Transformers documentation

mT5

Join the Hugging Face community

to get started

# mT5

## Overview

The mT5 model was presented in mT5: A massively multilingual pre-trained text-to-text transformer by Linting Xue, Noah Constant, Adam Roberts, Mihir Kale, Rami Al-Rfou, Aditya Siddhant, Aditya Barua, Colin Raffel.

The abstract from the paper is the following:

The recent “Text-to-Text Transfer Transformer” (T5) leveraged a unified text-to-text format and scale to attain state-of-the-art results on a wide variety of English-language NLP tasks. In this paper, we introduce mT5, a multilingual variant of T5 that was pre-trained on a new Common Crawl-based dataset covering 101 languages. We detail the design and modified training of mT5 and demonstrate its state-of-the-art performance on many multilingual benchmarks. We also describe a simple technique to prevent “accidental translation” in the zero-shot setting, where a generative model chooses to (partially) translate its prediction into the wrong language. All of the code and model checkpoints used in this work are publicly available.

Note: mT5 was only pre-trained on mC4 excluding any supervised training. Therefore, this model has to be fine-tuned before it is usable on a downstream task, unlike the original T5 model. Since mT5 was pre-trained unsupervisedly, there’s no real advantage to using a task prefix during single-task fine-tuning. If you are doing multi-task fine-tuning, you should use a prefix.

Google has released the following variants:

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

## MT5Config

### class transformers.MT5Config

< >

( vocab_size = 250112 d_model = 512 d_kv = 64 d_ff = 1024 num_layers = 8 num_decoder_layers = None num_heads = 6 relative_attention_num_buckets = 32 relative_attention_max_distance = 128 dropout_rate = 0.1 layer_norm_epsilon = 1e-06 initializer_factor = 1.0 feed_forward_proj = 'gated-gelu' is_encoder_decoder = True use_cache = True tokenizer_class = 'T5Tokenizer' tie_word_embeddings = False pad_token_id = 0 eos_token_id = 1 decoder_start_token_id = 0 **kwargs )

Parameters

• vocab_size (int, optional, defaults to 250112) — Vocabulary size of the T5 model. Defines the number of different tokens that can be represented by the inputs_ids passed when calling T5Model or TFT5Model.
• d_model (int, optional, defaults to 512) — Size of the encoder layers and the pooler layer.
• d_kv (int, optional, defaults to 64) — Size of the key, query, value projections per attention head. d_kv has to be equal to d_model // num_heads.
• d_ff (int, optional, defaults to 1024) — Size of the intermediate feed forward layer in each T5Block.
• num_layers (int, optional, defaults to 8) — Number of hidden layers in the Transformer encoder.
• num_decoder_layers (int, optional) — Number of hidden layers in the Transformer decoder. Will use the same value as num_layers if not set.
• num_heads (int, optional, defaults to 6) — Number of attention heads for each attention layer in the Transformer encoder.
• relative_attention_num_buckets (int, optional, defaults to 32) — The number of buckets to use for each attention layer.
• relative_attention_max_distance (int, optional, defaults to 128) — The maximum distance of the longer sequences for the bucket separation.
• dropout_rate (float, optional, defaults to 0.1) — The ratio for all dropout layers.
• layer_norm_eps (float, optional, defaults to 1e-6) — The epsilon used by the layer normalization layers.
• initializer_factor (float, optional, defaults to 1) — A factor for initializing all weight matrices (should be kept to 1, used internally for initialization testing).
• feed_forward_proj (string, optional, defaults to "gated-gelu") — Type of feed forward layer to be used. Should be one of "relu" or "gated-gelu".
• use_cache (bool, optional, defaults to True) — Whether or not the model should return the last key/values attentions (not used by all models).

This is the configuration class to store the configuration of a MT5Model or a TFMT5Model. It is used to instantiate a mT5 model according to the specified arguments, defining the model architecture. Instantiating a configuration with the defaults will yield a similar configuration to that of the mT5 google/mt5-small architecture.

