public/gpt-2/transformers/models/fsmt/modeling_fsmt.py

# coding=utf-8
# Copyright 2020 The Facebook AI Research Team Authors and The HuggingFace Inc. team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
# Original implementation: https://github.com/pytorch/fairseq/tree/master/examples/wmt19
# Authors:
# - @alexeib Alexei Baevski
# - @edunov Sergey Edunov
# - @michaelauli Michael Auli
# - @myleott Myle Ott
# - @nng555 Nathan Ng
# - David Grangier
# - Kyra Yee
#
# Paper: Facebook FAIR's WMT19 News Translation Task Submission https://arxiv.org/abs/1907.06616
#
"""PyTorch Fairseq model, ported from https://github.com/pytorch/fairseq/tree/master/examples/wmt19"""

import math
import random
from typing import Any, Dict, List, Optional, Tuple

import torch
from torch import Tensor, nn
from torch.nn import CrossEntropyLoss, LayerNorm

from ...activations import ACT2FN
from ...deepspeed import is_deepspeed_zero3_enabled
from ...file_utils import (
    add_code_sample_docstrings,
    add_end_docstrings,
    add_start_docstrings,
    add_start_docstrings_to_model_forward,
    replace_return_docstrings,
)
from ...modeling_outputs import (
    BaseModelOutput,
    BaseModelOutputWithPastAndCrossAttentions,
    Seq2SeqLMOutput,
    Seq2SeqModelOutput,
)
from ...modeling_utils import PreTrainedModel
from ...utils import logging
from .configuration_fsmt import FSMTConfig


logger = logging.get_logger(__name__)

_CHECKPOINT_FOR_DOC = "facebook/wmt19-ru-en"
_CONFIG_FOR_DOC = "FSMTConfig"
_TOKENIZER_FOR_DOC = "FSMTTokenizer"

# See all FSMT models at https://huggingface.co/models?filter=fsmt

# Porting notes:
# this one is modeled after BartModel*
#
# Currently only translation (fairseq also has weights for LM)
#
# fairseq provides weights for ru-en, en-ru and de-en, en-de pairs. All have been ported.
# - ru-en, en-ru use asymmetric vocab
# - de-en, en-de use a merged single vocab (but the code works as if they are separate)
#
# Differences with Bart:
# - not using bos token
# - 2 separate vocabs (src and target)
# - embed weights aren't tied
# - uses a model Ensemble (but that part isn't ported/implemented yet) - so we
#   aren't getting as good of a BLEU score
# - uses a projection layer at the end of the decoder
# - doesn't use final_logits_bias
# - beam search: stops as soon as num_beams == len(hypos) (whereas transformers
#   is not satisfied there and will continue searching until the next cycles
#   aren't promising something better), comparing BLEU scores - the transformers
#   algorithm is slightly superior, therefore using the latter. But if you want
#   to match fairseq outputs, you need to pass ``early_stopping=True`` to ``generate()``.
#
# SinusoidalPositionalEmbedding is slightly different from Bart's - generates
# different embeddings. This implementation is copied verbatim from fairseq with
# some small changes to make it work here.
#
# Other changes:
#  - doesn't support use_cache as Bart's version does
#
#
# FSMTConfig changes with BartConfig
#
#    Differences with BART:
#    - src/tgt vocabs aren't shared
#    - token embeddings aren't shared
#    - needs a language pair
#    - scale_embedding are True
#
#    some unused args were removed too
#
#
# TODO:
# - port model ensemble (fs uses 4 model checkpoints)
# - solve beam search discrepancies
# docstyle-ignore

"""

Here is how to compare BLEU scores against fairseq implementation:

# en-ru

export PAIR=en-ru
export DATA_DIR=data/$PAIR
export SAVE_DIR=data/$PAIR
export BS=8
export NUM_BEAMS=50
mkdir -p $DATA_DIR
sacrebleu -t wmt19 -l $PAIR --echo src > $DATA_DIR/val.source
sacrebleu -t wmt19 -l $PAIR --echo ref > $DATA_DIR/val.target
echo $PAIR
PYTHONPATH="src:examples/seq2seq" python examples/seq2seq/run_eval.py facebook/wmt19-$PAIR $DATA_DIR/val.source $SAVE_DIR/test_translations.txt --reference_path $DATA_DIR/val.target --score_path $SAVE_DIR/test_bleu.json --bs $BS --task translation --num_beams $NUM_BEAMS

# (fairseq BLEU: 36.4 http://matrix.statmt.org/matrix/output/1914?score_id=37605)


# ru-en

export PAIR=ru-en
export DATA_DIR=data/$PAIR
export SAVE_DIR=data/$PAIR
export BS=8
export NUM_BEAMS=50
mkdir -p $DATA_DIR
sacrebleu -t wmt19 -l $PAIR --echo src > $DATA_DIR/val.source
sacrebleu -t wmt19 -l $PAIR --echo ref > $DATA_DIR/val.target
PYTHONPATH="src:examples/seq2seq" python examples/seq2seq/run_eval.py facebook/wmt19-$PAIR $DATA_DIR/val.source $SAVE_DIR/test_translations.txt --reference_path $DATA_DIR/val.target --score_path $SAVE_DIR/test_bleu.json --bs $BS --task translation --num_beams $NUM_BEAMS


# (fairseq BLEU: 41.3 http://matrix.statmt.org/matrix/output/1907?run_id=6937)


# de-en

export PAIR=de-en
export DATA_DIR=data/$PAIR
export SAVE_DIR=data/$PAIR
export BS=8
export NUM_BEAMS=50
mkdir -p $DATA_DIR
sacrebleu -t wmt19 -l $PAIR --echo src > $DATA_DIR/val.source
sacrebleu -t wmt19 -l $PAIR --echo ref > $DATA_DIR/val.target
echo $PAIR
PYTHONPATH="src:examples/seq2seq" python examples/seq2seq/run_eval.py facebook/wmt19-$PAIR $DATA_DIR/val.source $SAVE_DIR/test_translations.txt --reference_path $DATA_DIR/val.target --score_path $SAVE_DIR/test_bleu.json --bs $BS --task translation --num_beams $NUM_BEAMS

