modeling_gpt_bert.py

from __future__ import annotations

import math

import torch
import torch.nn as nn
import torch.nn.functional as F
from torch import _softmax_backward_data as _softmax_backward_data
from .configuration_gpt_bert import ModelConfig

from transformers.modeling_utils import PreTrainedModel
from transformers.modeling_outputs import (
    BaseModelOutput,
    CausalLMOutput
)

from typing import Optional, Union


class Layer(nn.Module):

    def __init__(self: Layer, config: ModelConfig, layer_idx: int = 0):
        super().__init__()
        self.attention = Attention(config)
        self.mlp = FeedForward(config)

        self.mlp.mlp[1].weight.data *= math.sqrt(1.0 / (2.0 * (1 + layer_idx)))
        self.mlp.mlp[-2].weight.data *= math.sqrt(1.0 / (2.0 * (1 + layer_idx)))

    def forward(self: Layer, x: torch.Tensor, attention_mask: torch.Tensor, relative_embedding: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:
        attention: torch.Tensor
        attention_probs: torch.Tensor
        attention, attention_probs = self.attention(x, attention_mask, relative_embedding)
        x += attention
        x += self.mlp(x)

        return x, attention_probs


class MaskClassifier(nn.Module):

    def __init__(self: MaskClassifier, config: ModelConfig, subword_embedding: nn.Parameter):
        super().__init__()
        self.nonlinearity = nn.Sequential(
            nn.LayerNorm(config.hidden_size, config.layer_norm_eps, elementwise_affine=False),
            nn.Linear(config.hidden_size, config.hidden_size),
            nn.GELU(),
            nn.LayerNorm(config.hidden_size, config.layer_norm_eps, elementwise_affine=False),
            nn.Dropout(config.hidden_dropout_prob),
            nn.Linear(subword_embedding.size(1), subword_embedding.size(0))
        )
        self.initialize(config.hidden_size, subword_embedding)

    def initialize(self: MaskClassifier, hidden_size: int, embedding: nn.Parameter):
        std: float = math.sqrt(2.0 / (5.0 * hidden_size))
        nn.init.trunc_normal_(self.nonlinearity[1].weight, mean=0.0, std=std, a=-2*std, b=2*std)
        self.nonlinearity[-1].weight = embedding
        self.nonlinearity[1].bias.data.zero_()
        self.nonlinearity[-1].bias.data.zero_()

    def forward(self: MaskClassifier, x: torch.Tensor, masked_lm_labels: torch.Tensor | None = None) -> torch.Tensor:
        if masked_lm_labels is not None:
            x = torch.index_select(x.flatten(0, 1), 0, torch.nonzero(masked_lm_labels.flatten() != -100).squeeze())
        x = self.nonlinearity(x)

        return x


class GeGLU(nn.Module):
    def forward(self: GeGLU, x: torch.Tensor) -> torch.Tensor:
        gate: torch.Tensor
        x, gate = x.chunk(2, dim=-1)
        x = x * F.gelu(gate, approximate='tanh')
        return x


class FeedForward(nn.Module):
    def __init__(self: FeedForward, config: ModelConfig) -> None:
        super().__init__()
        self.mlp = nn.Sequential(
            nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps, elementwise_affine=False),
            nn.Linear(config.hidden_size, 2*config.intermediate_size, bias=False),
            GeGLU(),
            nn.LayerNorm(config.intermediate_size, eps=config.layer_norm_eps, elementwise_affine=False),
            nn.Linear(config.intermediate_size, config.hidden_size, bias=False),
            nn.Dropout(config.hidden_dropout_prob)
        )
        self.initialize(config.hidden_size)

    def initialize(self: FeedForward, hidden_size: int) -> None:
        std: float = math.sqrt(2.0 / (5.0 * hidden_size))
        nn.init.trunc_normal_(self.mlp[1].weight, mean=0.0, std=std, a=-2*std, b=2*std)
        nn.init.trunc_normal_(self.mlp[-2].weight, mean=0.0, std=std, a=-2*std, b=2*std)

    def forward(self: FeedForward, x: torch.Tensor) -> torch.Tensor:
        return self.mlp(x)


class MaskedSoftmax(torch.autograd.Function):
    @staticmethod
    def forward(self: MaskedSoftmax, x: torch.Tensor, mask: torch.Tensor, dim: int) -> torch.Tensor:
        self.dim = dim
        x.masked_fill_(mask, float('-inf'))
        x = torch.softmax(x, self.dim)
        x.masked_fill_(mask, 0.0)
        self.save_for_backward(x)
        return x