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

## MT5Tokenizer

### class transformers.T5Tokenizer

< >

( vocab_file eos_token = '</s>' unk_token = '<unk>' pad_token = '<pad>' extra_ids = 100 additional_special_tokens = None 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.
• extra_ids (int, optional, defaults to 100) — Add a number of extra ids added to the vocabulary for use as sentinels. These tokens are accessible as “id{%d}>” where ”{%d}” is a number between 0 and extra_ids-1. These tokens can be retrieved by calling get_sentinel_tokens method and token ids can be by calling get_sentinel_token_ids method 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.

• sp_model (SentencePieceProcessor) — The SentencePiece processor that is used for every conversion (string, tokens and IDs).

Construct a T5 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.

#### build_inputs_with_special_tokens

< >

( token_ids_0: typing.List[int] token_ids_1: typing.Optional[typing.List[int]] = None ) List[int]

Parameters

• token_ids_0 (List[int]) — List of IDs to which the special tokens will be added.
• token_ids_1 (List[int], optional) — Optional second list of IDs for sequence pairs.

Returns

List[int]

List of input IDs with the appropriate special tokens.

Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and adding special tokens. A sequence has the following format:

• single sequence: X </s>
• pair of sequences: A </s> B </s>

#### convert_tokens_to_string

< >

( tokens )

Converts a sequence of tokens (string) in a single string.

#### create_token_type_ids_from_sequences

< >

( token_ids_0: typing.List[int] token_ids_1: typing.Optional[typing.List[int]] = None ) List[int]

Parameters

• token_ids_0 (List[int]) — List of IDs.
• token_ids_1 (List[int], optional) — Optional second list of IDs for sequence pairs.

Returns

List[int]

List of zeros.

Create a mask from the two sequences passed to be used in a sequence-pair classification task. T5 does not make use of token type ids, therefore a list of zeros is returned.

< >

( token_ids_0: typing.List[int] token_ids_1: typing.Optional[typing.List[int]] = None already_has_special_tokens: bool = False ) List[int]

Parameters

• token_ids_0 (List[int]) — List of IDs.
• token_ids_1 (List[int], optional) — Optional second list of IDs for sequence pairs.
• already_has_special_tokens (bool, optional, defaults to False) — Whether or not the token list is already formatted with special tokens for the model.

Returns

List[int]

A list of integers in the range [0, 1]: 1 for a special token, 0 for a sequence token.

Retrieve sequence ids from a token list that has no special tokens added. This method is called when adding special tokens using the tokenizer prepare_for_model method.

See T5Tokenizer for all details.

## MT5TokenizerFast

### class transformers.T5TokenizerFast

< >

( vocab_file = None tokenizer_file = None eos_token = '</s>' unk_token = '<unk>' pad_token = '<pad>' extra_ids = 100 additional_special_tokens = 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.
• extra_ids (int, optional, defaults to 100) — Add a number of extra ids added to the vocabulary for use as sentinels. These tokens are accessible as “id{%d}>” where ”{%d}” is a number between 0 and extra_ids-1. These tokens can be retrieved by calling get_sentinel_tokens method and token ids can be by calling get_sentinel_token_ids method
• additional_special_tokens (List[str], optional) — Additional special tokens used by the tokenizer.

Construct a “fast” T5 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.

#### build_inputs_with_special_tokens

< >

( token_ids_0: typing.List[int] token_ids_1: typing.Optional[typing.List[int]] = None ) List[int]

Parameters

• token_ids_0 (List[int]) — List of IDs to which the special tokens will be added.
• token_ids_1 (List[int], optional) — Optional second list of IDs for sequence pairs.

Returns

List[int]

List of input IDs with the appropriate special tokens.

