transformers documentation

Reformer

# Reformer

DISCLAIMER: This model is still a work in progress, if you see something strange, file a Github Issue.

## Overview

The Reformer model was proposed in the paper Reformer: The Efficient Transformer by Nikita Kitaev, Łukasz Kaiser, Anselm Levskaya.

The abstract from the paper is the following:

Large Transformer models routinely achieve state-of-the-art results on a number of tasks but training these models can be prohibitively costly, especially on long sequences. We introduce two techniques to improve the efficiency of Transformers. For one, we replace dot-product attention by one that uses locality-sensitive hashing, changing its complexity from O(L^2) to O(Llog(L)), where L is the length of the sequence. Furthermore, we use reversible residual layers instead of the standard residuals, which allows storing activations only once in the training process instead of N times, where N is the number of layers. The resulting model, the Reformer, performs on par with Transformer models while being much more memory-efficient and much faster on long sequences.

This model was contributed by patrickvonplaten. The Authors’ code can be found here.

Note:

• Reformer does not work with torch.nn.DataParallel due to a bug in PyTorch, see issue #36035

## Axial Positional Encodings

Axial Positional Encodings were first implemented in Google’s trax library and developed by the authors of this model’s paper. In models that are treating very long input sequences, the conventional position id encodings store an embedings vector of size $d$ being the config.hidden_size for every position $i, \ldots, n_s$, with $n_s$ being config.max_embedding_size. This means that having a sequence length of $n_s = 2^{19} \approx 0.5M$ and a config.hidden_size of $d = 2^{10} \approx 1000$ would result in a position encoding matrix: $X_{i,j}, \text{ with } i \in \left[1,\ldots, d\right] \text{ and } j \in \left[1,\ldots, n_s\right]$

which alone has over 500M parameters to store. Axial positional encodings factorize $X_{i,j}$ into two matrices: $X^{1}_{i,j}, \text{ with } i \in \left[1,\ldots, d^1\right] \text{ and } j \in \left[1,\ldots, n_s^1\right]$

and $X^{2}_{i,j}, \text{ with } i \in \left[1,\ldots, d^2\right] \text{ and } j \in \left[1,\ldots, n_s^2\right]$

with: $d = d^1 + d^2 \text{ and } n_s = n_s^1 \times n_s^2 .$

Therefore the following holds: $X_{i,j} = \begin{cases} X^{1}_{i, k}, & \text{if }\ i < d^1 \text{ with } k = j \mod n_s^1 \\ X^{2}_{i - d^1, l}, & \text{if } i \ge d^1 \text{ with } l = \lfloor\frac{j}{n_s^1}\rfloor \end{cases}$

Intuitively, this means that a position embedding vector $x_j \in \mathbb{R}^{d}$ is now the composition of two factorized embedding vectors: $x^1_{k, l} + x^2_{l, k}$, where as the config.max_embedding_size dimension $j$ is factorized into $k \text{ and } l$. This design ensures that each position embedding vector $x_j$ is unique.

Using the above example again, axial position encoding with $d^1 = 2^5, d^2 = 2^5, n_s^1 = 2^9, n_s^2 = 2^{10}$ can drastically reduced the number of parameters to $2^{14} + 2^{15} \approx 49000$ parameters.

In practice, the parameter config.axial_pos_embds_dim is set to a tuple $(d^1, d^2)$ which sum has to be equal to config.hidden_size and config.axial_pos_shape is set to a tuple $(n_s^1, n_s^2)$ which product has to be equal to config.max_embedding_size, which during training has to be equal to the sequence length of the input_ids.

## LSH Self Attention

In Locality sensitive hashing (LSH) self attention the key and query projection weights are tied. Therefore, the key query embedding vectors are also tied. LSH self attention uses the locality sensitive hashing mechanism proposed in Practical and Optimal LSH for Angular Distance to assign each of the tied key query embedding vectors to one of config.num_buckets possible buckets. The premise is that the more “similar” key query embedding vectors (in terms of cosine similarity) are to each other, the more likely they are assigned to the same bucket.

The accuracy of the LSH mechanism can be improved by increasing config.num_hashes or directly the argument num_hashes of the forward function so that the output of the LSH self attention better approximates the output of the “normal” full self attention. The buckets are then sorted and chunked into query key embedding vector chunks each of length config.lsh_chunk_length. For each chunk, the query embedding vectors attend to its key vectors (which are tied to themselves) and to the key embedding vectors of config.lsh_num_chunks_before previous neighboring chunks and config.lsh_num_chunks_after following neighboring chunks.

