decoder_only_t5/modeling.py

import copy
from typing import Optional, Tuple, Union

import torch
from torch import nn
from torch.nn import CrossEntropyLoss
from transformers.models.t5 import modeling_t5
from transformers.modeling_outputs import CausalLMOutputWithPast
from transformers.utils import (
    add_start_docstrings_to_model_forward,
    logging,
    replace_return_docstrings,
)

from decoder_only_t5.config import DecoderOnlyT5Config


logger = logging.get_logger(__name__)
_CONFIG_FOR_DOC = "DecoderOnlyT5Config"


class DecoderOnlyT5LayerFF(modeling_t5.T5LayerFF):
    def __init__(self, config: DecoderOnlyT5Config):
        super(modeling_t5.T5LayerFF, self).__init__()
        if config.is_gated_act:
            self.DenseReluDense = modeling_t5.T5DenseGatedActDense(config)
        else:
            self.DenseReluDense = modeling_t5.T5DenseActDense(config)

        if not config.parallel_layers:
            self.layer_norm = modeling_t5.T5LayerNorm(
                config.d_model, eps=config.layer_norm_epsilon
            )
        else:
            self.layer_norm = nn.Identity()
        self.dropout = nn.Dropout(config.dropout_rate)


# LlamaRotaryEmbedding
class T5DecoderOnlyRotaryEmbedding(nn.Module):
    def __init__(self, dim, max_position_embeddings=2048, base=10000, device=None):
        super().__init__()

        self.dim = dim
        self.max_position_embeddings = max_position_embeddings
        self.base = base
        inv_freq = 1.0 / (
            self.base ** (torch.arange(0, self.dim, 2).float().to(device) / self.dim)
        )
        self.register_buffer("inv_freq", inv_freq, persistent=False)

        # Build here to make `torch.jit.trace` work.
        self._set_cos_sin_cache(
            seq_len=max_position_embeddings,
            device=self.inv_freq.device,
            dtype=torch.get_default_dtype(),
        )

    def _set_cos_sin_cache(self, seq_len, device, dtype):
        self.max_seq_len_cached = seq_len
        t = torch.arange(
            self.max_seq_len_cached, device=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().to(dtype), persistent=False)
        self.register_buffer("sin_cached", emb.sin().to(dtype), persistent=False)

    def forward(self, x, seq_len=None):
        # x: [bs, num_attention_heads, seq_len, head_size]
        if seq_len > self.max_seq_len_cached:
            self._set_cos_sin_cache(seq_len=seq_len, device=x.device, dtype=x.dtype)

        return (
            self.cos_cached[:seq_len].to(dtype=x.dtype),
            self.sin_cached[:seq_len].to(dtype=x.dtype),
        )


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, unsqueeze_dim=1):
    """Applies Rotary Position Embedding to the query and key tensors.

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


# https://github.com/huggingface/transformers/blob/7ee995fd9c692761c4601ddbffa2ac2ec9f27b0b/src/transformers/models/llama/modeling_llama.py#L263
def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
    """
    This is the equivalent of torch.repeat_interleave(x, dim=1, repeats=n_rep). The hidden states go from (batch,
    num_key_value_heads, seqlen, head_dim) to (batch, num_attention_heads, seqlen, head_dim)
    """
    batch, num_key_value_heads, slen, head_dim = hidden_states.shape
    if n_rep == 1:
        return hidden_states
    hidden_states = hidden_states[:, :, None, :, :].expand(
        batch, num_key_value_heads, n_rep, slen, head_dim
    )
    return hidden_states.reshape(batch, num_key_value_heads * n_rep, slen, head_dim)


class DecoderOnlyT5Attention(modeling_t5.T5Attention):
    """
    Supports both multi-head and multi-query attention.
    https://arxiv.org/abs/1911.02150
    https://github.com/google/flaxformer/blob/ea17eb012a1d340ddff017b7a534c2162aaec34c/flaxformer/components/attention/dense_attention.py#L292
    """