Build model inputs from a sequence or a pair of sequence for sequence classification tasks by concatenating and adding special tokens. A sequence has the following format:

• single sequence: X </s>
• pair of sequences: A </s> B </s>

#### create_token_type_ids_from_sequences

< >

( token_ids_0: typing.List[int] token_ids_1: typing.Optional[typing.List[int]] = None ) List[int]

Parameters

• token_ids_0 (List[int]) — List of IDs.
• token_ids_1 (List[int], optional) — Optional second list of IDs for sequence pairs.

Returns

List[int]

List of zeros.

Create a mask from the two sequences passed to be used in a sequence-pair classification task. T5 does not make use of token type ids, therefore a list of zeros is returned.

See T5TokenizerFast for all details.

## MT5Model

### class transformers.MT5Model

< >

( config: MT5Config )

Parameters

• config (MT5Config) — 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 MT5 Model transformer outputting raw hidden-states without any specific head on top.

The MT5 model was proposed in Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer by Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, Peter J. Liu. It’s an encoder decoder transformer pre-trained in a text-to-text denoising generative setting.

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.

Examples:

>>> from transformers import MT5Model, AutoTokenizer

>>> article = "UN Offizier sagt, dass weiter verhandelt werden muss in Syrien."
>>> summary = "Weiter Verhandlung in Syrien."
>>> inputs = tokenizer(article, return_tensors="pt")
>>> labels = tokenizer(text_target=summary, return_tensors="pt")

>>> outputs = model(input_ids=inputs["input_ids"], decoder_input_ids=labels["input_ids"])
>>> hidden_states = outputs.last_hidden_state

#### deparallelize

< >

( )

Moves the model to cpu from a model parallel state.

Example:

# On a 4 GPU machine with mt5-xl:
model = MT5ForConditionalGeneration.from_pretrained("Mt5-xl")
device_map = {
0: [0, 1, 2],
1: [3, 4, 5, 6, 7, 8, 9],
2: [10, 11, 12, 13, 14, 15, 16],
3: [17, 18, 19, 20, 21, 22, 23],
}
model.parallelize(device_map)  # Splits the model across several devices
model.deparallelize()  # Put the model back on cpu and cleans memory by calling torch.cuda.empty_cache()

#### forward

< >

( input_ids: typing.Optional[torch.LongTensor] = None attention_mask: typing.Optional[torch.FloatTensor] = None decoder_input_ids: typing.Optional[torch.LongTensor] = None decoder_attention_mask: typing.Optional[torch.BoolTensor] = None head_mask: typing.Optional[torch.FloatTensor] = None decoder_head_mask: typing.Optional[torch.FloatTensor] = None cross_attn_head_mask: typing.Optional[torch.Tensor] = None encoder_outputs: typing.Optional[typing.Tuple[typing.Tuple[torch.FloatTensor]]] = None past_key_values: typing.Optional[typing.Tuple[typing.Tuple[torch.FloatTensor]]] = None inputs_embeds: typing.Optional[torch.Tensor] = None decoder_inputs_embeds: typing.Optional[torch.Tensor] = None use_cache: typing.Optional[bool] = None output_attentions: typing.Optional[bool] = None output_hidden_states: typing.Optional[bool] = None return_dict: typing.Optional[bool] = None ) transformers.modeling_outputs.Seq2SeqModelOutput or tuple(torch.FloatTensor)

Parameters

• input_ids (torch.LongTensor of shape (batch_size, sequence_length)) — Indices of input sequence tokens in the vocabulary. MT5 is a model with relative position embeddings so you should be able to pad the inputs on both the right and the left.

Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for detail.

What are input IDs?

To know more on how to prepare input_ids for pretraining take a look a MT5 Training.

• 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.

• decoder_input_ids (torch.LongTensor of shape (batch_size, target_sequence_length), optional) — Indices of decoder input sequence tokens in the vocabulary.

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

What are decoder input IDs?

MT5 uses the pad_token_id as the starting token for decoder_input_ids generation. If past_key_values is used, optionally only the last decoder_input_ids have to be input (see past_key_values).

To know more on how to prepare decoder_input_ids for pretraining take a look at MT5 Training.