For more information, see the original Paper or this great blog post.

Note that config.num_buckets can also be factorized into a list $(n_{\text{buckets}}^1, n_{\text{buckets}}^2)$. This way instead of assigning the query key embedding vectors to one of $(1,\ldots, n_{\text{buckets}})$ they are assigned to one of $(1-1,\ldots, n_{\text{buckets}}^1-1, \ldots, 1-n_{\text{buckets}}^2, \ldots, n_{\text{buckets}}^1-n_{\text{buckets}}^2)$. This is crucial for very long sequences to save memory.

When training a model from scratch, it is recommended to leave config.num_buckets=None, so that depending on the sequence length a good value for num_buckets is calculated on the fly. This value will then automatically be saved in the config and should be reused for inference.

Using LSH self attention, the memory and time complexity of the query-key matmul operation can be reduced from $\mathcal{O}(n_s \times n_s)$ to $\mathcal{O}(n_s \times \log(n_s))$, which usually represents the memory and time bottleneck in a transformer model, with $n_s$ being the sequence length.

## Local Self Attention

Local self attention is essentially a “normal” self attention layer with key, query and value projections, but is chunked so that in each chunk of length config.local_chunk_length the query embedding vectors only attends to the key embedding vectors in its chunk and to the key embedding vectors of config.local_num_chunks_before previous neighboring chunks and config.local_num_chunks_after following neighboring chunks.

Using Local self attention, the memory and time complexity of the query-key matmul operation can be reduced from $\mathcal{O}(n_s \times n_s)$ to $\mathcal{O}(n_s \times \log(n_s))$, which usually represents the memory and time bottleneck in a transformer model, with $n_s$ being the sequence length.

## Training

During training, we must ensure that the sequence length is set to a value that can be divided by the least common multiple of config.lsh_chunk_length and config.local_chunk_length and that the parameters of the Axial Positional Encodings are correctly set as described above. Reformer is very memory efficient so that the model can easily be trained on sequences as long as 64000 tokens.

For training, the ReformerModelWithLMHead should be used as follows:

input_ids = tokenizer.encode("This is a sentence from the training data", return_tensors="pt")
loss = model(input_ids, labels=input_ids)[0]

## ReformerConfig

class transformers.ReformerConfig < >

( attention_head_size = 64 attn_layers = ['local', 'lsh', 'local', 'lsh', 'local', 'lsh'] axial_norm_std = 1.0 axial_pos_embds = True axial_pos_shape = [64, 64] axial_pos_embds_dim = [64, 192] chunk_size_lm_head = 0 eos_token_id = 2 feed_forward_size = 512 hash_seed = None hidden_act = 'relu' hidden_dropout_prob = 0.05 hidden_size = 256 initializer_range = 0.02 is_decoder = False layer_norm_eps = 1e-12 local_num_chunks_before = 1 local_num_chunks_after = 0 local_attention_probs_dropout_prob = 0.05 local_attn_chunk_length = 64 lsh_attn_chunk_length = 64 lsh_attention_probs_dropout_prob = 0.0 lsh_num_chunks_before = 1 lsh_num_chunks_after = 0 max_position_embeddings = 4096 num_attention_heads = 12 num_buckets = None num_hashes = 1 pad_token_id = 0 vocab_size = 320 tie_word_embeddings = False use_cache = True classifier_dropout = None **kwargs )

Parameters

• attention_head_size (int, optional, defaults to 64) — Dimensionality of the projected key, query and value vectors
• attn_layers (List[str], optional, defaults to ["local", "lsh", "local", "lsh", "local", "lsh"]) — List of attention layer types in ascending order. It can be chosen between a LSHSelfAttention layer ("lsh") and a LocalSelfAttention layer ("local").

• axial_pos_embds (bool, optional, defaults to True) — Whether or not to use axial position embeddings. For more information on how axial position embeddings work, see Axial Position Encodings.
• axial_norm_std (float, optional, defaults to 1.0) — The standard deviation of the normal_initializer for initializing the weight matrices of the axial positional encodings.
• axial_pos_shape (List[int], optional, defaults to [64, 64]) — The position dims of the axial position encodings. During training, the product of the position dims has to be equal to the sequence length.

For more information on how axial position embeddings work, see Axial Position Encodings.

• axial_pos_embds_dim (List[int], optional, defaults to [64, 192]) — The embedding dims of the axial position encodings. The sum of the embedding dims has to be equal to the hidden size.