# (fairseq BLEU: 42.3 http://matrix.statmt.org/matrix/output/1902?run_id=6750)



# en-de

export PAIR=en-de
export DATA_DIR=data/$PAIR
export SAVE_DIR=data/$PAIR
export BS=8
mkdir -p $DATA_DIR
sacrebleu -t wmt19 -l $PAIR --echo src > $DATA_DIR/val.source
sacrebleu -t wmt19 -l $PAIR --echo ref > $DATA_DIR/val.target
echo $PAIR
PYTHONPATH="src:examples/seq2seq" python examples/seq2seq/run_eval.py facebook/wmt19-$PAIR $DATA_DIR/val.source $SAVE_DIR/test_translations.txt --reference_path $DATA_DIR/val.target --score_path $SAVE_DIR/test_bleu.json --bs $BS --task translation --num_beams $NUM_BEAMS

# (fairseq BLEU: 43.1 http://matrix.statmt.org/matrix/output/1909?run_id=6862)

"""


FSMT_START_DOCSTRING = r"""

    This model inherits from :class:`~transformers.PreTrainedModel`. Check the superclass documentation for the generic
    methods the library implements for all its model (such as downloading or saving, resizing the input embeddings,
    pruning heads etc.)

    This model is also a PyTorch `torch.nn.Module <https://pytorch.org/docs/stable/nn.html#torch.nn.Module>`__
    subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to
    general usage and behavior.

    Parameters:
        config (:class:`~transformers.FSMTConfig`): Model configuration class with all the parameters of the model.
            Initializing with a config file does not load the weights associated with the model, only the
            configuration. Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model
            weights.

"""
FSMT_GENERATION_EXAMPLE = r"""
    Translation example::

        from transformers import FSMTTokenizer, FSMTForConditionalGeneration

        mname = "facebook/wmt19-ru-en"
        model = FSMTForConditionalGeneration.from_pretrained(mname)
        tokenizer = FSMTTokenizer.from_pretrained(mname)

        src_text = "Машинное обучение - это здорово, не так ли?"
        input_ids = tokenizer.encode(src_text, return_tensors='pt')
        outputs = model.generate(input_ids, num_beams=5, num_return_sequences=3)
        for i, output in enumerate(outputs):
            decoded = tokenizer.decode(output, skip_special_tokens=True)
            print(f"{i}: {decoded})
         # 1: Machine learning is great, isn't it? ...

"""

FSMT_INPUTS_DOCSTRING = r"""
    Args:
        input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`):
            Indices of input sequence tokens in the vocabulary.

            IIndices can be obtained using :class:`~transformers.FSTMTokenizer`. See
            :meth:`transformers.PreTrainedTokenizer.encode` and :meth:`transformers.PreTrainedTokenizer.__call__` for
            details.

            `What are input IDs? <../glossary.html#input-ids>`__
        attention_mask (:obj:`torch.Tensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
            Mask to avoid performing attention on padding token indices. Mask values selected in ``[0, 1]``:

            - 1 for tokens that are **not masked**,
            - 0 for tokens that are **masked**.

            `What are attention masks? <../glossary.html#attention-mask>`__
        decoder_input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch_size, target_sequence_length)`, `optional`):
            Indices of decoder input sequence tokens in the vocabulary.

            Indices can be obtained using :class:`~transformers.FSMTTokenizer`. See
            :meth:`transformers.PreTrainedTokenizer.encode` and :meth:`transformers.PreTrainedTokenizer.__call__` for
            details.

            `What are decoder input IDs? <../glossary.html#decoder-input-ids>`__

            FSMT uses the :obj:`eos_token_id` as the starting token for :obj:`decoder_input_ids` generation. If
            :obj:`past_key_values` is used, optionally only the last :obj:`decoder_input_ids` have to be input (see
            :obj:`past_key_values`).
        decoder_attention_mask (:obj:`torch.BoolTensor` of shape :obj:`(batch_size, target_sequence_length)`, `optional`):
            Default behavior: generate a tensor that ignores pad tokens in :obj:`decoder_input_ids`. Causal mask will
            also be used by default.
        head_mask (:obj:`torch.Tensor` of shape :obj:`(encoder_layers, encoder_attention_heads)`, `optional`):
            Mask to nullify selected heads of the attention modules in the encoder. Mask values selected in ``[0, 1]``:

            - 1 indicates the head is **not masked**,
            - 0 indicates the head is **masked**.

        decoder_head_mask (:obj:`torch.Tensor` of shape :obj:`(decoder_layers, decoder_attention_heads)`, `optional`):
            Mask to nullify selected heads of the attention modules in the decoder. Mask values selected in ``[0, 1]``:

            - 1 indicates the head is **not masked**,
            - 0 indicates the head is **masked**.

        cross_attn_head_mask (:obj:`torch.Tensor` of shape :obj:`(decoder_layers, decoder_attention_heads)`, `optional`):
            Mask to nullify selected heads of the cross-attention modules in the decoder. Mask values selected in ``[0,
            1]``:

            - 1 indicates the head is **not masked**,
            - 0 indicates the head is **masked**.

        encoder_outputs (:obj:`Tuple(torch.FloatTensor)`, `optional`):
            Tuple consists of (:obj:`last_hidden_state`, `optional`: :obj:`hidden_states`, `optional`:
            :obj:`attentions`) :obj:`last_hidden_state` of shape :obj:`(batch_size, sequence_length, hidden_size)` is a
            sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention of
            the decoder.
        past_key_values (:obj:`Tuple(torch.FloatTensor)` of length :obj:`config.n_layers` with each tuple having 4 tensors of shape :obj:`(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`):
            Contains precomputed key and value hidden-states of the attention blocks. Can be used to speed up decoding.
            If :obj:`past_key_values` are used, the user can optionally input only the last :obj:`decoder_input_ids`
            (those that don't have their past key value states given to this model) of shape :obj:`(batch_size, 1)`
            instead of all :obj:`decoder_input_ids` of shape :obj:`(batch_size, sequence_length)`.
        use_cache (:obj:`bool`, `optional`, defaults to :obj:`True`):
            If set to :obj:`True`, :obj:`past_key_values` key value states are returned and can be used to speed up
            decoding (see :obj:`past_key_values`).
        output_attentions (:obj:`bool`, `optional`):
            Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under returned
            tensors for more detail.
        output_hidden_states (:obj:`bool`, `optional`):
            Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors for
            more detail.
        return_dict (:obj:`bool`, `optional`):
            Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple.
"""