    def __init__(self, config: DecoderOnlyT5Config, has_relative_attention_bias=False):
        super(modeling_t5.T5Attention, self).__init__()
        self.is_decoder = config.is_decoder
        self.has_relative_attention_bias = has_relative_attention_bias
        self.relative_attention_num_buckets = config.relative_attention_num_buckets
        self.relative_attention_max_distance = config.relative_attention_max_distance
        self.d_model = config.d_model
        self.key_value_proj_dim = config.d_kv
        self.n_heads = config.num_heads
        self.n_kv_heads = 1 if config.multi_query_attention else self.n_heads
        self.n_kv_groups = self.n_heads // self.n_kv_heads
        self.dropout = config.dropout_rate
        self.inner_dim = self.n_heads * self.key_value_proj_dim
        self.kv_inner_dim = self.n_kv_heads * self.key_value_proj_dim
        if config.use_rotary_embedding:
            self.rotary_embedding = T5DecoderOnlyRotaryEmbedding(
                self.key_value_proj_dim,
                max_position_embeddings=config.relative_attention_max_distance,
                base=config.rotary_embedding_max_timescale,
            )
        else:
            self.rotary_embedding = None

        # Mesh TensorFlow initialization to avoid scaling before softmax
        self.q = nn.Linear(self.d_model, self.inner_dim, bias=False)
        self.k = nn.Linear(self.d_model, self.kv_inner_dim, bias=False)
        self.v = nn.Linear(self.d_model, self.kv_inner_dim, bias=False)
        self.o = nn.Linear(self.inner_dim, self.d_model, bias=False)

        if self.has_relative_attention_bias:
            self.relative_attention_bias = nn.Embedding(
                self.relative_attention_num_buckets, self.n_heads
            )
        self.pruned_heads = set()
        self.gradient_checkpointing = False

    def forward(
        self,
        hidden_states,
        mask=None,
        key_value_states=None,
        position_bias=None,
        position_ids=None,
        past_key_value=None,
        layer_head_mask=None,
        query_length=None,
        use_cache=False,
        output_attentions=False,
    ):
        """
        Self-attention (if key_value_states is None) or attention over source sentence (provided by key_value_states).
        """
        # Input is (batch_size, seq_length, dim)
        # Mask is (batch_size, key_length) (non-causal) or (batch_size, key_length, key_length)
        # past_key_value[0] is (batch_size, n_kv_heads, q_len - 1, dim_per_head)
        batch_size, seq_length = hidden_states.shape[:2]

        real_seq_length = seq_length

        if past_key_value is not None:
            if len(past_key_value) != 2:
                raise ValueError(
                    f"past_key_value should have 2 past states: keys and values. Got { len(past_key_value)} past states"
                )
            real_seq_length += (
                past_key_value[0].shape[2] if query_length is None else query_length
            )

        key_length = (
            real_seq_length if key_value_states is None else key_value_states.shape[1]
        )

        def shape(states, n_heads):
            """projection"""
            return states.view(
                batch_size, -1, n_heads, self.key_value_proj_dim
            ).transpose(1, 2)

        def unshape(states):
            """reshape"""
            return (
                states.transpose(1, 2).contiguous().view(batch_size, -1, self.inner_dim)
            )

        def project(hidden_states, proj_layer, key_value_states, past_key_value):
            """projects hidden states correctly to key/query states"""
            if key_value_states is None:
                # self-attn
                # (batch_size, n_kv_heads, seq_length, dim_per_head)
                hidden_states = shape(proj_layer(hidden_states), self.n_kv_heads)
            elif past_key_value is None:
                # cross-attn
                # (batch_size, n_kv_heads, seq_length, dim_per_head)
                hidden_states = shape(proj_layer(key_value_states), self.n_kv_heads)
            return hidden_states

        def concat_past_key_value(hidden_states, past_key_value, key_value_states):
            if key_value_states is None:
                # self-attn
                # (batch_size, n_kv_heads, key_length, dim_per_head)
                hidden_states = torch.cat([past_key_value, hidden_states], dim=2)
            elif past_key_value.shape[2] != key_value_states.shape[1]:
                # checking that the `sequence_length` of the `past_key_value` is the same as
                # the provided `key_value_states` to support prefix tuning
                # cross-attn
                # (batch_size, n_kv_heads, seq_length, dim_per_head)
                raise NotImplementedError(
                    "cross attention with RoPE and past KV is not implemented"
                )
                # hidden_states = shape(proj_layer(key_value_states), self.n_kv_heads)
            else:
                # cross-attn
                hidden_states = past_key_value
            return hidden_states