For more information on how axial position embeddings work, see Axial Position Encodings.

• chunk_size_lm_head (int, optional, defaults to 0) — The chunk size of the final language model feed forward head layer. A chunk size of 0 means that the feed forward layer is not chunked. A chunk size of n means that the feed forward layer processes n < sequence_length embeddings at a time.

For more information on feed forward chunking, see How does Feed Forward Chunking work?.

• eos_token_id (int, optional, defaults to 2) — The token id for the end-of-sentence token.
• feed_forward_size (int, optional, defaults to 512) — Dimensionality of the feed_forward layer in the residual attention block.
• hash_seed (int, optional) — Seed that can be used to make local sensitive hashing in LSHSelfAttention deterministic. This should only be set for testing purposed. For evaluation and training purposes hash_seed should be left as None to ensure fully random rotations in local sensitive hashing scheme.
• hidden_act (str or Callable, optional, defaults to "relu") — The non-linear activation function (function or string) in the feed forward layer in the residual attention block. If string, "gelu", "relu", "silu" and "gelu_new" are supported.
• hidden_dropout_prob (float, optional, defaults to 0.05) — The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
• hidden_size (int, optional, defaults to 256) — Dimensionality of the output hidden states of the residual attention blocks.
• initializer_range (float, optional, defaults to 0.02) — The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
• is_decoder (bool, optional, defaults to False) — Whether or not to use a causal mask in addition to the attention_mask passed to ReformerModel. When using the Reformer for causal language modeling, this argument should be set to True.
• layer_norm_eps (float, optional, defaults to 1e-12) — The epsilon used by the layer normalization layers.
• local_chunk_length (int, optional, defaults to 64) — Length of chunk which attends to itself in LocalSelfAttention. Chunking reduces memory complexity from sequence length x sequence length (self attention) to chunk length x chunk length x sequence length / chunk length (chunked self attention).
• local_num_chunks_before (int, optional, defaults to 1) — Number of previous neighbouring chunks to attend to in LocalSelfAttention layer to itself.
• local_num_chunks_after (int, optional, defaults to 0) — Number of following neighbouring chunks to attend to in LocalSelfAttention layer in addition to itself.
• local_attention_probs_dropout_prob (float, optional, defaults to 0.1) — The dropout ratio for the attention probabilities in LocalSelfAttention.
• lsh_attn_chunk_length (int, optional, defaults to 64) — Length of chunk which attends to itself in LSHSelfAttention. Chunking reduces memory complexity from sequence length x sequence length (self attention) to chunk length x chunk length x sequence length / chunk length (chunked self attention).
• lsh_num_chunks_before (int, optional, defaults to 1) — Number of previous neighbouring chunks to attend to in LSHSelfAttention layer to itself.
• lsh_num_chunks_after (int, optional, defaults to 0) — Number of following neighbouring chunks to attend to in LSHSelfAttention layer to itself.
• lsh_attention_probs_dropout_prob (float, optional, defaults to 0.1) — The dropout ratio for the attention probabilities in LSHSelfAttention.
• max_position_embeddings (int, optional, defaults to 4096) — The maximum sequence length that this model might ever be used with. Typically set this to something large just in case (e.g., 512 or 1024 or 2048).
• num_attention_heads (int, optional, defaults to 12) — Number of attention heads for each attention layer in the Transformer encoder.
• num_buckets (int or List[int], optional) — Number of buckets, the key query vectors can be “hashed into” using the locality sensitive hashing scheme. Each query key vector is hashed into a hash in 1, ..., num_buckets. The number of buckets can also be factorized into a list for improved memory complexity. In this case, each query key vector is hashed into a hash in 1-1, 1-2, ..., num_buckets[0]-1, ..., num_buckets[0]-num_buckets[1] if num_buckets is factorized into two factors. The number of buckets (or the product the factors) should approximately equal sequence length / lsh_chunk_length. If num_buckets not set, a good value is calculated on the fly.
• num_hashes (int, optional, defaults to 1) — Number of hashing rounds (e.g., number of random rotations) in Local Sensitive Hashing scheme. The higher num_hashes, the more accurate the LSHSelfAttention becomes, but also the more memory and time intensive the hashing becomes.
• pad_token_id (int, optional, defaults to 0) — The token id for the padding token.
• vocab_size (int, optional, defaults to 320) —\ Vocabulary size of the Reformer model. Defines the number of different tokens that can be represented by the inputs_ids passed when calling ReformerModel.
• tie_word_embeddings (bool, optional, defaults to False) — Whether to tie input and output embeddings.
• use_cache (bool, optional, defaults to True) — Whether or not the model should return the last key/values attentions (not used by all models).
• classifier_dropout (float, optional) — The dropout ratio for the classification head.