def invert_mask(attention_mask):
    """Turns 1->0, 0->1, False->True, True-> False"""
    assert attention_mask.dim() == 2
    return attention_mask.eq(0)


def triu_onnx(x, diagonal=0):
    l = x.shape[0]
    arange = torch.arange(l, device=x.device)
    mask = arange.expand(l, l)
    arange = arange.unsqueeze(-1)
    if diagonal:
        arange = arange + diagonal
    mask = mask >= arange
    return x.masked_fill(mask == 0, 0)


def _prepare_fsmt_decoder_inputs(
    config,
    input_ids,
    decoder_input_ids=None,
    decoder_padding_mask=None,
    causal_mask_dtype=torch.float32,
):
    """
    Prepare masks that ignore padding tokens in the decoder and a causal mask for the decoder if none are provided.
    This mimics the default behavior in fairseq. To override it pass in masks. Note: this is not called during
    generation
    """
    pad_token_id = config.pad_token_id
    if decoder_input_ids is None:
        decoder_input_ids = shift_tokens_right(input_ids, pad_token_id)
    bsz, tgt_len = decoder_input_ids.size()
    if decoder_padding_mask is None:
        decoder_padding_mask = make_padding_mask(decoder_input_ids, pad_token_id)
    else:
        decoder_padding_mask = invert_mask(decoder_padding_mask)
    causal_mask = triu_onnx(fill_with_neg_inf(torch.zeros(tgt_len, tgt_len)), 1).to(
        dtype=causal_mask_dtype, device=decoder_input_ids.device
    )
    return decoder_input_ids, decoder_padding_mask, causal_mask


class PretrainedFSMTModel(PreTrainedModel):
    config_class = FSMTConfig
    base_model_prefix = "model"

    def _init_weights(self, module):
        std = self.config.init_std
        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, SinusoidalPositionalEmbedding):
            pass
        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_()

    @property
    def dummy_inputs(self):
        pad_token = self.config.pad_token_id
        input_ids = torch.tensor([[0, 6, 10, 4, 2], [0, 8, 12, 2, pad_token]], device=self.device)
        dummy_inputs = {
            "attention_mask": input_ids.ne(pad_token),
            "input_ids": input_ids,
        }
        return dummy_inputs


def _make_linear_from_emb(emb):
    vocab_size, emb_size = emb.weight.shape
    lin_layer = nn.Linear(vocab_size, emb_size, bias=False)
    lin_layer.weight.data = emb.weight.data
    return lin_layer


# Helper Functions, mostly for making masks
def _check_shapes(shape_1, shape2):
    if shape_1 != shape2:
        raise AssertionError(f"shape mismatch: {shape_1} != {shape2}")


def shift_tokens_right(input_ids, pad_token_id):
    """Shift input ids one token to the right, and wrap the last non pad token (usually <eos>)."""
    prev_output_tokens = input_ids.clone()
    index_of_eos = (input_ids.ne(pad_token_id).sum(dim=1) - 1).unsqueeze(-1)
    prev_output_tokens[:, 0] = input_ids.gather(1, index_of_eos).squeeze()
    prev_output_tokens[:, 1:] = input_ids[:, :-1]
    return prev_output_tokens


def make_padding_mask(input_ids, padding_idx=1):
    """True for pad tokens"""
    padding_mask = input_ids.eq(padding_idx)
    if not padding_mask.any():
        padding_mask = None
    return padding_mask


# Helper Modules


class EncoderLayer(nn.Module):
    def __init__(self, config: FSMTConfig):
        super().__init__()
        self.embed_dim = config.d_model
        self.self_attn = Attention(self.embed_dim, config.encoder_attention_heads, dropout=config.attention_dropout)
        self.self_attn_layer_norm = LayerNorm(self.embed_dim)
        self.dropout = config.dropout
        self.activation_fn = ACT2FN[config.activation_function]
        self.activation_dropout = config.activation_dropout
        self.fc1 = nn.Linear(self.embed_dim, config.encoder_ffn_dim)
        self.fc2 = nn.Linear(config.encoder_ffn_dim, self.embed_dim)
        self.final_layer_norm = LayerNorm(self.embed_dim)

    def forward(self, x, encoder_padding_mask, layer_head_mask, output_attentions=False):
        """
        Args:
            x (:obj:`torch.Tensor`): input to the layer of shape `(seq_len, batch, embed_dim)`
            encoder_padding_mask (:obj:`torch.ByteTensor`): binary ByteTensor of shape
                `(batch, src_len)` where padding elements are indicated by ``1``.
            for t_tgt, t_src is excluded (or masked out), =0 means it is
            included in attention
            layer_head_mask (:obj:`torch.FloatTensor`): mask for attention heads in a given layer of size
                `(config.encoder_attention_heads,)`.

        Returns:
            encoded output of shape `(seq_len, batch, embed_dim)`
        """
        residual = x
        x, attn_weights = self.self_attn(
            query=x,
            key=x,
            key_padding_mask=encoder_padding_mask,
            layer_head_mask=layer_head_mask,
            output_attentions=output_attentions,
        )
        x = nn.functional.dropout(x, p=self.dropout, training=self.training)
        x = residual + x
        x = self.self_attn_layer_norm(x)

        residual = x
        x = self.activation_fn(self.fc1(x))
        x = nn.functional.dropout(x, p=self.activation_dropout, training=self.training)
        x = self.fc2(x)
        x = nn.functional.dropout(x, p=self.dropout, training=self.training)
        x = residual + x
        x = self.final_layer_norm(x)
        return x, attn_weights


class FSMTEncoder(nn.Module):
    """
    Transformer encoder consisting of *config.encoder_layers* self attention layers. Each layer is a
    :class:`EncoderLayer`.

