        super().__init__(config)
        self.dropout = config.dropout
        self.layerdrop = config.encoder_layerdrop

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

        self.conv1 = nn.Conv1d(self.num_mel_bins, embed_dim, kernel_size=3, padding=1)
        self.conv2 = nn.Conv1d(embed_dim, embed_dim, kernel_size=3, stride=2, padding=1)

        self.embed_positions = nn.Embedding(self.max_source_positions, embed_dim)

        self.layers = nn.ModuleList([WhisperEncoderLayer(config) for _ in range(config.encoder_layers)])
        self.layer_norm = nn.LayerNorm(config.d_model)

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

    def _freeze_parameters(self):
        for param in self.parameters():
            param.requires_grad = False
        self._requires_grad = False

    def get_input_embeddings(self) -> nn.Module:
        return self.conv1

    def set_input_embeddings(self, value: nn.Module):
        self.conv1 = value

    def forward(
        self,
        input_features,
        attention_mask=None,
        head_mask=None,
        output_attentions=None,
        output_hidden_states=None,
        return_dict=None,
    ):
        r"""
        Args:
            input_features (`torch.LongTensor` of shape `(batch_size, feature_size, sequence_length)`):
                Float values of mel features extracted from the raw speech waveform. Raw speech waveform can be
                obtained by loading a `.flac` or `.wav` audio file into an array of type `List[float]` or a
                `numpy.ndarray`, *e.g.* via the soundfile library (`pip install soundfile`). To prepare the array into
                `input_features`, the [`AutoFeatureExtractor`] should be used for extracting the mel features, padding
                and conversion into a tensor of type `torch.FloatTensor`. See [`~WhisperFeatureExtractor.__call__`]
            attention_mask (`torch.Tensor`)`, *optional*):
                Whisper does not support masking of the `input_features`, this argument is preserved for compatibility,
                but it is not used. By default the silence in the input log mel spectrogram are ignored.
            head_mask (`torch.Tensor` of shape `(encoder_layers, encoder_attention_heads)`, *optional*):
                Mask to nullify selected heads of the attention modules. Mask values selected in `[0, 1]`:

                - 1 indicates the head is **not masked**,
                - 0 indicates the head is **masked**.
            output_attentions (`bool`, *optional*):
                Whether or not to return the attentions tensors of all attention layers. See `attentions` under
                returned tensors for more detail.
            output_hidden_states (`bool`, *optional*):
                Whether or not to return the hidden states of all layers. See `hidden_states` under returned tensors
                for more detail.
            return_dict (`bool`, *optional*):
                Whether or not to return a [`~utils.ModelOutput`] instead of a plain tuple.
        """
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
        inputs_embeds = nn.functional.gelu(self.conv1(input_features))
        inputs_embeds = nn.functional.gelu(self.conv2(inputs_embeds))

        inputs_embeds = inputs_embeds.permute(0, 2, 1)
        embed_pos = self.embed_positions.weight

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

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

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

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

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

                        return custom_forward

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

                hidden_states = layer_outputs[0]

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

        hidden_states = self.layer_norm(hidden_states)
        if output_hidden_states:
            encoder_states = encoder_states + (hidden_states,)

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

class MM_LLMs_Config(PretrainedConfig):
    model_type = 'mm_llms'
    is_composition = True

    def __init__(self, n_frames=6, attention_heads=8, image_conv_kernel=48, image_conv_stride=36, 
    video_conv_kernel=36, video_conv_stride=30, audio_conv_kernel=240, audio_conv_stride=220,
    clip_config=None, whisper_config=None, llm_config=None, **kwargs):

        self.image_config = clip_config
        self.audio_config = whisper_config
        self.llm_config = llm_config
        self.n_frames = n_frames
        self.attention_heads = attention_heads
        self.image_conv_kernel = image_conv_kernel
        self.image_conv_stride = image_conv_stride
        self.video_conv_kernel = video_conv_kernel
        self.video_conv_stride = video_conv_stride
        self.audio_conv_kernel = audio_conv_kernel
        self.audio_conv_stride = audio_conv_stride

        self.hidden_size = max(llm_config.hidden_size, clip_config.projection_dim, whisper_config.d_model, clip_config.projection_dim)

        super().__init__(**kwargs)

    def to_dict(self):
        """
        Serializes this instance to a Python dictionary. Override the default [`~PretrainedConfig.to_dict`].