• decoder_attention_mask (torch.BoolTensor of shape (batch_size, target_sequence_length), optional) — Default behavior: generate a tensor that ignores pad tokens in decoder_input_ids. Causal mask will also be used by default.
• head_mask (torch.FloatTensor of shape (num_heads,) or (num_layers, num_heads), optional) — Mask to nullify selected heads of the self-attention modules in the encoder. Mask values selected in [0, 1]:

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

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

• encoder_outputs (tuple(tuple(torch.FloatTensor), optional) — Tuple consists of (last_hidden_state, optional: hidden_states, optional: attentions) last_hidden_state of shape (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 (tuple(tuple(torch.FloatTensor)) of length config.n_layers with each tuple having 4 tensors of shape (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 past_key_values are used, the user can optionally input only the last decoder_input_ids (those that don’t have their past key value states given to this model) of shape (batch_size, 1) instead of all decoder_input_ids of shape (batch_size, sequence_length).

• 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.
• decoder_inputs_embeds (torch.FloatTensor of shape (batch_size, target_sequence_length, hidden_size), optional) — Optionally, instead of passing decoder_input_ids you can choose to directly pass an embedded representation. If past_key_values is used, optionally only the last decoder_inputs_embeds have to be input (see past_key_values). This is useful if you want more control over how to convert decoder_input_ids indices into associated vectors than the model’s internal embedding lookup matrix.

If decoder_input_ids and decoder_inputs_embeds are both unset, decoder_inputs_embeds takes the value of inputs_embeds.

• 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.modeling_outputs.Seq2SeqModelOutput or tuple(torch.FloatTensor)

A transformers.modeling_outputs.Seq2SeqModelOutput 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 (MT5Config) and inputs.

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

If past_key_values is used only the last hidden-state of the sequences of shape (batch_size, 1, hidden_size) is output.

• past_key_values (tuple(tuple(torch.FloatTensor)), optional, returned when use_cache=True is passed or when config.use_cache=True) — Tuple of tuple(torch.FloatTensor) of length config.n_layers, with each tuple having 2 tensors of shape (batch_size, num_heads, sequence_length, embed_size_per_head)) and 2 additional tensors of shape (batch_size, num_heads, encoder_sequence_length, embed_size_per_head).

Contains pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention blocks) that can be used (see past_key_values input) to speed up sequential decoding.

• decoder_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, if the model has an embedding layer, + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

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

• decoder_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 of the decoder, after the attention softmax, used to compute the weighted average in the self-attention heads.

• cross_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 of the decoder’s cross-attention layer, after the attention softmax, used to compute the weighted average in the cross-attention heads.

• encoder_last_hidden_state (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) — Sequence of hidden-states at the output of the last layer of the encoder of the model.

• encoder_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, if the model has an embedding layer, + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

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

• encoder_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 of the encoder, after the attention softmax, used to compute the weighted average in the self-attention heads.

The MT5Model 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 AutoTokenizer, MT5Model

>>> tokenizer = AutoTokenizer.from_pretrained("mt5-small")
>>> model = MT5Model.from_pretrained("mt5-small")

>>> input_ids = tokenizer(
...     "Studies have been shown that owning a dog is good for you", return_tensors="pt"
... ).input_ids  # Batch size 1
>>> decoder_input_ids = tokenizer("Studies show that", return_tensors="pt").input_ids  # Batch size 1

>>> # preprocess: Prepend decoder_input_ids with start token which is pad token for MT5Model.
>>> # This is not needed for torch's MT5ForConditionalGeneration as it does this internally using labels arg.
>>> decoder_input_ids = model._shift_right(decoder_input_ids)

>>> # forward pass
>>> outputs = model(input_ids=input_ids, decoder_input_ids=decoder_input_ids)
>>> last_hidden_states = outputs.last_hidden_state

#### parallelize

< >

( device_map = None )

Parameters

• device_map (Dict[int, list], optional, defaults to None) — A dictionary that maps attention modules to devices. Note that the embedding module and LMHead are always automatically mapped to the first device (for esoteric reasons). That means that the first device should have fewer attention modules mapped to it than other devices. For reference, the mt5 models have the following number of attention modules:

• mt5-small: 6
• mt5-base: 12
• mt5-large: 24
• mt5-xl: 24
• mt5-xxl: 24

This is an experimental feature and is a subject to change at a moment’s notice.