    Args:
        config: FSMTConfig
    """

    def __init__(self, config: FSMTConfig, embed_tokens):
        super().__init__()
        self.dropout = config.dropout
        self.layerdrop = config.encoder_layerdrop
        self.padding_idx = embed_tokens.padding_idx
        self.embed_tokens = embed_tokens
        embed_dim = embed_tokens.embedding_dim
        self.embed_scale = math.sqrt(embed_dim) if config.scale_embedding else 1.0
        self.embed_positions = SinusoidalPositionalEmbedding(
            config.max_position_embeddings + self.padding_idx + 1, embed_dim, self.padding_idx
        )
        self.layers = nn.ModuleList(
            [EncoderLayer(config) for _ in range(config.encoder_layers)]
        )  # type: List[EncoderLayer]

    def forward(
        self,
        input_ids,
        attention_mask=None,
        head_mask=None,
        output_attentions=False,
        output_hidden_states=False,
        return_dict=True,
    ):
        """
        Args:
            input_ids (:obj:`torch.LongTensor`): tokens in the source language of shape
                `(batch, src_len)`
            attention_mask (:obj:`torch.LongTensor`): indicating which indices are padding tokens
            head_mask (:obj:`torch.Tensor` of shape :obj:`(num_layers, num_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**.

        Returns:
            BaseModelOutput or Tuple comprised of:

                - **x** (:obj:`torch.Tensor`): the last encoder layer's output of shape `(src_len, batch, embed_dim)`
                - **encoder_states** (:obj:`Tuple(torch.FloatTensor`)): all intermediate hidden states of shape
                  `(src_len, batch, embed_dim)`. Only populated if *output_hidden_states:* is True.
                - **all_attentions** (:obj:`Tuple(torch.FloatTensor`)): Attention weights for each layer.
                During training might not be of length n_layers because of layer dropout.
        """
        # check attention mask and invert
        if attention_mask is not None:
            attention_mask = invert_mask(attention_mask)

        inputs_embeds = self.embed_tokens(input_ids) * self.embed_scale
        embed_pos = self.embed_positions(input_ids)
        x = inputs_embeds + embed_pos
        x = nn.functional.dropout(x, p=self.dropout, training=self.training)

        # B x T x C -> T x B x C
        x = x.transpose(0, 1)

        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:
                x = x.transpose(0, 1)  # T x B x C -> B x T x C
                encoder_states += (x,)
                x = x.transpose(0, 1)  # B x T x C -> T x B x C
            # 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
                attn = None
            else:
                x, attn = encoder_layer(
                    x,
                    attention_mask,
                    layer_head_mask=(head_mask[idx] if head_mask is not None else None),
                    output_attentions=output_attentions,
                )

            if output_attentions:
                all_attentions = all_attentions + (attn,)

        # T x B x C -> B x T x C
        x = x.transpose(0, 1)

        if output_hidden_states:
            encoder_states += (x,)

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


class DecoderLayer(nn.Module):
    def __init__(self, config: FSMTConfig):
        super().__init__()
        self.embed_dim = config.d_model

        self.self_attn = Attention(
            embed_dim=self.embed_dim,
            num_heads=config.decoder_attention_heads,
            dropout=config.attention_dropout,
        )
        self.dropout = config.dropout
        self.activation_fn = ACT2FN[config.activation_function]
        self.activation_dropout = config.activation_dropout

        self.self_attn_layer_norm = LayerNorm(self.embed_dim)
        self.encoder_attn = Attention(
            self.embed_dim,
            config.decoder_attention_heads,
            dropout=config.attention_dropout,
            encoder_decoder_attention=True,
        )
        self.encoder_attn_layer_norm = LayerNorm(self.embed_dim)
        self.fc1 = nn.Linear(self.embed_dim, config.decoder_ffn_dim)
        self.fc2 = nn.Linear(config.decoder_ffn_dim, self.embed_dim)
        self.final_layer_norm = LayerNorm(self.embed_dim)

    def forward(
        self,
        x,
        encoder_hidden_states,
        encoder_attn_mask=None,
        layer_state=None,
        causal_mask=None,
        layer_head_mask=None,
        cross_attn_layer_head_mask=None,
        decoder_padding_mask=None,
        output_attentions=False,
    ):
        residual = x

        if layer_state is None:
            layer_state = {}

        # Self Attention
        x, self_attn_weights = self.self_attn(
            query=x,
            key=x,
            layer_state=layer_state,  # adds keys to layer state
            key_padding_mask=decoder_padding_mask,
            attn_mask=causal_mask,
            layer_head_mask=layer_head_mask,
            output_attentions=output_attentions,
        )
        x = nn.functional.dropout(x, p=self.dropout, training=self.training)
        x = residual + x
        x = self.self_attn_layer_norm(x)

        # Cross attention
        residual = x
        assert self.encoder_attn.cache_key != self.self_attn.cache_key
        x, cross_attn_weights = self.encoder_attn(
            query=x,
            key=encoder_hidden_states,
            key_padding_mask=encoder_attn_mask,
            layer_state=layer_state,  # mutates layer state
            layer_head_mask=cross_attn_layer_head_mask,
            output_attentions=output_attentions,
        )
        x = nn.functional.dropout(x, p=self.dropout, training=self.training)
        x = residual + x
        x = self.encoder_attn_layer_norm(x)

        # Fully Connected
        residual = x
        x = self.activation_fn(self.fc1(x))
        x = nn.functional.dropout(x, p=self.activation_dropout, training=self.training)
        x = self.fc2(x)
        x = nn.functional.dropout(x, p=self.dropout, training=self.training)
        x = residual + x
        x = self.final_layer_norm(x)
        return (
            x,
            self_attn_weights,
            layer_state,
            cross_attn_weights,
        )  # layer_state = cache for decoding


class FSMTDecoder(nn.Module):
    """
    Transformer decoder consisting of *config.decoder_layers* layers. Each layer is a :class:`DecoderLayer`