    @staticmethod
    def backward(self: MaskedSoftmax, grad_output: torch.Tensor) -> tuple[torch.Tensor, None, None]:
        output: torch.Tensor
        output, = self.saved_tensors
        inputGrad: torch.Tensor = _softmax_backward_data(grad_output, output, self.dim, output.dtype)
        return inputGrad, None, None


class Attention(nn.Module):
    def __init__(self: Attention, config: ModelConfig) -> None:
        super().__init__()

        self.config: ModelConfig = config

        if config.hidden_size % config.num_attention_heads != 0:
            raise ValueError(f"The hidden size {config.hidden_size} is not a multiple of the number of attention heads {config.num_attention_heads}")

        self.hidden_size: int = config.hidden_size
        self.num_heads: int = config.num_attention_heads
        self.head_size: int = config.hidden_size // config.num_attention_heads

        self.in_proj_qk = nn.Linear(config.hidden_size, 2*config.hidden_size, bias=True)
        self.in_proj_vg = nn.Linear(config.hidden_size, 2*config.hidden_size, bias=True)
        self.out_proj = nn.Linear(config.hidden_size, config.hidden_size, bias=True)

        self.pre_layer_norm = nn.LayerNorm(config.hidden_size, config.layer_norm_eps, elementwise_affine=False)
        self.post_layer_norm = nn.LayerNorm(config.hidden_size, config.layer_norm_eps, elementwise_affine=False)

        position_indices: torch.Tensor = torch.arange(config.max_position_embeddings, dtype=torch.long).unsqueeze(1) \
            - torch.arange(config.max_position_embeddings, dtype=torch.long).unsqueeze(0)
        position_indices: torch.Tensor = self.make_log_bucket_position(position_indices, config.position_bucket_size, config.max_position_embeddings)
        position_indices = config.position_bucket_size - 1 + position_indices
        self.register_buffer("position_indices", position_indices, persistent=False)

        self.dropout = nn.Dropout(config.attention_probs_dropout_prob)
        self.scale: float = 1.0 / math.sqrt(3 * self.head_size)
        self.initialize()

    def make_log_bucket_position(self: Attention, relative_pos: torch.Tensor, bucket_size: int, max_position: int) -> torch.Tensor:
        sign: torch.Tensor = torch.sign(relative_pos)
        mid: int = bucket_size // 2
        abs_pos: torch.Tensor = torch.where((relative_pos < mid) & (relative_pos > -mid), mid - 1, torch.abs(relative_pos).clamp(max=max_position - 1))
        log_pos: torch.Tensor = torch.ceil(torch.log(abs_pos / mid) / math.log((max_position-1) / mid) * (mid - 1)).int() + mid
        bucket_pos: torch.Tensor = torch.where(abs_pos <= mid, relative_pos, log_pos * sign).long()
        return bucket_pos

    def initialize(self: Attention) -> None:
        std: float = math.sqrt(2.0 / (5.0 * self.hidden_size))
        nn.init.trunc_normal_(self.in_proj_qk.weight, mean=0.0, std=std, a=-2*std, b=2*std)
        nn.init.trunc_normal_(self.in_proj_vg.weight, mean=0.0, std=std, a=-2*std, b=2*std)
        nn.init.trunc_normal_(self.out_proj.weight, mean=0.0, std=std, a=-2*std, b=2*std)
        self.in_proj_qk.bias.data.zero_()
        self.in_proj_vg.bias.data.zero_()
        self.out_proj.bias.data.zero_()

    def _create_position_tensors(self: Attention, relative_embedding: torch.Tensor, query_len: int, key_len: int) -> tuple[torch.Tensor, torch.Tensor]:
        pos = self.in_proj_qk(self.dropout(relative_embedding))  # shape: [2T-1, 2D]
        pos = F.embedding(self.position_indices[:query_len, :key_len], pos)  # shape: [T, T, 2D]
        query_pos, key_pos = pos.chunk(2, dim=-1)
        query_pos = query_pos.view(query_len, key_len, self.num_heads, self.head_size)
        key_pos = key_pos.view(query_len, key_len, self.num_heads, self.head_size)