        # get query states
        query_states = shape(
            self.q(hidden_states), self.n_heads
        )  # (batch_size, n_heads, seq_length, dim_per_head)

        # get key/value states
        key_states = project(hidden_states, self.k, key_value_states, past_key_value)
        value_states = project(hidden_states, self.v, key_value_states, past_key_value)

        # RoPE
        if self.rotary_embedding is not None:
            kv_seq_len = key_states.shape[-2]
            if past_key_value:
                kv_seq_len += past_key_value[0].shape[-2]
            cos, sin = self.rotary_embedding(query_states, seq_len=kv_seq_len)
            query_states, key_states = apply_rotary_pos_emb(
                query_states, key_states, cos, sin, position_ids
            )

        # concat past
        if past_key_value is not None:
            key_states = concat_past_key_value(
                key_states,
                past_key_value[0],
                key_value_states,
            )
            value_states = concat_past_key_value(
                value_states,
                past_key_value[1],
                key_value_states,
            )

        # MultiQueryDotProductAttention
        key_states = repeat_kv(key_states, self.n_kv_groups)
        value_states = repeat_kv(value_states, self.n_kv_groups)

        # compute scores
        scores = torch.matmul(
            query_states, key_states.transpose(3, 2)
        )  # equivalent of torch.einsum("bnqd,bnkd->bnqk", query_states, key_states), compatible with onnx op>9

        if position_bias is None:
            if not self.has_relative_attention_bias:
                position_bias = torch.zeros(
                    (1, self.n_heads, real_seq_length, key_length),
                    device=scores.device,
                    dtype=scores.dtype,
                )
                if self.gradient_checkpointing and self.training:
                    position_bias.requires_grad = True
            else:
                position_bias = self.compute_bias(
                    real_seq_length, key_length, device=scores.device
                )

            # if key and values are already calculated
            # we want only the last query position bias
            if past_key_value is not None:
                position_bias = position_bias[:, :, -hidden_states.size(1) :, :]

            if mask is not None:
                position_bias = (
                    position_bias + mask
                )  # (batch_size, n_heads, seq_length, key_length)

        if self.pruned_heads:
            mask = torch.ones(position_bias.shape[1])
            mask[list(self.pruned_heads)] = 0
            position_bias_masked = position_bias[:, mask.bool()]
        else:
            position_bias_masked = position_bias

        scores += position_bias_masked
        attn_weights = nn.functional.softmax(scores.float(), dim=-1).type_as(
            scores
        )  # (batch_size, n_heads, seq_length, key_length)
        attn_weights = nn.functional.dropout(
            attn_weights, p=self.dropout, training=self.training
        )  # (batch_size, n_heads, seq_length, key_length)

        # Mask heads if we want to
        if layer_head_mask is not None:
            attn_weights = attn_weights * layer_head_mask

        attn_output = unshape(
            torch.matmul(attn_weights, value_states)
        )  # (batch_size, seq_length, dim)
        attn_output = self.o(attn_output)

        present_key_value_state = (
            (key_states, value_states) if (self.is_decoder and use_cache) else None
        )
        outputs = (attn_output,) + (present_key_value_state,) + (position_bias,)

        if output_attentions:
            outputs = outputs + (attn_weights,)
        return outputs


class DecoderOnlyT5LayerSelfAttention(modeling_t5.T5LayerSelfAttention):
    def __init__(self, config, has_relative_attention_bias=False):
        super(modeling_t5.T5LayerSelfAttention, self).__init__()
        self.SelfAttention = DecoderOnlyT5Attention(
            config, has_relative_attention_bias=has_relative_attention_bias
        )
        self.layer_norm = modeling_t5.T5LayerNorm(
            config.d_model, eps=config.layer_norm_epsilon
        )
        self.dropout = nn.Dropout(config.dropout_rate)
        self.parallel_layers = config.parallel_layers