        Returns:
            `Dict[str, any]`: Dictionary of all the attributes that make up this configuration instance,
        """
        output = copy.deepcopy(self.__dict__)
        output["image_config"] = self.image_config.to_dict()
        output["audio_config"] = self.audio_config.to_dict()
        output['llm_config'] = self.llm_config.to_dict()
        output['n_frames'] = self.n_frames
        output['attention_heads'] = self.attention_heads
        output['image_conv_kernel'] = self.image_conv_kernel
        output['image_conv_stride'] = self.image_conv_stride
        output['video_conv_kernel'] = self.video_conv_kernel
        output['video_conv_stride'] = self.video_conv_stride
        output['audio_conv_kernel'] = self.audio_conv_kernel
        output['audio_conv_stride'] = self.audio_conv_stride
        output['hidden_size'] = self.hidden_size
        output["model_type"] = self.__class__.model_type
        return output
    @classmethod
    def from_pretrained(cls, pretrained_model_name_or_path, **kwargs):
        config_dict, kwargs = cls.get_config_dict(pretrained_model_name_or_path, **kwargs)

        clip_config = CLIPConfig.from_dict(config_dict['image_config'])
        whisper_config = WhisperConfig.from_dict(config_dict['audio_config'])
        llm_config = LlamaConfig.from_dict(config_dict['llm_config'])

        return cls(clip_config=clip_config, whisper_config=whisper_config, llm_config=llm_config, **kwargs)

class MM_LLMs(PreTrainedModel):
    def __init__(self, config):
        super().__init__(config)
        self.config = config

        self.temporal_position_embeddings = nn.Embedding(config.n_frames, 
        config.image_config.projection_dim)

        self.image_encoder = CLIPModel(config.image_config)

        self.video_encoder = CLIPModel(config.image_config)

        self.audio_encoder = WhisperModel(config.audio_config)

        self.llm = LlamaForCausalLM(config.llm_config)

        attn_dropout = 0.1
        is_add_bias_kv = True
        is_add_zero_attn = True
        self.temporal_self_attention = nn.MultiheadAttention(config.image_config.projection_dim, 
                                                             config.attention_heads,
                                                             dropout=attn_dropout,
                                                             add_bias_kv=is_add_bias_kv,
                                                             add_zero_attn=is_add_zero_attn)

        self.video_align_attention = nn.MultiheadAttention(config.llm_config.hidden_size, 
                                                             config.attention_heads * 2,
                                                             dropout=attn_dropout,
                                                             add_bias_kv=is_add_bias_kv,
                                                             add_zero_attn=is_add_zero_attn)

        self.audio_align_attention = nn.MultiheadAttention(config.llm_config.hidden_size, 
                                                             config.attention_heads * 2,
                                                             dropout=attn_dropout,
                                                             add_bias_kv=is_add_bias_kv,
                                                             add_zero_attn=is_add_zero_attn)

        self.image_align_attention = nn.MultiheadAttention(config.llm_config.hidden_size, 
                                                             config.attention_heads * 2,
                                                             dropout=attn_dropout,
                                                             add_bias_kv=is_add_bias_kv,
                                                             add_zero_attn=is_add_zero_attn)
        
        self.video_long_self_attention = nn.MultiheadAttention(config.image_config.projection_dim,
                                                            config.attention_heads,
                                                            dropout=attn_dropout,
                                                            add_bias_kv=is_add_bias_kv,
                                                            add_zero_attn=is_add_zero_attn)

        self.transform_video_to_hidden = nn.Linear(config.image_config.projection_dim, 
                                                   config.llm_config.hidden_size)
        self.transform_audio_to_hidden = nn.Linear(config.audio_config.d_model, 
                                                   config.llm_config.hidden_size)
        self.transform_image_to_hidden = nn.Linear(config.image_config.projection_dim, 
                                                   config.llm_config.hidden_size)

        self.project_image = nn.Conv1d(config.image_config.projection_dim, config.image_config.projection_dim, 
        kernel_size=config.image_conv_kernel, stride=config.image_conv_stride)
        self.project_video = nn.Conv1d(config.image_config.projection_dim, config.image_config.projection_dim, 
        kernel_size=config.video_conv_kernel, stride=config.video_conv_stride)
        self.project_audio = nn.Conv1d(config.audio_config.d_model, config.audio_config.d_model, 
        kernel_size=config.audio_conv_kernel, stride=config.audio_conv_stride)

        
        self.logit_scale = nn.Parameter(torch.ones([]) * np.log(1 / 0.07))