This is the configuration class to store the configuration of a ReformerModel. It is used to instantiate a Reformer model according to the specified arguments, defining the model architecture.

Configuration objects inherit from PretrainedConfig and can be used to control the model outputs. Read the documentation from PretrainedConfig for more information.

Examples:

>>> from transformers import ReformerModel, ReformerConfig

>>> # Initializing a Reformer configuration
>>> configuration = ReformerConfig()

>>> # Initializing a Reformer model
>>> model = ReformerModel(configuration)

>>> # Accessing the model configuration
>>> configuration = model.config

## ReformerTokenizer

class transformers.ReformerTokenizer < >

( vocab_file eos_token = '</s>' unk_token = '<unk>' additional_special_tokens = [] sp_model_kwargs: typing.Union[typing.Dict[str, typing.Any], NoneType] = None **kwargs )

Parameters

• vocab_file (str) — SentencePiece file (generally has a .spm extension) that contains the vocabulary necessary to instantiate a tokenizer.
• eos_token (str, optional, defaults to "</s>") — The end of sequence token.

When building a sequence using special tokens, this is not the token that is used for the end of sequence. The token used is the sep_token.

• unk_token (str, optional, defaults to "<unk>") — The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this token instead.
• pad_token (str, optional, defaults to "<pad>") — The token used for padding, for example when batching sequences of different lengths.
• additional_special_tokens (List[str], optional) — Additional special tokens used by the tokenizer.
• sp_model_kwargs (dict, optional) — Will be passed to the SentencePieceProcessor.__init__() method. The Python wrapper for SentencePiece can be used, among other things, to set:

• enable_sampling: Enable subword regularization.

• nbest_size: Sampling parameters for unigram. Invalid for BPE-Dropout.

• nbest_size = {0,1}: No sampling is performed.
• nbest_size > 1: samples from the nbest_size results.
• nbest_size < 0: assuming that nbest_size is infinite and samples from the all hypothesis (lattice) using forward-filtering-and-backward-sampling algorithm.
• alpha: Smoothing parameter for unigram sampling, and dropout probability of merge operations for BPE-dropout.

Construct a Reformer tokenizer. Based on SentencePiece .

This tokenizer inherits from PreTrainedTokenizer which contains most of the main methods. Users should refer to this superclass for more information regarding those methods.

## ReformerTokenizerFast

class transformers.ReformerTokenizerFast < >

( vocab_file = None tokenizer_file = None eos_token = '</s>' unk_token = '<unk>' additional_special_tokens = [] **kwargs )

Parameters

• vocab_file (str) — SentencePiece file (generally has a .spm extension) that contains the vocabulary necessary to instantiate a tokenizer.
• eos_token (str, optional, defaults to "</s>") — The end of sequence token.

When building a sequence using special tokens, this is not the token that is used for the end of sequence. The token used is the sep_token.

• unk_token (str, optional, defaults to "<unk>") — The unknown token. A token that is not in the vocabulary cannot be converted to an ID and is set to be this token instead.
• pad_token (str, optional, defaults to "<pad>") — The token used for padding, for example when batching sequences of different lengths.
• additional_special_tokens (List[str], optional) — Additional special tokens used by the tokenizer.

Construct a “fast” Reformer tokenizer (backed by HuggingFace’s tokenizers library). Based on Unigram.

This tokenizer inherits from PreTrainedTokenizerFast which contains most of the main methods. Users should refer to this superclass for more information regarding those methods.

## ReformerModel

class transformers.ReformerModel < >

( config )

Parameters

• config (ReformerConfig) — 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 from_pretrained() method to load the model weights.

The bare Reformer Model transformer outputting raw hidden-stateswithout any specific head on top. Reformer was proposed in Reformer: The Efficient Transformer by Nikita Kitaev, Łukasz Kaiser, Anselm Levskaya.

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

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward < >

( input_ids = None attention_mask = None position_ids = None head_mask = None inputs_embeds = None num_hashes = None past_buckets_states = None use_cache = None output_hidden_states = None output_attentions = None return_dict = None ) transformers.models.reformer.modeling_reformer.ReformerModelOutputor tuple(torch.FloatTensor)

Parameters

• input_ids (torch.LongTensor of shape (batch_size, sequence_length)) — Indices of input sequence tokens in the vocabulary. During training the input_ids sequence_length has to be a multiple of the relevant model’s chunk lengths (lsh’s, local’s or both). During evaluation, the indices are automatically padded to be a multiple of the chunk length.