    def forward(
        self,
        hidden_states,
        attention_mask=None,
        position_bias=None,
        position_ids=None,
        layer_head_mask=None,
        past_key_value=None,
        use_cache=False,
        output_attentions=False,
    ):
        if not self.parallel_layers:
            x = self.layer_norm(hidden_states)
        else:
            x = hidden_states
        attention_output = self.SelfAttention(
            x,
            mask=attention_mask,
            position_bias=position_bias,
            position_ids=position_ids,
            layer_head_mask=layer_head_mask,
            past_key_value=past_key_value,
            use_cache=use_cache,
            output_attentions=output_attentions,
        )
        if not self.parallel_layers:
            # When parallel_layers is True, the residual connection is applied
            # in the decoder block instead of here.
            hidden_states = hidden_states + self.dropout(attention_output[0])
        else:
            hidden_states = attention_output[0]
        outputs = (hidden_states,) + attention_output[
            1:
        ]  # add attentions if we output them
        return outputs


class DecoderOnlyT5Block(modeling_t5.T5Block):
    def __init__(self, config, has_relative_attention_bias=False):
        super(modeling_t5.T5Block, self).__init__()
        self.is_decoder = config.is_decoder
        self.is_decoder_only = config.is_decoder_only
        self.layer = nn.ModuleList()
        self.layer.append(
            DecoderOnlyT5LayerSelfAttention(
                config, has_relative_attention_bias=has_relative_attention_bias
            )
        )
        if self.is_decoder:
            if config.is_decoder_only:
                self.layer.append(nn.Identity())
            else:
                self.layer.append(modeling_t5.T5LayerCrossAttention(config))
        self.parallel_layers = config.parallel_layers
        self.layer.append(DecoderOnlyT5LayerFF(config))

    def forward(
        self,
        hidden_states,
        attention_mask=None,
        position_bias=None,
        position_ids=None,
        encoder_hidden_states=None,
        encoder_attention_mask=None,
        encoder_decoder_position_bias=None,
        layer_head_mask=None,
        cross_attn_layer_head_mask=None,
        past_key_value=None,
        use_cache=False,
        output_attentions=False,
        return_dict=True,
    ):
        if past_key_value is not None:
            if not self.is_decoder:
                logger.warning(
                    "`past_key_values` is passed to the encoder. Please make sure this is intended."
                )
            expected_num_past_key_values = 2 if encoder_hidden_states is None else 4

            if len(past_key_value) != expected_num_past_key_values:
                raise ValueError(
                    f"There should be {expected_num_past_key_values} past states. "
                    f"{'2 (past / key) for cross attention. ' if expected_num_past_key_values == 4 else ''}"
                    f"Got {len(past_key_value)} past key / value states"
                )

            self_attn_past_key_value = past_key_value[:2]
            cross_attn_past_key_value = past_key_value[2:]
        else:
            self_attn_past_key_value, cross_attn_past_key_value = None, None

        ff_layer = self.layer[-1]
        if self.parallel_layers:
            # https://github.com/google/flaxformer/blob/ea17eb012a1d340ddff017b7a534c2162aaec34c/flaxformer/architectures/t5/t5_architecture.py#L563-L568
            x = self.layer[0].layer_norm(hidden_states)
            ff_output = ff_layer(x)
        else:
            x = hidden_states

        self_attention_outputs = self.layer[0](
            x,
            attention_mask=attention_mask,
            position_bias=position_bias,
            position_ids=position_ids,
            layer_head_mask=layer_head_mask,
            past_key_value=self_attn_past_key_value,
            use_cache=use_cache,
            output_attentions=output_attentions,
        )
        x, present_key_value_state = self_attention_outputs[:2]
        attention_outputs = self_attention_outputs[
            2:
        ]  # Keep self-attention outputs and relative position weights