Uses a device map to distribute attention modules of the model across several devices. If no device map is given, it will evenly distribute blocks across all devices.

Example:

# Here is an example of a device map on a machine with 4 GPUs using mt5-xl, which has a total of 24 attention modules:
model = MT5ForConditionalGeneration.from_pretrained("mt5-xl")
device_map = {
0: [0, 1, 2],
1: [3, 4, 5, 6, 7, 8, 9],
2: [10, 11, 12, 13, 14, 15, 16],
3: [17, 18, 19, 20, 21, 22, 23],
}
model.parallelize(device_map)

## MT5ForConditionalGeneration

### class transformers.MT5ForConditionalGeneration

< >

( config: MT5Config )

Parameters

• config (MT5Config) — 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.

MT5 Model with a language modeling head on top.

The MT5 model was proposed in Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer by Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, Peter J. Liu. It’s an encoder decoder transformer pre-trained in a text-to-text denoising generative setting.

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.

Examples:

>>> from transformers import MT5ForConditionalGeneration, AutoTokenizer

>>> article = "UN Offizier sagt, dass weiter verhandelt werden muss in Syrien."
>>> summary = "Weiter Verhandlung in Syrien."
>>> inputs = tokenizer(article, text_target=summary, return_tensors="pt")

>>> outputs = model(**inputs)
>>> loss = outputs.loss

#### deparallelize

< >

( )

Moves the model to cpu from a model parallel state.

Example:

# On a 4 GPU machine with mt5-xl:
model = MT5ForConditionalGeneration.from_pretrained("Mt5-xl")
device_map = {
0: [0, 1, 2],
1: [3, 4, 5, 6, 7, 8, 9],
2: [10, 11, 12, 13, 14, 15, 16],
3: [17, 18, 19, 20, 21, 22, 23],
}
model.parallelize(device_map)  # Splits the model across several devices
model.deparallelize()  # Put the model back on cpu and cleans memory by calling torch.cuda.empty_cache()

#### forward

< >

( input_ids: typing.Optional[torch.LongTensor] = None attention_mask: typing.Optional[torch.FloatTensor] = None decoder_input_ids: typing.Optional[torch.LongTensor] = None decoder_attention_mask: typing.Optional[torch.BoolTensor] = None head_mask: typing.Optional[torch.FloatTensor] = None decoder_head_mask: typing.Optional[torch.FloatTensor] = None cross_attn_head_mask: typing.Optional[torch.Tensor] = None encoder_outputs: typing.Optional[typing.Tuple[typing.Tuple[torch.Tensor]]] = None past_key_values: typing.Optional[typing.Tuple[typing.Tuple[torch.Tensor]]] = None inputs_embeds: typing.Optional[torch.FloatTensor] = None decoder_inputs_embeds: typing.Optional[torch.FloatTensor] = None labels: typing.Optional[torch.LongTensor] = None use_cache: typing.Optional[bool] = None output_attentions: typing.Optional[bool] = None output_hidden_states: typing.Optional[bool] = None return_dict: typing.Optional[bool] = None ) transformers.modeling_outputs.Seq2SeqLMOutput or tuple(torch.FloatTensor)

Parameters

• input_ids (torch.LongTensor of shape (batch_size, sequence_length)) — Indices of input sequence tokens in the vocabulary. MT5 is a model with relative position embeddings so you should be able to pad the inputs on both the right and the left.

Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for detail.

What are input IDs?

To know more on how to prepare input_ids for pretraining take a look a MT5 Training.