        self.layer_norm = nn.LayerNorm(config.image_config.projection_dim)
        self.softmax = nn.Softmax(dim=-1)
        self.relu = nn.ReLU()
        self.gelu = nn.GELU()
        self.elu = nn.ELU()

        self.sigmoid = nn.Sigmoid()

        self.loss_fct = CrossEntropyLoss()

        self.init_weights()

    def forward(self, inputs=None):
        # """
        # :param inputs:
        #             video_frames: (B x F)
        #             audios: B x 1
        #             images: B x 1
        #             input_ids: B x L
        #             labels: B x L
        #
        # :return: loss when training else None
        # """
        text_embeddings, attention_mask, labels = self.prepare_inputs_for_generation(inputs)

        if 'inference' in inputs and inputs['inference'] is True:
            # generate_ids = self.llm.generate(input_ids=inputs['input_ids'], inputs_embeds=text_embeddings, max_new_tokens=128)
            # generate_ids = self.llm.generate(inputs_embeds=text_embeddings, max_new_tokens=128)

            # The code below will possibly trigger an error in : https://github.com/microsoft/DeepSpeed/issues/3156 (the solution only partially resolves the bug for me)
            generate_ids = self.llm.generate(inputs_embeds=text_embeddings, max_new_tokens=128, eos_token_id=2, bos_token_id=1, pad_token_id=32006)
            return generate_ids
        outputs = self.llm(inputs_embeds=text_embeddings, attention_mask=attention_mask, labels=labels)

        return outputs

    def prepare_inputs_for_generation(self, inputs):

        image_features = self.encode_image(inputs['images']) if inputs['images'] is not None else None
        audio_features = self.encode_audio(inputs['audios']) if inputs['audios'] is not None else None
        video_features = self.encode_video_long(inputs['videos']) if inputs['videos'] is not None else None
        # embed_tokens = self.llm.model.model.embed_tokens
        embed_tokens = self.llm.model.embed_tokens
        text_embeddings = embed_tokens(inputs['input_ids'])

        token_embeddings = embed_tokens.weight.unsqueeze(0).repeat(
            text_embeddings.size(0), 1, 1).transpose(0, 1)

        ingore_num = 0
        if video_features is not None:
            video_starts = embed_tokens(inputs['video_starts']).unsqueeze(1)
            video_ends = embed_tokens(inputs['video_ends']).unsqueeze(1)

            video_features = self.project_video(video_features.transpose(1, 2).contiguous()).transpose(1, 2).contiguous()

            video_features = self.transform_video_to_hidden(video_features)
            
            video_features = self.video_align_attention(video_features.transpose(0, 1), 
            token_embeddings, token_embeddings)[0].transpose(0, 1).contiguous()

            video_inputs = torch.cat([torch.cat([video_starts, video_features], dim=1), video_ends], dim=1)

            text_embeddings = torch.cat([torch.cat([text_embeddings[:, 0, :].unsqueeze(1), video_inputs], dim=1), text_embeddings[:, 1:, :]], dim=1)

            ingore_num += (video_inputs.size(1))

        if audio_features is not None:
            audio_starts = embed_tokens(inputs['audio_starts']).unsqueeze(1)
            audio_ends = embed_tokens(inputs['audio_ends']).unsqueeze(1)
            
            audio_features = self.project_audio(audio_features.transpose(1, 2).contiguous()).transpose(1, 2).contiguous()

            audio_features = self.transform_audio_to_hidden(audio_features)
            
            # mean pooling
            # audio_features = torch.sum(audio_features, dim=1) / audio_features.size(1) 
            # audio_features = audio_features.unsqueeze(1)

            audio_features = self.audio_align_attention(audio_features.transpose(0, 1), 
            token_embeddings, token_embeddings)[0].transpose(0, 1).contiguous()

            audio_inputs = torch.cat([torch.cat([audio_starts, audio_features], dim=1), audio_ends], dim=1)

            text_embeddings = torch.cat(
                [torch.cat([text_embeddings[:, 0, :].unsqueeze(1), audio_inputs], dim=1), text_embeddings[:, 1:, :]],
                dim=1)

            ingore_num += (audio_inputs.size(1))

        if image_features is not None:
            image_starts = embed_tokens(inputs['image_starts']).unsqueeze(1)
            image_ends = embed_tokens(inputs['image_ends']).unsqueeze(1)