        # clamp inf values to enable fp16 training
        if x.dtype == torch.float16:
            clamp_value = torch.where(
                torch.isinf(x).any(),
                torch.finfo(x.dtype).max - 1000,
                torch.finfo(x.dtype).max,
            )
            x = torch.clamp(x, min=-clamp_value, max=clamp_value)

        do_cross_attention = (
            self.is_decoder
            and not self.is_decoder_only
            and encoder_hidden_states is not None
        )
        if do_cross_attention:
            # the actual query length is unknown for cross attention
            # if using past key value states. Need to inject it here
            if present_key_value_state is not None:
                query_length = present_key_value_state[0].shape[2]
            else:
                query_length = None

            cross_attention_outputs = self.layer[1](
                x,
                key_value_states=encoder_hidden_states,
                attention_mask=encoder_attention_mask,
                position_bias=encoder_decoder_position_bias,
                # position_ids ?
                layer_head_mask=cross_attn_layer_head_mask,
                past_key_value=cross_attn_past_key_value,
                query_length=query_length,
                use_cache=use_cache,
                output_attentions=output_attentions,
            )
            x = cross_attention_outputs[0]

            # clamp inf values to enable fp16 training
            if x.dtype == torch.float16:
                clamp_value = torch.where(
                    torch.isinf(x).any(),
                    torch.finfo(x.dtype).max - 1000,
                    torch.finfo(x.dtype).max,
                )
                x = torch.clamp(x, min=-clamp_value, max=clamp_value)

            # Combine self attn and cross attn key value states
            if present_key_value_state is not None:
                present_key_value_state = (
                    present_key_value_state + cross_attention_outputs[1]
                )

            # Keep cross-attention outputs and relative position weights
            attention_outputs = attention_outputs + cross_attention_outputs[2:]

        if self.parallel_layers:
            # https://github.com/google/flaxformer/blob/ea17eb012a1d340ddff017b7a534c2162aaec34c/flaxformer/architectures/t5/t5_architecture.py#L534-L578
            x = x + ff_output
            x *= 2**-0.5
            hidden_states = hidden_states + self.layer[0].dropout(x)
        else:
            hidden_states = ff_layer(x)

        # clamp inf values to enable fp16 training
        if hidden_states.dtype == torch.float16:
            clamp_value = torch.where(
                torch.isinf(hidden_states).any(),
                torch.finfo(hidden_states.dtype).max - 1000,
                torch.finfo(hidden_states.dtype).max,
            )
            hidden_states = torch.clamp(
                hidden_states, min=-clamp_value, max=clamp_value
            )

        outputs = (hidden_states,)

        if use_cache:
            outputs = outputs + (present_key_value_state,) + attention_outputs
        else:
            outputs = outputs + attention_outputs

        return outputs  # hidden-states, present_key_value_states, (self-attention position bias), (self-attention weights), (cross-attention position bias), (cross-attention weights)


class DecoderOnlyT5Stack(modeling_t5.T5Stack):
    def __init__(self, config, embed_tokens=None):
        super(modeling_t5.T5Stack, self).__init__(config)

        self.embed_tokens = embed_tokens
        self.is_decoder = config.is_decoder

        self.block = nn.ModuleList(
            [
                DecoderOnlyT5Block(
                    config,
                    has_relative_attention_bias=(
                        config.has_relative_attention_bias and bool(i == 0)
                    ),
                )
                for i in range(config.num_layers)
            ]
        )
        if not config.parallel_layers:
            self.final_layer_norm = modeling_t5.T5LayerNorm(
                config.d_model, eps=config.layer_norm_epsilon
            )
        else:
            self.final_layer_norm = nn.Identity()
        self.dropout = nn.Dropout(config.dropout_rate)

        # Initialize weights and apply final processing
        self.post_init()
        # Model parallel
        self.model_parallel = False
        self.device_map = None
        self.gradient_checkpointing = False