        return query_pos, key_pos

    def attention_operation(self: Attention, query: torch.Tensor, key: torch.Tensor, value: torch.Tensor, attention_mask: torch.Tensor, query_pos: torch.Tensor, key_pos: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:
        key_len: int
        batch_size: int
        key_len, batch_size, _ = key.size()
        query_len: int
        query_len, _, _ = query.size()

        query = query.reshape(query_len, batch_size * self.num_heads, self.head_size).transpose(0, 1)
        key = key.reshape(key_len, batch_size * self.num_heads, self.head_size).transpose(0, 1)
        value = value.reshape(key_len, batch_size * self.num_heads, self.head_size).transpose(0, 1)

        attention_probs: torch.Tensor = torch.bmm(query, key.transpose(1, 2) * self.scale)

        query = query.view(batch_size, self.num_heads, query_len, self.head_size)
        key = key.view(batch_size, self.num_heads, query_len, self.head_size)
        attention_probs = attention_probs.view(batch_size, self.num_heads, query_len, key_len)
        attention_probs.add_(torch.einsum("bhqd,qkhd->bhqk", query, key_pos * self.scale))
        attention_probs.add_(torch.einsum("bhkd,qkhd->bhqk", key * self.scale, query_pos))

        attention_probs = MaskedSoftmax.apply(attention_probs, attention_mask, -1)

        attention_probs = self.dropout(attention_probs)
        attention_output: torch.Tensor = torch.bmm(attention_probs.flatten(0, 1), value)  # shape: [B*H, Q, D]
        attention_output = attention_output.transpose(0, 1).reshape(query_len, batch_size, self.hidden_size)  # shape: [Q, B, H*D]

        return attention_output, attention_probs

    def forward(self: Attention, hidden_states: torch.Tensor, attention_mask: torch.Tensor, relative_embedding: torch.Tensor) -> tuple[torch.Tensor, torch.Tensor]:
        key_len: int
        batch_size: int
        key_len, batch_size, _ = hidden_states.size()
        query_len: int = key_len

        if self.position_indices.size(0) < query_len:
            position_indices = torch.arange(query_len, dtype=torch.long).unsqueeze(1) \
                - torch.arange(query_len, dtype=torch.long).unsqueeze(0)
            position_indices = self.make_log_bucket_position(position_indices, self.config.position_bucket_size, 512)
            position_indices = self.config.position_bucket_size - 1 + position_indices
            self.register_buffer("position_indices", position_indices.to(hidden_states.device), persistent=True)

        hidden_states = self.pre_layer_norm(hidden_states)
        query, key = self.in_proj_qk(hidden_states).chunk(2, dim=2)  # shape: [T, B, D]
        value, gate = self.in_proj_vg(hidden_states).chunk(2, dim=2)  # shape: [T, B, D]
        gate = F.gelu(gate)

        query_pos: torch.Tensor
        key_pos: torch.Tensor
        query_pos, key_pos = self._create_position_tensors(relative_embedding, query_len, key_len)

        attention_output: torch.Tensor
        attention_probs: torch.Tensor
        attention_output, attention_probs = self.attention_operation(query, key, value, attention_mask, query_pos, key_pos)
        attention_output = attention_output * gate
        attention_output = self.post_layer_norm(attention_output)
        attention_output = self.out_proj(attention_output)
        attention_output = self.dropout(attention_output)

        return attention_output, attention_probs


class Embedding(nn.Module):
    def __init__(self: Embedding, config: ModelConfig):
        super().__init__()
        self.hidden_size: int = config.hidden_size

        self.word_embedding = nn.Embedding(config.vocab_size, config.hidden_size)
        self.word_layer_norm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps, elementwise_affine=False)
        self.dropout = nn.Dropout(config.hidden_dropout_prob)

        self.relative_embedding = nn.Parameter(torch.empty(2 * config.position_bucket_size - 1, config.hidden_size))
        self.relative_layer_norm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps)

        self.initialize()

    def initialize(self: Embedding):
        std: float = math.sqrt(2.0 / (5.0 * self.hidden_size))
        nn.init.trunc_normal_(self.relative_embedding, mean=0.0, std=std, a=-2*std, b=2*std)
        nn.init.trunc_normal_(self.word_embedding.weight, mean=0.0, std=std, a=-2*std, b=2*std)

    def forward(self: Embedding, input_ids: torch.Tensor):
        word_embedding: torch.Tensor = self.dropout(self.word_layer_norm(self.word_embedding(input_ids)))
        relative_embeddings: torch.Tensor = self.relative_layer_norm(self.relative_embedding)
        return word_embedding, relative_embeddings