            image_features = self.project_image(image_features.transpose(1, 2).contiguous()).transpose(1, 2).contiguous()

            image_features = self.transform_image_to_hidden(image_features)
            image_features = self.image_align_attention(image_features.transpose(0, 1),
             token_embeddings, token_embeddings)[0].transpose(0, 1).contiguous()

            image_inputs = torch.cat([torch.cat([image_starts, image_features], dim=1), image_ends], dim=1)

            text_embeddings = torch.cat(
                [torch.cat([text_embeddings[:, 0, :].unsqueeze(1), image_inputs], dim=1), 
                text_embeddings[:, 1:, :]], dim=1)

            ingore_num += (image_inputs.size(1))

        if 'attention_mask' in inputs:
            attention_mask = torch.tensor([1]*ingore_num*text_embeddings.size(0), device=text_embeddings.device).view(text_embeddings.size(0), -1)
            attention_mask = torch.cat([attention_mask, inputs['attention_mask']], dim=1)
        else:
            attention_mask = None

        if 'labels' in inputs and inputs['labels'] is not None:
            labels = torch.tensor([-100]*ingore_num*text_embeddings.size(0), device=text_embeddings.device).view(text_embeddings.size(0), -1)
            labels = torch.cat([labels, inputs['labels']], dim=1)
        else:
            labels = None

        return text_embeddings, attention_mask, labels

    def encode_video(self, videos):
        # simple image encoding without temporal embedding and self attention
        videos = videos.view(-1, videos.size(-3), videos.size(-2), videos.size(-1))
        video_outputs = self.video_encoder.get_image_features(videos)
        video_features = video_outputs
        temporal_pos = torch.tensor(
            [[i for i in range(self.config.n_frames)] 
            for j in range(videos.size(0) // self.config.n_frames)],
            dtype=torch.int, device=video_features.device).view(-1)

        frame_temporal_pos_embed = self.temporal_position_embeddings(temporal_pos)

        video_features = (video_features + frame_temporal_pos_embed).view(videos.size(0) // self.config.n_frames,
                                                                          self.config.n_frames, -1)

        video_features = video_features.transpose(0, 1).contiguous()
        self_attn_video_features = self.temporal_self_attention(video_features, video_features, video_features)[0]

        return self_attn_video_features.transpose(0, 1).contiguous()
    
    def encode_video_long(self, videos):
        # simple image encoding without temporal embedding and self attention
        videos = videos.view(-1, videos.size(-3), videos.size(-2), videos.size(-1))
        video_features = self.video_encoder.visual_projection(self.video_encoder.vision_model(videos)[0])[:, 1:, :]
        video_features = video_features.reshape(videos.size(0) // self.config.n_frames,
                                                                          self.config.n_frames * video_features.size(1), -1).contiguous()
        
        video_features = add_positional_encoding(video_features).transpose(0, 1).contiguous()
        video_self_attn_features = self.video_long_self_attention(video_features, video_features, video_features)[0].transpose(0, 1).contiguous()
        return video_self_attn_features

    def encode_audio(self, audios):
        audio_features = self.audio_encoder.encoder(audios)
        return audio_features[0]

    def encode_image(self, images):
        # vision_outputs = self.image_encoder.get_image_features(images)
        # image_features = vision_outputs  # pooled_output
        # image_features = self.visual_projection(pooled_output)
        # image_features = image_features.unsqueeze(1)


        image_features = self.image_encoder.visual_projection(self.image_encoder.vision_model(images)[0])[:, 1:, :]
        return image_features

def create_positional_encoding(L, h):
    # Create a tensor to store the position encoding
    position_encoding = torch.zeros(L, h)

    # Fill the position encoding tensor
    for pos in range(L):
        for i in range(0, h, 2):
            div_term = torch.exp(torch.tensor(-(math.log(10000.0) / h * (2 * i))))
            position_encoding[pos, i] = torch.sin(pos * div_term)
            position_encoding[pos, i + 1] = torch.cos(pos * div_term)

    return position_encoding

def add_positional_encoding(tensor):
    N, L, h = tensor.size()  # batch size, sequence length, and feature dimension

    # Create position embedding tensor
    position_embedding = create_positional_encoding(L, h).to(tensor.device).to(tensor.dtype)

    # Expand position embedding to match input tensor dimensions
    position_embedding = position_embedding.unsqueeze(0).expand(N, -1, -1)

    # Add position embedding to the input tensor
    return tensor + position_embedding