    def forward(
        self,
        input_ids=None,
        position_ids=None,
        attention_mask=None,
        encoder_hidden_states=None,
        encoder_attention_mask=None,
        inputs_embeds=None,
        head_mask=None,
        cross_attn_head_mask=None,
        past_key_values=None,
        use_cache=None,
        output_attentions=None,
        output_hidden_states=None,
        return_dict=None,
    ):
        # Model parallel
        if self.model_parallel:
            torch.cuda.set_device(self.first_device)
            self.embed_tokens = self.embed_tokens.to(self.first_device)
        use_cache = use_cache if use_cache is not None else self.config.use_cache
        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 and inputs_embeds is not None:
            err_msg_prefix = "decoder_" if self.is_decoder else ""
            raise ValueError(
                f"You cannot specify both {err_msg_prefix}input_ids and {err_msg_prefix}inputs_embeds at the same time"
            )
        elif input_ids is not None:
            input_shape = input_ids.size()
            input_ids = input_ids.view(-1, input_shape[-1])
        elif inputs_embeds is not None:
            input_shape = inputs_embeds.size()[:-1]
        else:
            err_msg_prefix = "decoder_" if self.is_decoder else ""
            raise ValueError(
                f"You have to specify either {err_msg_prefix}input_ids or {err_msg_prefix}inputs_embeds"
            )

        if position_ids is None:
            seq_length = input_ids.shape[1]
            past_key_values_length = (
                0 if past_key_values is None else past_key_values[0][0].shape[2]
            )
            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)

        if inputs_embeds is None:
            if self.embed_tokens is None:
                raise ValueError(
                    "You have to initialize the model with valid token embeddings"
                )
            inputs_embeds = self.embed_tokens(input_ids)

        batch_size, seq_length = input_shape

        # required mask seq length can be calculated via length of past
        mask_seq_length = (
            past_key_values[0][0].shape[2] + seq_length
            if past_key_values is not None
            else seq_length
        )

        if use_cache is True:
            if not self.is_decoder:
                raise ValueError(
                    f"`use_cache` can only be set to `True` if {self} is used as a decoder"
                )

        if attention_mask is None:
            attention_mask = torch.ones(
                batch_size, mask_seq_length, device=inputs_embeds.device
            )
        if (
            self.is_decoder
            and encoder_attention_mask is None
            and encoder_hidden_states is not None
        ):
            encoder_seq_length = encoder_hidden_states.shape[1]
            encoder_attention_mask = torch.ones(
                batch_size,
                encoder_seq_length,
                device=inputs_embeds.device,
                dtype=torch.long,
            )

        # initialize past_key_values with `None` if past does not exist
        if past_key_values is None:
            past_key_values = [None] * len(self.block)

        # We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length]
        # ourselves in which case we just need to make it broadcastable to all heads.
        extended_attention_mask = self.get_extended_attention_mask(
            attention_mask, input_shape
        )

        # If a 2D or 3D attention mask is provided for the cross-attention
        # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
        if self.is_decoder and encoder_hidden_states is not None:
            (
                encoder_batch_size,
                encoder_sequence_length,
                _,
            ) = encoder_hidden_states.size()
            encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
            if encoder_attention_mask is None:
                encoder_attention_mask = torch.ones(
                    encoder_hidden_shape, device=inputs_embeds.device
                )
            encoder_extended_attention_mask = self.invert_attention_mask(
                encoder_attention_mask
            )
        else:
            encoder_extended_attention_mask = None

        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

        # Prepare head mask if needed
        head_mask = self.get_head_mask(head_mask, self.config.num_layers)
        cross_attn_head_mask = self.get_head_mask(
            cross_attn_head_mask, self.config.num_layers
        )
        present_key_value_states = () if use_cache else None
        all_hidden_states = () if output_hidden_states else None
        all_attentions = () if output_attentions else None
        all_cross_attentions = () if (output_attentions and self.is_decoder) else None
        position_bias = None
        encoder_decoder_position_bias = None

        hidden_states = self.dropout(inputs_embeds)