Indices can be obtained using ReformerTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.

• attention_mask (torch.FloatTensor of shape (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.
• position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

• head_mask (torch.FloatTensor of shape (num_heads,) or (num_layers, num_heads), optional) — Mask to nullify selected heads of the self-attention modules. Mask values selected in [0, 1]:

• inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) — Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the model’s internal embedding lookup matrix.
• num_hashes (int, optional) — The number of hashing rounds that should be performed during bucketing. Setting this argument overwrites the default defined in config.num_hashes.

For more information, see num_hashes in ReformerConfig.

• past_buckets_states (List[Tuple(torch.LongTensor, torch.FloatTensor)], optional) — List of Tuple(torch.LongTensor, torch.FloatTensor of length config.n_layers, with the first element being the previous buckets of shape (batch_size, num_heads, num_hashes, sequence_length)) and the second being the previous hidden_states of shape (batch_size, sequence_length, hidden_size)).

Contains precomputed hidden-states and buckets (only relevant for LSH Self-Attention). Can be used to speed up sequential decoding.

• use_cache (bool, optional) — If set to True, past_key_values key value states are returned and can be used to speed up decoding (see past_key_values).
• output_attentions (bool, optional) — Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.
• output_hidden_states (bool, optional) — Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.
• return_dict (bool, optional) — Whether or not to return a ModelOutput instead of a plain tuple.

Returns

transformers.models.reformer.modeling_reformer.ReformerModelOutputor tuple(torch.FloatTensor)

A transformers.models.reformer.modeling_reformer.ReformerModelOutputor a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (ReformerConfig) and inputs.

• last_hidden_state (torch.FloatTensor of shape (batch_size, num_predict, hidden_size)) — Sequence of hidden-states at the last layer of the model.

num_predict corresponds to target_mapping.shape[1]. If target_mapping is None, then num_predict corresponds to sequence_length.

• past_buckets_states (List[Tuple(torch.LongTensor, torch.FloatTensor)], optional, returned when use_cache=True is passed or when config.use_cache=True) — List of Tuple(torch.LongTensor, torch.FloatTensor of length config.n_layers, with the first element being the previous buckets of shape (batch_size, num_heads, num_hashes, sequence_length)) and the second being the previous hidden_states of shape (batch_size, sequence_length, hidden_size)).

Contains precomputed buckets and hidden-states that can be used (see past_buckets_states input) to speed up sequential decoding.

• hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings and one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

Hidden-states of the model at the output of each layer plus the initial embedding outputs.

• attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

The ReformerModel forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Example:

>>> from transformers import ReformerTokenizer, ReformerModel
>>> import torch

>>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
>>> outputs = model(**inputs)

>>> last_hidden_states = outputs.last_hidden_state

( config )

Parameters

• config (ReformerConfig) — 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 from_pretrained() method to load the model weights.

Reformer Model with a language modeling head on top. Reformer was proposed in Reformer: The Efficient Transformer by Nikita Kitaev, Łukasz Kaiser, Anselm Levskaya.

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

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward < >

( input_ids = None position_ids = None attention_mask = None head_mask = None inputs_embeds = None num_hashes = None past_buckets_states = None use_cache = None output_hidden_states = None output_attentions = None return_dict = None labels = None ) transformers.modeling_outputs.CausalLMOutput or tuple(torch.FloatTensor)

Parameters

• input_ids (torch.LongTensor of shape (batch_size, sequence_length)) — Indices of input sequence tokens in the vocabulary. During training the input_ids sequence_length has to be a multiple of the relevant model’s chunk lengths (lsh’s, local’s or both). During evaluation, the indices are automatically padded to be a multiple of the chunk length.

Indices can be obtained using ReformerTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.

• attention_mask (torch.FloatTensor of shape (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.
• position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

• head_mask (torch.FloatTensor of shape (num_heads,) or (num_layers, num_heads), optional) — Mask to nullify selected heads of the self-attention modules. Mask values selected in [0, 1]:

• inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) — Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the model’s internal embedding lookup matrix.
• num_hashes (int, optional) — The number of hashing rounds that should be performed during bucketing. Setting this argument overwrites the default defined in config.num_hashes.

For more information, see num_hashes in ReformerConfig.