    Args:
        config: FSMTConfig
        embed_tokens (nn.Embedding): output embedding
    """

    def __init__(self, config: FSMTConfig, embed_tokens: nn.Embedding):
        super().__init__()
        self.dropout = config.dropout
        self.layerdrop = config.decoder_layerdrop
        self.padding_idx = embed_tokens.padding_idx
        self.embed_scale = math.sqrt(config.d_model) if config.scale_embedding else 1.0
        self.embed_tokens = embed_tokens
        embed_dim = embed_tokens.embedding_dim
        self.embed_positions = SinusoidalPositionalEmbedding(
            config.max_position_embeddings + self.padding_idx + 1, embed_dim, self.padding_idx
        )
        self.layers = nn.ModuleList(
            [DecoderLayer(config) for _ in range(config.decoder_layers)]
        )  # type: List[DecoderLayer]

        if is_deepspeed_zero3_enabled():
            import deepspeed

            with deepspeed.zero.GatheredParameters(self.embed_tokens.weight, modifier_rank=None):
                embed_tokens_weight_shape = self.embed_tokens.weight.shape
        else:
            embed_tokens_weight_shape = self.embed_tokens.weight.shape
        self.output_projection = nn.Linear(embed_tokens_weight_shape[1], embed_tokens_weight_shape[0], bias=False)
        self.output_projection.weight = self.embed_tokens.weight

    def forward(
        self,
        input_ids,
        encoder_hidden_states,
        encoder_padding_mask,
        decoder_padding_mask,
        decoder_causal_mask,
        head_mask=None,
        cross_attn_head_mask=None,
        past_key_values=None,
        use_cache=False,
        output_attentions=False,
        output_hidden_states=False,
        return_dict=True,
    ):
        """
        Includes several features from "Jointly Learning to Align and Translate with Transformer Models" (Garg et al.,
        EMNLP 2019).

        Args:
            input_ids (:obj:`torch.LongTensor` of shape :obj:`(batch, tgt_len)`):
                previous decoder outputs for teacher forcing
            encoder_hidden_states: output from the encoder, used for
                encoder-side attention
            encoder_padding_mask: for ignoring pad tokens
            past_key_values (dict or None): dictionary used for storing state during generation
            head_mask (:obj:`torch.Tensor` of shape :obj:`(num_layers, num_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**.

            cross_attn_head_mask (:obj:`torch.Tensor` of shape :obj:`(num_layers, num_heads)`, `optional`):
                Mask to nullify selected heads of the cross-attention modules. Mask values selected in ``[0, 1]``:

                - 1 indicates the head is **not masked**,
                - 0 indicates the head is **masked**.

        Returns:
            BaseModelOutputWithPast or tuple:

                - the decoder's features of shape `(batch, tgt_len, embed_dim)`
                - the cache
                - hidden states
                - attentions
        """
        # check attention mask and invert
        if encoder_padding_mask is not None:
            encoder_padding_mask = invert_mask(encoder_padding_mask)

        # embed positions
        positions = self.embed_positions(input_ids)  # , use_cache=use_cache)

        if use_cache:
            input_ids = input_ids[:, -1:]
            positions = positions[:, -1:]  # happens after we embed them
            # assert input_ids.ne(self.padding_idx).any()

        x = self.embed_tokens(input_ids) * self.embed_scale
        x += positions
        x = nn.functional.dropout(x, p=self.dropout, training=self.training)

        # Convert to FSMT output format: (seq_len, BS, model_dim) -> (BS, seq_len, model_dim)
        x = x.transpose(0, 1)
        encoder_hidden_states = encoder_hidden_states.transpose(0, 1)

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

        # check if head_mask has a correct number of layers specified if desired
        for attn_mask, mask_name in zip([head_mask, cross_attn_head_mask], ["head_mask", "cross_attn_head_mask"]):
            if attn_mask is not None:
                assert attn_mask.size()[0] == (
                    len(self.layers)
                ), f"The `{mask_name}` should be specified for {len(self.layers)} layers, but it is for {head_mask.size()[0]}."
        for idx, decoder_layer in enumerate(self.layers):
            # add LayerDrop (see https://arxiv.org/abs/1909.11556 for description)
            if output_hidden_states:
                x = x.transpose(0, 1)
                all_hidden_states += (x,)
                x = x.transpose(0, 1)
            dropout_probability = random.uniform(0, 1)
            if self.training and (dropout_probability < self.layerdrop):
                continue

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

            x, layer_self_attn, layer_past, layer_cross_attn = decoder_layer(
                x,
                encoder_hidden_states,
                encoder_attn_mask=encoder_padding_mask,
                decoder_padding_mask=decoder_padding_mask,
                layer_state=layer_state,
                causal_mask=decoder_causal_mask,
                layer_head_mask=(head_mask[idx] if head_mask is not None else None),
                cross_attn_layer_head_mask=(cross_attn_head_mask[idx] if cross_attn_head_mask is not None else None),
                output_attentions=output_attentions,
            )

            if use_cache:
                next_decoder_cache.append(layer_past.copy())

            if output_attentions:
                all_self_attns += (layer_self_attn,)
                all_cross_attns += (layer_cross_attn,)

        # add hidden states from the last decoder layer
        if output_hidden_states:
            x = x.transpose(0, 1)
            all_hidden_states += (x,)
            x = x.transpose(0, 1)

        # Convert to standard output format: (seq_len, BS, model_dim) -> (BS, seq_len, model_dim)
        x = x.transpose(0, 1)
        encoder_hidden_states = encoder_hidden_states.transpose(0, 1)

        x = self.output_projection(x)

        next_cache = next_decoder_cache if use_cache else None

        if not return_dict:
            return tuple(
                v for v in [x, next_cache, all_hidden_states, all_self_attns, all_cross_attns] if v is not None
            )
        return BaseModelOutputWithPastAndCrossAttentions(
            last_hidden_state=x,
            past_key_values=next_cache,
            hidden_states=all_hidden_states,
            attentions=all_self_attns,
            cross_attentions=all_cross_attns,
        )


def _reorder_buffer(attn_cache, new_order):
    for k, input_buffer_k in attn_cache.items():
        if input_buffer_k is not None:
            attn_cache[k] = input_buffer_k.index_select(0, new_order)
    return attn_cache