        for i, (layer_module, past_key_value) in enumerate(
            zip(self.block, past_key_values)
        ):
            layer_head_mask = head_mask[i]
            cross_attn_layer_head_mask = cross_attn_head_mask[i]
            # Model parallel
            if self.model_parallel:
                torch.cuda.set_device(hidden_states.device)
                # Ensure that attention_mask is always on the same device as hidden_states
                if attention_mask is not None:
                    attention_mask = attention_mask.to(hidden_states.device)
                if position_bias is not None:
                    position_bias = position_bias.to(hidden_states.device)
                if encoder_hidden_states is not None:
                    encoder_hidden_states = encoder_hidden_states.to(
                        hidden_states.device
                    )
                if encoder_extended_attention_mask is not None:
                    encoder_extended_attention_mask = (
                        encoder_extended_attention_mask.to(hidden_states.device)
                    )
                if encoder_decoder_position_bias is not None:
                    encoder_decoder_position_bias = encoder_decoder_position_bias.to(
                        hidden_states.device
                    )
                if layer_head_mask is not None:
                    layer_head_mask = layer_head_mask.to(hidden_states.device)
                if cross_attn_layer_head_mask is not None:
                    cross_attn_layer_head_mask = cross_attn_layer_head_mask.to(
                        hidden_states.device
                    )
            if output_hidden_states:
                all_hidden_states = all_hidden_states + (hidden_states,)

            if self.gradient_checkpointing and self.training:
                layer_outputs = self._gradient_checkpointing_func(
                    layer_module.forward,
                    hidden_states,
                    extended_attention_mask,
                    position_bias,
                    encoder_hidden_states,
                    encoder_extended_attention_mask,
                    encoder_decoder_position_bias,
                    layer_head_mask,
                    cross_attn_layer_head_mask,
                    None,  # past_key_value is always None with gradient checkpointing
                    use_cache,
                    output_attentions,
                )
            else:
                layer_outputs = layer_module(
                    hidden_states,
                    attention_mask=extended_attention_mask,
                    position_bias=position_bias,
                    position_ids=position_ids,
                    encoder_hidden_states=encoder_hidden_states,
                    encoder_attention_mask=encoder_extended_attention_mask,
                    encoder_decoder_position_bias=encoder_decoder_position_bias,
                    layer_head_mask=layer_head_mask,
                    cross_attn_layer_head_mask=cross_attn_layer_head_mask,
                    past_key_value=past_key_value,
                    use_cache=use_cache,
                    output_attentions=output_attentions,
                )

            # layer_outputs is a tuple with:
            # hidden-states, key-value-states, (self-attention position bias), (self-attention weights), (cross-attention position bias), (cross-attention weights)
            if use_cache is False:
                layer_outputs = layer_outputs[:1] + (None,) + layer_outputs[1:]

            hidden_states, present_key_value_state = layer_outputs[:2]

            # We share the position biases between the layers - the first layer store them
            # layer_outputs = hidden-states, key-value-states (self-attention position bias), (self-attention weights),
            # (cross-attention position bias), (cross-attention weights)
            position_bias = layer_outputs[2]
            if self.is_decoder and encoder_hidden_states is not None:
                encoder_decoder_position_bias = layer_outputs[
                    4 if output_attentions else 3
                ]
            # append next layer key value states
            if use_cache:
                present_key_value_states = present_key_value_states + (
                    present_key_value_state,
                )

            if output_attentions:
                all_attentions = all_attentions + (layer_outputs[3],)
                if self.is_decoder:
                    all_cross_attentions = all_cross_attentions + (layer_outputs[5],)

            # Model Parallel: If it's the last layer for that device, put things on the next device
            if self.model_parallel:
                for k, v in self.device_map.items():
                    if i == v[-1] and "cuda:" + str(k) != self.last_device:
                        hidden_states = hidden_states.to("cuda:" + str(k + 1))

        hidden_states = self.final_layer_norm(hidden_states)
        hidden_states = self.dropout(hidden_states)

        # Add last layer
        if output_hidden_states:
            all_hidden_states = all_hidden_states + (hidden_states,)

        if not return_dict:
            return tuple(
                v
                for v in [
                    hidden_states,
                    present_key_value_states,
                    all_hidden_states,
                    all_attentions,
                    all_cross_attentions,
                ]
                if v is not None
            )
        return modeling_t5.BaseModelOutputWithPastAndCrossAttentions(
            last_hidden_state=hidden_states,
            past_key_values=present_key_value_states,
            hidden_states=all_hidden_states,
            attentions=all_attentions,
            cross_attentions=all_cross_attentions,
        )