• past_buckets_states (List[Tuple(torch.LongTensor, torch.FloatTensor)], optional) — List of Tuple(torch.LongTensor, torch.FloatTensor of length config.n_layers, with the first element being the previous buckets of shape (batch_size, num_heads, num_hashes, sequence_length)) and the second being the previous hidden_states of shape (batch_size, sequence_length, hidden_size)).

Contains precomputed hidden-states and buckets (only relevant for LSH Self-Attention). Can be used to speed up sequential decoding.

• use_cache (bool, optional) — If set to True, past_key_values key value states are returned and can be used to speed up decoding (see past_key_values).
• output_attentions (bool, optional) — Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.
• output_hidden_states (bool, optional) — Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.
• return_dict (bool, optional) — Whether or not to return a ModelOutput instead of a plain tuple.
• 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

transformers.modeling_outputs.CausalLMOutput or tuple(torch.FloatTensor)

A transformers.modeling_outputs.CausalLMOutput or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (ReformerConfig) and inputs.

• loss (torch.FloatTensor of shape (1,), optional, returned when labels is provided) — Language modeling loss (for next-token prediction).

• logits (torch.FloatTensor of shape (batch_size, sequence_length, config.vocab_size)) — Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).

• hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

Hidden-states of the model at the output of each layer plus the initial embedding outputs.

• attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

The ReformerModelWithLMHead forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Example:

>>> import torch
>>> from transformers import ReformerTokenizer, ReformerModelWithLMHead

>>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
>>> outputs = model(**inputs, labels=inputs["input_ids"])
>>> loss = outputs.loss
>>> logits = outputs.logits

( config )

Parameters

• config (ReformerConfig) — 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 from_pretrained() method to load the model weights.

Reformer Model with a language modeling head on top. Reformer was proposed in Reformer: The Efficient Transformer by Nikita Kitaev, Łukasz Kaiser, Anselm Levskaya.

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

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward < >

( input_ids = None position_ids = None attention_mask = None head_mask = None inputs_embeds = None num_hashes = None labels = None output_hidden_states = None output_attentions = None return_dict = None ) transformers.modeling_outputs.MaskedLMOutput or tuple(torch.FloatTensor)

Parameters

• input_ids (torch.LongTensor of shape (batch_size, sequence_length)) — Indices of input sequence tokens in the vocabulary. During training the input_ids sequence_length has to be a multiple of the relevant model’s chunk lengths (lsh’s, local’s or both). During evaluation, the indices are automatically padded to be a multiple of the chunk length.

Indices can be obtained using ReformerTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.

• attention_mask (torch.FloatTensor of shape (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.
• position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

• head_mask (torch.FloatTensor of shape (num_heads,) or (num_layers, num_heads), optional) — Mask to nullify selected heads of the self-attention modules. Mask values selected in [0, 1]:

• inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) — Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the model’s internal embedding lookup matrix.
• num_hashes (int, optional) — The number of hashing rounds that should be performed during bucketing. Setting this argument overwrites the default defined in config.num_hashes.

For more information, see num_hashes in ReformerConfig.

• past_buckets_states (List[Tuple(torch.LongTensor, torch.FloatTensor)], optional) — List of Tuple(torch.LongTensor, torch.FloatTensor of length config.n_layers, with the first element being the previous buckets of shape (batch_size, num_heads, num_hashes, sequence_length)) and the second being the previous hidden_states of shape (batch_size, sequence_length, hidden_size)).

Contains precomputed hidden-states and buckets (only relevant for LSH Self-Attention). Can be used to speed up sequential decoding.

• use_cache (bool, optional) — If set to True, past_key_values key value states are returned and can be used to speed up decoding (see past_key_values).
• output_attentions (bool, optional) — Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.
• output_hidden_states (bool, optional) — Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.
• return_dict (bool, optional) — Whether or not to return a ModelOutput instead of a plain tuple.
• labels (torch.LongTensor of shape (batch_size, sequence_length), optional) — Labels for computing the masked language modeling loss. Indices should be in [-100, 0, ..., config.vocab_size] (see input_ids docstring) Tokens with indices set to -100 are ignored (masked), the loss is only computed for the tokens with labels

Returns

transformers.modeling_outputs.MaskedLMOutput or tuple(torch.FloatTensor)

A transformers.modeling_outputs.MaskedLMOutput or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (ReformerConfig) and inputs.

• loss (torch.FloatTensor of shape (1,), optional, returned when labels is provided) — Masked language modeling (MLM) loss.

• logits (torch.FloatTensor of shape (batch_size, sequence_length, config.vocab_size)) — Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).

• hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

Hidden-states of the model at the output of each layer plus the initial embedding outputs.

• attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

The ReformerForMaskedLM forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Example:

>>> from transformers import ReformerTokenizer, ReformerForMaskedLM
>>> import torch

>>> inputs = tokenizer("The capital of France is [MASK].", return_tensors="pt")
>>> labels = tokenizer("The capital of France is Paris.", return_tensors="pt")["input_ids"]

>>> outputs = model(**inputs, labels=labels)
>>> loss = outputs.loss
>>> logits = outputs.logits

## ReformerForSequenceClassification

class transformers.ReformerForSequenceClassification < >

( config )

Parameters

• config (ReformerConfig) — 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 from_pretrained() method to load the model weights.

Reformer Model transformer with a sequence classification/regression head on top (a linear layer on top of the pooled output) e.g. for GLUE tasks.

Reformer was proposed in Reformer: The Efficient Transformer by Nikita Kitaev, Łukasz Kaiser, Anselm Levskaya.

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

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward < >

( input_ids = None position_ids = None attention_mask = None head_mask = None inputs_embeds = None num_hashes = None labels = None output_hidden_states = None output_attentions = None return_dict = None ) transformers.modeling_outputs.SequenceClassifierOutput or tuple(torch.FloatTensor)

Parameters

• input_ids (torch.LongTensor of shape (batch_size, sequence_length)) — Indices of input sequence tokens in the vocabulary. During training the input_ids sequence_length has to be a multiple of the relevant model’s chunk lengths (lsh’s, local’s or both). During evaluation, the indices are automatically padded to be a multiple of the chunk length.

Indices can be obtained using ReformerTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.

• attention_mask (torch.FloatTensor of shape (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.
• position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

• head_mask (torch.FloatTensor of shape (num_heads,) or (num_layers, num_heads), optional) — Mask to nullify selected heads of the self-attention modules. Mask values selected in [0, 1]:

• inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) — Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the model’s internal embedding lookup matrix.
• num_hashes (int, optional) — The number of hashing rounds that should be performed during bucketing. Setting this argument overwrites the default defined in config.num_hashes.

For more information, see num_hashes in ReformerConfig.

• past_buckets_states (List[Tuple(torch.LongTensor, torch.FloatTensor)], optional) — List of Tuple(torch.LongTensor, torch.FloatTensor of length config.n_layers, with the first element being the previous buckets of shape (batch_size, num_heads, num_hashes, sequence_length)) and the second being the previous hidden_states of shape (batch_size, sequence_length, hidden_size)).

Contains precomputed hidden-states and buckets (only relevant for LSH Self-Attention). Can be used to speed up sequential decoding.

• use_cache (bool, optional) — If set to True, past_key_values key value states are returned and can be used to speed up decoding (see past_key_values).
• output_attentions (bool, optional) — Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.
• output_hidden_states (bool, optional) — Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.
• return_dict (bool, optional) — Whether or not to return a ModelOutput instead of a plain tuple.
• labels (torch.LongTensor of shape (batch_size,), optional) — Labels for computing the sequence classification/regression loss. Indices should be in [0, ..., config.num_labels - 1]. If config.num_labels == 1 a regression loss is computed (Mean-Square loss), If config.num_labels > 1 a classification loss is computed (Cross-Entropy).

Returns

transformers.modeling_outputs.SequenceClassifierOutput or tuple(torch.FloatTensor)

A transformers.modeling_outputs.SequenceClassifierOutput or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (ReformerConfig) and inputs.

• loss (torch.FloatTensor of shape (1,), optional, returned when labels is provided) — Classification (or regression if config.num_labels==1) loss.

• logits (torch.FloatTensor of shape (batch_size, config.num_labels)) — Classification (or regression if config.num_labels==1) scores (before SoftMax).

• hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

Hidden-states of the model at the output of each layer plus the initial embedding outputs.

• attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

The ReformerForSequenceClassification forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Example of single-label classification:

>>> from transformers import ReformerTokenizer, ReformerForSequenceClassification
>>> import torch

>>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
>>> labels = torch.tensor([1]).unsqueeze(0)  # Batch size 1
>>> outputs = model(**inputs, labels=labels)
>>> loss = outputs.loss
>>> logits = outputs.logits

Example of multi-label classification:

>>> from transformers import ReformerTokenizer, ReformerForSequenceClassification
>>> import torch

>>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
>>> labels = torch.tensor([[1, 1]], dtype=torch.float)  # need dtype=float for BCEWithLogitsLoss
>>> outputs = model(**inputs, labels=labels)
>>> loss = outputs.loss
>>> logits = outputs.logits

( config )

Parameters

• config (ReformerConfig) — 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 from_pretrained() method to load the model weights.