class GPTBERTPreTrainedModel(PreTrainedModel):
    config_class = ModelConfig
    supports_gradient_checkpointing = False

    def _set_gradient_checkpointing(self, module, value=False):
        raise NotImplementedError("Gradient checkpointing is not supported by this model")

    def _init_weights(self, module):
        std = math.sqrt(2.0 / (5.0 * self.hidden_size))

        if isinstance(module, nn.Linear):
            nn.init.trunc_normal_(module.weight.data, mean=0.0, std=std, a=-2*std, b=2*std)
            if module.bias is not None:
                module.bias.data.zero_()
        elif isinstance(module, nn.Embedding):
            nn.init.trunc_normal_(module.weight.data, mean=0.0, std=std, a=-2*std, b=2*std)
        elif isinstance(module, nn.LayerNorm):
            module.bias.data.zero_()
            module.weight.data.fill_(1.0)


class GPTBERT(GPTBERTPreTrainedModel):

    def __init__(self, config: ModelConfig, is_causal: bool, **kwargs):
        super().__init__(config, **kwargs)
        self.config = config
        self.hidden_size = config.hidden_size

        self.embedding = Embedding(config)
        self.layers = nn.ModuleList([Layer(config) for _ in range(config.num_layers)])
        self.is_causal = is_causal

    def get_input_embeddings(self):
        return self.embedding.word_embedding

    def set_input_embeddings(self, value):
        self.embedding.word_embedding = value

    def get_contextualized_embeddings(self, input_ids: torch.Tensor, attention_mask: Optional[torch.Tensor] = None) -> list[torch.Tensor]:
        """
        """
        input_shape = input_ids.size()

        batch_size, seq_length = input_shape

        if attention_mask is None:
            attention_mask = input_ids.new_zeros((batch_size, seq_length), dtype=torch.bool).unsqueeze(1).unsqueeze(2)
        else:
            attention_mask = ~attention_mask.bool()

        if len(attention_mask.size()) == 2:
            attention_mask = attention_mask.unsqueeze(1).unsqueeze(2)
        elif len(attention_mask.size()) == 3:
            attention_mask = attention_mask.unsqueeze(1)

        if self.is_causal:
            attention_mask = attention_mask | input_ids.new_ones((seq_length, seq_length), dtype=torch.bool).triu(1).unsqueeze(0).unsqueeze(0)

        static_embeddings, relative_embeddings = self.embedding(input_ids.t())
        contextualized_embeddings = [static_embeddings]
        attention_probs = []
        for layer in self.layers:
            layer_embeddings, layer_attention_probs = layer(contextualized_embeddings[-1], attention_mask, relative_embeddings)
            contextualized_embeddings.append(layer_embeddings)
            attention_probs.append(layer_attention_probs)
        contextualized_embeddings = [emb.transpose(0, 1) for emb in contextualized_embeddings]
        last_layer = contextualized_embeddings[-1]
        return last_layer, contextualized_embeddings, attention_probs

    def forward(
        self,
        input_ids: torch.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
        token_type_ids: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.Tensor] = None,
        output_hidden_states: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        **kwargs
    ) -> Union[tuple[torch.Tensor], BaseModelOutput]:
        """
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        sequence_output, contextualized_embeddings, attention_probs = self.get_contextualized_embeddings(input_ids, attention_mask)

        if not return_dict:
            return (
                sequence_output,
                *([contextualized_embeddings] if output_hidden_states else []),
                *([attention_probs] if output_attentions else [])
            )

        return BaseModelOutput(
            last_hidden_state=sequence_output,
            hidden_states=contextualized_embeddings if output_hidden_states else None,
            attentions=attention_probs if output_attentions else None
        )

# To do Masked Language Modeling instead, you can replace MyModelForCausalLM by MyModelForMaskedLM
# and change the output type from CausalLMOutput to MaskedLMOutput.