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

    def __init__(
        self,
        embed_dim,
        num_heads,
        dropout=0.0,
        bias=True,
        encoder_decoder_attention=False,  # otherwise self_attention
    ):
        super().__init__()
        self.embed_dim = embed_dim
        self.num_heads = num_heads
        self.dropout = dropout
        self.head_dim = embed_dim // num_heads
        assert self.head_dim * num_heads == self.embed_dim, "embed_dim must be divisible by num_heads"
        self.scaling = self.head_dim ** -0.5

        self.encoder_decoder_attention = encoder_decoder_attention
        self.k_proj = nn.Linear(embed_dim, embed_dim, bias=bias)
        self.v_proj = nn.Linear(embed_dim, embed_dim, bias=bias)
        self.q_proj = nn.Linear(embed_dim, embed_dim, bias=bias)
        self.out_proj = nn.Linear(embed_dim, embed_dim, bias=bias)
        self.cache_key = "encoder_decoder" if self.encoder_decoder_attention else "self"

    def _shape(self, tensor, seq_len, bsz):
        return tensor.contiguous().view(seq_len, bsz * self.num_heads, self.head_dim).transpose(0, 1)

    def forward(
        self,
        query,
        key: Optional[Tensor],
        key_padding_mask: Optional[Tensor] = None,
        layer_state: Optional[Dict[str, Optional[Tensor]]] = None,
        attn_mask: Optional[Tensor] = None,
        layer_head_mask: Optional[Tensor] = None,
        output_attentions=False,
    ) -> Tuple[Tensor, Optional[Tensor]]:
        """Input shape: Time(SeqLen) x Batch x Channel"""
        static_kv: bool = self.encoder_decoder_attention
        tgt_len, bsz, embed_dim = query.size()
        assert embed_dim == self.embed_dim
        assert list(query.size()) == [tgt_len, bsz, embed_dim]
        # get here for encoder decoder cause of static_kv
        if layer_state is not None:  # reuse k,v and encoder_padding_mask
            saved_state = layer_state.get(self.cache_key, {})
            if "prev_key" in saved_state and static_kv:
                # previous time steps are cached - no need to recompute key and value if they are static
                key = None
        else:
            saved_state = None
            layer_state = {}

        q = self.q_proj(query) * self.scaling
        if static_kv:
            if key is None:
                k = v = None
            else:
                k = self.k_proj(key)
                v = self.v_proj(key)
        else:
            k = self.k_proj(query)
            v = self.v_proj(query)

        q = self._shape(q, tgt_len, bsz)
        if k is not None:
            k = self._shape(k, -1, bsz)
        if v is not None:
            v = self._shape(v, -1, bsz)

        if saved_state is not None:
            k, v, key_padding_mask = self._use_saved_state(k, v, saved_state, key_padding_mask, static_kv, bsz)

        # Update cache
        layer_state[self.cache_key] = {
            "prev_key": k.view(bsz, self.num_heads, -1, self.head_dim),
            "prev_value": v.view(bsz, self.num_heads, -1, self.head_dim),
            "prev_key_padding_mask": key_padding_mask if not static_kv else None,
        }

        assert k is not None
        src_len = k.size(1)
        attn_weights = torch.bmm(q, k.transpose(1, 2))
        assert attn_weights.size() == (bsz * self.num_heads, tgt_len, src_len)

        if attn_mask is not None:
            attn_weights = attn_weights.view(bsz, self.num_heads, tgt_len, src_len) + attn_mask
            attn_weights = attn_weights.view(bsz * self.num_heads, tgt_len, src_len)

        # This is part of a workaround to get around fork/join parallelism not supporting Optional types.
        if key_padding_mask is not None and key_padding_mask.dim() == 0:
            key_padding_mask = None
        assert key_padding_mask is None or key_padding_mask.size()[:2] == (
            bsz,
            src_len,
        )

        if key_padding_mask is not None:  # don't attend to padding symbols
            attn_weights = attn_weights.view(bsz, self.num_heads, tgt_len, src_len)
            reshaped = key_padding_mask.unsqueeze(1).unsqueeze(2)
            attn_weights = attn_weights.masked_fill(reshaped, float("-inf"))
            attn_weights = attn_weights.view(bsz * self.num_heads, tgt_len, src_len)

        attn_weights = nn.functional.softmax(attn_weights, dim=-1)

        if layer_head_mask is not None:
            assert layer_head_mask.size() == (
                self.num_heads,
            ), f"Head mask for a single layer should be of size {(self.num_heads,)}, but is {layer_head_mask.size()}"
            attn_weights = layer_head_mask.view(1, -1, 1, 1) * attn_weights.view(bsz, self.num_heads, tgt_len, src_len)
            attn_weights = attn_weights.view(bsz * self.num_heads, tgt_len, src_len)

        if output_attentions:
            # make sure that attn_weights are included in graph
            attn_weights_reshaped = attn_weights.view(bsz, self.num_heads, tgt_len, src_len)
            attn_weights = attn_weights_reshaped.view(bsz * self.num_heads, tgt_len, src_len)
        else:
            attn_weights_reshaped = None

        attn_probs = nn.functional.dropout(
            attn_weights,
            p=self.dropout,
            training=self.training,
        )

        assert v is not None
        attn_output = torch.bmm(attn_probs, v)
        assert attn_output.size() == (bsz * self.num_heads, tgt_len, self.head_dim)
        attn_output = attn_output.transpose(0, 1).contiguous().view(tgt_len, bsz, embed_dim)
        attn_output = self.out_proj(attn_output)

        return attn_output, attn_weights_reshaped

    def _use_saved_state(self, k, v, saved_state, key_padding_mask, static_kv, bsz):
        # saved states are stored with shape (bsz, num_heads, seq_len, head_dim)
        if "prev_key" in saved_state:
            _prev_key = saved_state["prev_key"]
            assert _prev_key is not None
            prev_key = _prev_key.view(bsz * self.num_heads, -1, self.head_dim)
            if static_kv:
                k = prev_key
            else:
                assert k is not None
                k = torch.cat([prev_key, k], dim=1)
        if "prev_value" in saved_state:
            _prev_value = saved_state["prev_value"]
            assert _prev_value is not None
            prev_value = _prev_value.view(bsz * self.num_heads, -1, self.head_dim)
            if static_kv:
                v = prev_value
            else:
                assert v is not None
                v = torch.cat([prev_value, v], dim=1)
        assert k is not None and v is not None
        prev_key_padding_mask: Optional[Tensor] = saved_state.get("prev_key_padding_mask", None)
        if prev_key_padding_mask is not None:
            if static_kv:
                new_key_padding_mask = prev_key_padding_mask
            else:
                new_key_padding_mask = torch.cat([prev_key_padding_mask, key_padding_mask], dim=1)
        else:
            new_key_padding_mask = key_padding_mask
        return k, v, new_key_padding_mask


def fill_with_neg_inf(t):
    """FP16-compatible function that fills a input_ids with -inf."""
    return t.float().fill_(float("-inf")).type_as(t)