Reformer Model with a span classification head on top for extractive question-answering tasks like SQuAD / TriviaQA ( a linear layer on top of hidden-states output to compute span start logits and span end logits.

Reformer was proposed in Reformer: The Efficient Transformer by Nikita Kitaev, Łukasz Kaiser, Anselm Levskaya.

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

This model is also a PyTorch torch.nn.Module subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior.

forward < >

( input_ids = None position_ids = None attention_mask = None head_mask = None inputs_embeds = None num_hashes = None start_positions = None end_positions = None output_hidden_states = None output_attentions = None return_dict = None ) transformers.modeling_outputs.QuestionAnsweringModelOutput or tuple(torch.FloatTensor)

Parameters

• input_ids (torch.LongTensor of shape (batch_size, sequence_length)) — Indices of input sequence tokens in the vocabulary. During training the input_ids sequence_length has to be a multiple of the relevant model’s chunk lengths (lsh’s, local’s or both). During evaluation, the indices are automatically padded to be a multiple of the chunk length.

Indices can be obtained using ReformerTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for details.

• attention_mask (torch.FloatTensor of shape (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.
• position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

• head_mask (torch.FloatTensor of shape (num_heads,) or (num_layers, num_heads), optional) — Mask to nullify selected heads of the self-attention modules. Mask values selected in [0, 1]:

• inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) — Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the model’s internal embedding lookup matrix.
• num_hashes (int, optional) — The number of hashing rounds that should be performed during bucketing. Setting this argument overwrites the default defined in config.num_hashes.

For more information, see num_hashes in ReformerConfig.

• past_buckets_states (List[Tuple(torch.LongTensor, torch.FloatTensor)], optional) — List of Tuple(torch.LongTensor, torch.FloatTensor of length config.n_layers, with the first element being the previous buckets of shape (batch_size, num_heads, num_hashes, sequence_length)) and the second being the previous hidden_states of shape (batch_size, sequence_length, hidden_size)).

Contains precomputed hidden-states and buckets (only relevant for LSH Self-Attention). Can be used to speed up sequential decoding.

• use_cache (bool, optional) — If set to True, past_key_values key value states are returned and can be used to speed up decoding (see past_key_values).
• output_attentions (bool, optional) — Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.
• output_hidden_states (bool, optional) — Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.
• return_dict (bool, optional) — Whether or not to return a ModelOutput instead of a plain tuple.
• start_positions (torch.LongTensor of shape (batch_size,), optional) — Labels for position (index) of the start of the labelled span for computing the token classification loss. Positions are clamped to the length of the sequence (sequence_length). Position outside of the sequence are not taken into account for computing the loss.
• end_positions (torch.LongTensor of shape (batch_size,), optional) — Labels for position (index) of the end of the labelled span for computing the token classification loss. Positions are clamped to the length of the sequence (sequence_length). Position outside of the sequence are not taken into account for computing the loss.

Returns

transformers.modeling_outputs.QuestionAnsweringModelOutput or tuple(torch.FloatTensor)

A transformers.modeling_outputs.QuestionAnsweringModelOutput or a tuple of torch.FloatTensor (if return_dict=False is passed or when config.return_dict=False) comprising various elements depending on the configuration (ReformerConfig) and inputs.

• loss (torch.FloatTensor of shape (1,), optional, returned when labels is provided) — Total span extraction loss is the sum of a Cross-Entropy for the start and end positions.

• start_logits (torch.FloatTensor of shape (batch_size, sequence_length)) — Span-start scores (before SoftMax).

• end_logits (torch.FloatTensor of shape (batch_size, sequence_length)) — Span-end scores (before SoftMax).

• hidden_states (tuple(torch.FloatTensor), optional, returned when output_hidden_states=True is passed or when config.output_hidden_states=True) — Tuple of torch.FloatTensor (one for the output of the embeddings + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

Hidden-states of the model at the output of each layer plus the initial embedding outputs.

• attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True is passed or when config.output_attentions=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, num_heads, sequence_length, sequence_length).

Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads.

The ReformerForQuestionAnswering forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

Example:

>>> from transformers import ReformerTokenizer, ReformerForQuestionAnswering
>>> import torch

>>> end_scores = outputs.end_logits