class GPTBERTForCausalLM(GPTBERTPreTrainedModel):
    _keys_to_ignore_on_load_unexpected = ["head"]

    def __init__(self, config, **kwargs):
        super().__init__(config, **kwargs)
        self.model = GPTBERT(config, is_causal=True, **kwargs)
        self.vocab_size = config.vocab_size
        self.lm_head = MaskClassifier(config, self.model.embedding.word_embedding.weight)
        self.hidden_size = config.hidden_size

    def get_output_embeddings(self):
        return self.lm_head.nonlinearity[-1].weight

    def set_output_embeddings(self, new_embeddings):
        self.lm_head.nonlinearity[-1].weight = new_embeddings

    def get_input_embeddings(self):
        return self.model.embedding.word_embedding

    def set_input_embeddings(self, value):
        self.model.embedding.word_embedding = value

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

    def get_decoder(self):
        return self.model

    def can_generate(self):
        return True

    def forward(
        self,
        input_ids: torch.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
        token_type_ids: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.Tensor] = None,
        output_hidden_states: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        labels: Optional[torch.LongTensor] = None,
        **kwargs
    ) -> Union[tuple, CausalLMOutput]:

        sequence_output, contextualized_embeddings, attention_probs = self.model.get_contextualized_embeddings(input_ids, attention_mask)
        subword_prediction = self.lm_head(sequence_output)

        loss = None
        if labels is not None:
            gold_labels = labels.flatten()
            gold_labels = gold_labels[gold_labels != -100]

            loss = F.cross_entropy(subword_prediction, gold_labels)

        if not return_dict:
            output = (
                subword_prediction,
                *([contextualized_embeddings] if output_hidden_states else []),
                *([attention_probs] if output_attentions else [])
            )
            return ((loss,) + output) if loss is not None else output

        return CausalLMOutput(
            loss=loss,
            logits=subword_prediction,
            hidden_states=contextualized_embeddings if output_hidden_states else None,
            attentions=attention_probs if output_attentions else None
        )

    def prepare_inputs_for_generation(
        self,
        input_ids: torch.Tensor,
        past_key_values: Optional[torch.Tensor] = None,
        attention_mask: Optional[torch.Tensor] = None,
        inputs_embeds: Optional[torch.Tensor] = None,
        cache_position: Optional[torch.LongTensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        use_cache: bool = True,
        num_logits_to_keep: Optional[int] = 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}
        else:
            model_inputs = {"input_ids": input_ids.contiguous()}  # `contiguous()` needed for compilation use cases

        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


class GPTBERTForMaskedLM(GPTBERTPreTrainedModel):
    _keys_to_ignore_on_load_unexpected = ["head"]

    def __init__(self, config, **kwargs):
        super().__init__(config, **kwargs)
        self.model = GPTBERT(config, is_causal=False, **kwargs)
        self.vocab_size = config.vocab_size
        self.lm_head = MaskClassifier(config, self.model.embedding.word_embedding.weight)
        self.hidden_size = config.hidden_size

    def get_output_embeddings(self):
        return self.lm_head.nonlinearity[-1].weight

    def set_output_embeddings(self, new_embeddings):
        self.lm_head.nonlinearity[-1].weight = new_embeddings

    def get_input_embeddings(self):
        return self.model.embedding.word_embedding

    def set_input_embeddings(self, value):
        self.model.embedding.word_embedding = value

    def set_encoder(self, encoder):
        self.model = encoder

    def get_encoder(self):
        return self.model

    def can_generate(self):
        return True

    def forward(
        self,
        input_ids: torch.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
        token_type_ids: Optional[torch.Tensor] = None,
        position_ids: Optional[torch.Tensor] = None,
        output_hidden_states: Optional[bool] = None,
        output_attentions: Optional[bool] = None,
        return_dict: Optional[bool] = None,
        labels: Optional[torch.LongTensor] = None,
        **kwargs
    ) -> Union[tuple, CausalLMOutput]:

        sequence_output, contextualized_embeddings, attention_probs = self.model.get_contextualized_embeddings(input_ids, attention_mask)
        subword_prediction = self.lm_head(sequence_output)

        loss = None
        if labels is not None:
            gold_labels = labels.flatten()
            gold_labels = gold_labels[gold_labels != -100]

            loss = F.cross_entropy(subword_prediction, gold_labels)

        if not return_dict:
            output = (
                subword_prediction,
                *([contextualized_embeddings] if output_hidden_states else []),
                *([attention_probs] if output_attentions else [])
            )
            return ((loss,) + output) if loss is not None else output

        return CausalLMOutput(
            loss=loss,
            logits=subword_prediction,
            hidden_states=contextualized_embeddings if output_hidden_states else None,
            attentions=attention_probs if output_attentions else None
        )