class DecoderOnlyT5Model(modeling_t5.T5ForConditionalGeneration):
    def __init__(self, config: DecoderOnlyT5Config):
        super(modeling_t5.T5ForConditionalGeneration, self).__init__(config)
        self.model_dim = config.d_model

        self.shared = nn.Embedding(config.vocab_size, config.d_model)
        assert (
            self.config.num_layers == 0
        ), "Decoder only model cannot have encoder layers"
        self.encoder = None

        decoder_config = copy.deepcopy(config)
        decoder_config.is_decoder = True
        decoder_config.is_encoder_decoder = False
        decoder_config.num_layers = config.num_decoder_layers
        self.decoder = DecoderOnlyT5Stack(decoder_config, self.shared)

        self.lm_head = nn.Linear(config.d_model, config.vocab_size, bias=False)

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

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

    def _tie_weights(self):
        if not self.config.tie_word_embeddings:
            return
        if self.encoder:
            self._tie_or_clone_weights(self.encoder.embed_tokens, self.shared)
        if self.decoder:
            self._tie_or_clone_weights(self.decoder.embed_tokens, self.shared)

    @add_start_docstrings_to_model_forward(modeling_t5.T5_INPUTS_DOCSTRING)
    @replace_return_docstrings(
        output_type=CausalLMOutputWithPast, config_class=_CONFIG_FOR_DOC
    )
    def forward(
        self,
        input_ids: Optional[torch.LongTensor] = None,
        position_ids: Optional[torch.LongTensor] = None,
        attention_mask: Optional[torch.FloatTensor] = None,
        past_key_values: Optional[Tuple[Tuple[torch.Tensor]]] = 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]:
        r"""
        labels (`torch.LongTensor` of shape `(batch_size,)`, *optional*):
            Labels for computing the sequence classification/regression loss. Indices should be in `[-100, 0, ...,
            config.vocab_size - 1]`. All labels set to `-100` are ignored (masked), the loss is only computed for
            labels in `[0, ..., config.vocab_size]`

        Returns:

        Examples:

        ```"""
        use_cache = use_cache if use_cache is not None else self.config.use_cache
        return_dict = (
            return_dict if return_dict is not None else self.config.use_return_dict
        )

        if self.model_parallel:
            torch.cuda.set_device(self.decoder.first_device)

        # Set device for model parallelism
        if self.model_parallel:
            torch.cuda.set_device(self.decoder.first_device)
            if input_ids is not None:
                input_ids = input_ids.to(self.decoder.first_device)
            if attention_mask is not None:
                attention_mask = attention_mask.to(self.decoder.first_device)

        # Decode
        outputs = self.decoder(
            input_ids=input_ids,
            position_ids=position_ids,
            attention_mask=attention_mask,
            inputs_embeds=inputs_embeds,
            past_key_values=past_key_values,
            encoder_hidden_states=None,
            encoder_attention_mask=None,
            head_mask=None,
            cross_attn_head_mask=None,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        sequence_output = outputs[0]

        # Set device for model parallelism
        if self.model_parallel:
            torch.cuda.set_device(self.decoder.first_device)
            self.lm_head = self.lm_head.to(self.decoder.first_device)
            sequence_output = sequence_output.to(self.lm_head.weight.device)

        if self.config.tie_word_embeddings:
            # Rescale output before projecting on vocab
            # See https://github.com/tensorflow/mesh/blob/fa19d69eafc9a482aff0b59ddd96b025c0cb207d/mesh_tensorflow/transformer/transformer.py#L586
            sequence_output = sequence_output * (self.model_dim**-0.5)

        lm_logits = self.lm_head(sequence_output)

        loss = None
        if labels is not None:
            loss_fct = CrossEntropyLoss(ignore_index=-100)
            # move labels to correct device to enable PP
            labels = labels.to(lm_logits.device)
            loss = loss_fct(lm_logits.view(-1, lm_logits.size(-1)), labels.view(-1))
            # TODO(thom): Add z_loss https://github.com/tensorflow/mesh/blob/fa19d69eafc9a482aff0b59ddd96b025c0cb207d/mesh_tensorflow/layers.py#L666

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

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