# Public API
def _get_shape(t):
    return getattr(t, "shape", None)


@add_start_docstrings(
    "The bare FSMT Model outputting raw hidden-states without any specific head on top.",
    FSMT_START_DOCSTRING,
)
class FSMTModel(PretrainedFSMTModel):
    def __init__(self, config: FSMTConfig):
        super().__init__(config)

        padding_idx = config.pad_token_id
        encoder_embed_tokens = nn.Embedding(config.src_vocab_size, config.d_model, padding_idx)
        decoder_embed_tokens = nn.Embedding(config.tgt_vocab_size, config.d_model, padding_idx)

        self.encoder = FSMTEncoder(config, encoder_embed_tokens)
        self.decoder = FSMTDecoder(config, decoder_embed_tokens)

        self.init_weights()

    @add_start_docstrings_to_model_forward(FSMT_INPUTS_DOCSTRING)
    @add_code_sample_docstrings(
        tokenizer_class=_TOKENIZER_FOR_DOC,
        checkpoint=_CHECKPOINT_FOR_DOC,
        output_type=Seq2SeqModelOutput,
        config_class=_CONFIG_FOR_DOC,
    )
    def forward(
        self,
        input_ids,
        attention_mask=None,
        decoder_input_ids=None,
        decoder_attention_mask=None,
        head_mask=None,
        decoder_head_mask=None,
        cross_attn_head_mask=None,
        encoder_outputs: Optional[Tuple] = None,
        past_key_values=None,
        use_cache=None,
        output_attentions=None,
        output_hidden_states=None,
        return_dict=None,
    ):
        if decoder_input_ids is None:
            use_cache = False

        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

        # make masks if user doesn't supply
        if not use_cache:
            decoder_input_ids, decoder_padding_mask, causal_mask = _prepare_fsmt_decoder_inputs(
                self.config,
                input_ids,
                decoder_input_ids=decoder_input_ids,
                decoder_padding_mask=decoder_attention_mask,
                causal_mask_dtype=self.decoder.embed_tokens.weight.dtype,
            )
        else:
            decoder_padding_mask, causal_mask = None, None

        assert decoder_input_ids is not None

        if encoder_outputs is None:
            encoder_outputs = self.encoder(
                input_ids=input_ids,
                attention_mask=attention_mask,
                head_mask=head_mask,
                output_attentions=output_attentions,
                output_hidden_states=output_hidden_states,
                return_dict=return_dict,
            )
        # If the user passed a tuple for encoder_outputs, we wrap it in a BaseModelOutput when return_dict=False
        elif return_dict and not isinstance(encoder_outputs, BaseModelOutput):
            encoder_outputs = BaseModelOutput(
                last_hidden_state=encoder_outputs[0],
                hidden_states=encoder_outputs[1] if len(encoder_outputs) > 1 else None,
                attentions=encoder_outputs[2] if len(encoder_outputs) > 2 else None,
            )

        # decoder outputs consists of (dec_features, layer_state, dec_hidden, dec_attn)
        decoder_outputs = self.decoder(
            decoder_input_ids,
            encoder_outputs[0],
            attention_mask,
            decoder_padding_mask,
            decoder_causal_mask=causal_mask,
            head_mask=decoder_head_mask,
            cross_attn_head_mask=cross_attn_head_mask,
            past_key_values=past_key_values,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        if not return_dict:
            return decoder_outputs + encoder_outputs

        return Seq2SeqModelOutput(
            last_hidden_state=decoder_outputs.last_hidden_state,
            past_key_values=decoder_outputs.past_key_values,
            decoder_hidden_states=decoder_outputs.hidden_states,
            decoder_attentions=decoder_outputs.attentions,
            cross_attentions=decoder_outputs.cross_attentions,
            encoder_last_hidden_state=encoder_outputs.last_hidden_state,
            encoder_hidden_states=encoder_outputs.hidden_states,
            encoder_attentions=encoder_outputs.attentions,
        )

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

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

    def get_output_embeddings(self):
        return self.decoder.embed_tokens

    def set_output_embeddings(self, value):
        self.decoder.embed_tokens = value


@add_start_docstrings(
    "The FSMT Model with a language modeling head. Can be used for summarization.", FSMT_START_DOCSTRING
)
class FSMTForConditionalGeneration(PretrainedFSMTModel):
    base_model_prefix = "model"
    _keys_to_ignore_on_load_missing = [
        "model.encoder.embed_positions.weight",
        "model.decoder.embed_positions.weight",
    ]
    _keys_to_ignore_on_save = [
        "model.encoder.embed_positions.weight",
        "model.decoder.embed_positions.weight",
    ]

    def __init__(self, config: FSMTConfig):
        super().__init__(config)
        base_model = FSMTModel(config)
        self.model = base_model

    @add_start_docstrings_to_model_forward(FSMT_INPUTS_DOCSTRING)
    @replace_return_docstrings(output_type=Seq2SeqLMOutput, config_class=_CONFIG_FOR_DOC)
    @add_end_docstrings(FSMT_GENERATION_EXAMPLE)
    def forward(
        self,
        input_ids,
        attention_mask=None,
        decoder_input_ids=None,
        decoder_attention_mask=None,
        head_mask=None,
        decoder_head_mask=None,
        cross_attn_head_mask=None,
        encoder_outputs=None,
        past_key_values=None,
        labels=None,
        use_cache=None,
        output_attentions=None,
        output_hidden_states=None,
        return_dict=None,
    ):
        r"""
        labels (:obj:`torch.LongTensor` of shape :obj:`(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:

        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        if labels is not None:
            use_cache = False

        outputs = self.model(
            input_ids,
            attention_mask=attention_mask,
            decoder_input_ids=decoder_input_ids,
            encoder_outputs=encoder_outputs,
            decoder_attention_mask=decoder_attention_mask,
            head_mask=head_mask,
            decoder_head_mask=decoder_head_mask,
            cross_attn_head_mask=cross_attn_head_mask,
            past_key_values=past_key_values,
            use_cache=use_cache,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )
        lm_logits = outputs[0]

        masked_lm_loss = None
        if labels is not None:
            loss_fct = CrossEntropyLoss()
            # TODO(SS): do we need to ignore pad tokens in labels?
            masked_lm_loss = loss_fct(lm_logits.view(-1, self.config.tgt_vocab_size), labels.view(-1))

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

        return Seq2SeqLMOutput(
            loss=masked_lm_loss,
            logits=lm_logits,
            past_key_values=outputs.past_key_values,
            decoder_hidden_states=outputs.decoder_hidden_states,
            decoder_attentions=outputs.decoder_attentions,
            cross_attentions=outputs.cross_attentions,
            encoder_last_hidden_state=outputs.encoder_last_hidden_state,
            encoder_hidden_states=outputs.encoder_hidden_states,
            encoder_attentions=outputs.encoder_attentions,
        )

    def prepare_inputs_for_generation(
        self,
        decoder_input_ids,
        past=None,
        attention_mask=None,
        head_mask=None,
        decoder_head_mask=None,
        cross_attn_head_mask=None,
        use_cache=None,
        encoder_outputs=None,
        **kwargs
    ):
        return {
            "input_ids": None,  # encoder_outputs is defined. input_ids not needed
            "encoder_outputs": encoder_outputs,
            "past_key_values": past,
            "decoder_input_ids": decoder_input_ids,
            "attention_mask": attention_mask,
            "head_mask": head_mask,
            "decoder_head_mask": decoder_head_mask,
            "cross_attn_head_mask": cross_attn_head_mask,
            "use_cache": use_cache,  # change this to avoid caching (presumably for debugging)
        }

    def prepare_decoder_input_ids_from_labels(self, labels: torch.Tensor):
        return shift_tokens_right(labels, self.config.pad_token_id)

    @staticmethod
    def _reorder_cache(past, beam_idx):
        reordered_past = []
        for layer_past in past:
            # get the correct batch idx from decoder layer's batch dim for cross and self-attn
            layer_past_new = {
                attn_key: _reorder_buffer(attn_cache, beam_idx) for attn_key, attn_cache in layer_past.items()
            }
            reordered_past.append(layer_past_new)
        return reordered_past

    def get_encoder(self):
        return self.model.encoder

    def get_output_embeddings(self):
        return self.model.decoder.embed_tokens

    def set_output_embeddings(self, value):
        self.model.decoder.embed_tokens = value


class SinusoidalPositionalEmbedding(nn.Embedding):
    """
    This module produces sinusoidal positional embeddings of any length.

    We don't want to save the weight of this embedding since it's not trained (deterministic) and it can be huge.

    Padding symbols are ignored.

    These embeddings get automatically extended in forward if more positions is needed.
    """

    def __init__(self, num_positions, embedding_dim, padding_idx):
        self.make_weight(num_positions, embedding_dim, padding_idx)

    def make_weight(self, num_positions, embedding_dim, padding_idx):
        weight = self.get_embedding(num_positions, embedding_dim, padding_idx)
        if not hasattr(self, "weight"):
            # in ___init__
            super().__init__(num_positions, embedding_dim, padding_idx, _weight=weight)
        else:
            # in forward
            weight = weight.to(self.weight.device)
            self.weight = nn.Parameter(weight)
        self.weight.detach_()
        self.weight.requires_grad = False

    @staticmethod
    def get_embedding(num_embeddings, embedding_dim, padding_idx):
        """
        Build sinusoidal embeddings.

        This matches the implementation in tensor2tensor, but differs slightly from the description in Section 3.5 of
        "Attention Is All You Need".
        """
        half_dim = embedding_dim // 2
        emb = math.log(10000) / (half_dim - 1)
        emb = torch.exp(torch.arange(half_dim, dtype=torch.float) * -emb)
        emb = torch.arange(num_embeddings, dtype=torch.float).unsqueeze(1) * emb.unsqueeze(0)
        emb = torch.cat([torch.sin(emb), torch.cos(emb)], dim=1).view(num_embeddings, -1)
        if embedding_dim % 2 == 1:
            # zero pad
            emb = torch.cat([emb, torch.zeros(num_embeddings, 1)], dim=1)
        if padding_idx is not None:
            emb[padding_idx, :] = 0
        return emb

    @staticmethod
    def make_positions(tensor, padding_idx: int):
        """
        Replace non-padding symbols with their position numbers.

        Position numbers begin at padding_idx+1. Padding symbols are ignored.
        """
        # The series of casts and type-conversions here are carefully
        # balanced to both work with ONNX export and XLA. In particular XLA
        # prefers ints, cumsum defaults to output longs, and ONNX doesn't know
        # how to handle the dtype kwarg in cumsum.
        mask = tensor.ne(padding_idx).int()
        return (torch.cumsum(mask, dim=1).type_as(mask) * mask).long() + padding_idx

    def forward(
        self,
        input,
        incremental_state: Optional[Any] = None,
        timestep: Optional[Tensor] = None,
    ):
        """Input is expected to be of size [bsz x seqlen]."""
        bsz, seq_len = input.shape[:2]
        max_pos = self.padding_idx + 1 + seq_len
        if max_pos > self.weight.size(0):
            # expand embeddings if needed
            self.make_weight(max_pos, self.embedding_dim, self.padding_idx)
        positions = self.make_positions(input, self.padding_idx)
        return super().forward(positions)