• 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.

• decoder_input_ids (torch.LongTensor of shape (batch_size, target_sequence_length), optional) — Indices of decoder input sequence tokens in the vocabulary.

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

What are decoder input IDs?

MT5 uses the pad_token_id as the starting token for decoder_input_ids generation. If past_key_values is used, optionally only the last decoder_input_ids have to be input (see past_key_values).

To know more on how to prepare decoder_input_ids for pretraining take a look at MT5 Training.

• decoder_attention_mask (torch.BoolTensor of shape (batch_size, target_sequence_length), optional) — Default behavior: generate a tensor that ignores pad tokens in decoder_input_ids. Causal mask will also be used by default.
• head_mask (torch.FloatTensor of shape (num_heads,) or (num_layers, num_heads), optional) — Mask to nullify selected heads of the self-attention modules in the encoder. Mask values selected in [0, 1]:

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

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

• encoder_outputs (tuple(tuple(torch.FloatTensor), optional) — Tuple consists of (last_hidden_state, optional: hidden_states, optional: attentions) last_hidden_state of shape (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 (tuple(tuple(torch.FloatTensor)) of length config.n_layers with each tuple having 4 tensors of shape (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 past_key_values are used, the user can optionally input only the last decoder_input_ids (those that don’t have their past key value states given to this model) of shape (batch_size, 1) instead of all decoder_input_ids of shape (batch_size, sequence_length).

• 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.
• decoder_inputs_embeds (torch.FloatTensor of shape (batch_size, target_sequence_length, hidden_size), optional) — Optionally, instead of passing decoder_input_ids you can choose to directly pass an embedded representation. If past_key_values is used, optionally only the last decoder_inputs_embeds have to be input (see past_key_values). This is useful if you want more control over how to convert decoder_input_ids indices into associated vectors than the model’s internal embedding lookup matrix.

If decoder_input_ids and decoder_inputs_embeds are both unset, decoder_inputs_embeds takes the value of inputs_embeds.

• 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.Seq2SeqLMOutput or tuple(torch.FloatTensor)

A transformers.modeling_outputs.Seq2SeqLMOutput 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 (MT5Config) and inputs.

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

• past_key_values (tuple(tuple(torch.FloatTensor)), optional, returned when use_cache=True is passed or when config.use_cache=True) — Tuple of tuple(torch.FloatTensor) of length config.n_layers, with each tuple having 2 tensors of shape (batch_size, num_heads, sequence_length, embed_size_per_head)) and 2 additional tensors of shape (batch_size, num_heads, encoder_sequence_length, embed_size_per_head).

Contains pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention blocks) that can be used (see past_key_values input) to speed up sequential decoding.

• decoder_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, if the model has an embedding layer, + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

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

• decoder_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 of the decoder, after the attention softmax, used to compute the weighted average in the self-attention heads.

• cross_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 of the decoder’s cross-attention layer, after the attention softmax, used to compute the weighted average in the cross-attention heads.

• encoder_last_hidden_state (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) — Sequence of hidden-states at the output of the last layer of the encoder of the model.

• encoder_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, if the model has an embedding layer, + one for the output of each layer) of shape (batch_size, sequence_length, hidden_size).

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

• encoder_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 of the encoder, after the attention softmax, used to compute the weighted average in the self-attention heads.

The MT5ForConditionalGeneration 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.

Examples:

>>> from transformers import AutoTokenizer, MT5ForConditionalGeneration

>>> tokenizer = AutoTokenizer.from_pretrained("mt5-small")
>>> model = MT5ForConditionalGeneration.from_pretrained("mt5-small")

>>> # training
>>> input_ids = tokenizer("The <extra_id_0> walks in <extra_id_1> park", return_tensors="pt").input_ids
>>> labels = tokenizer("<extra_id_0> cute dog <extra_id_1> the <extra_id_2>", return_tensors="pt").input_ids
>>> outputs = model(input_ids=input_ids, labels=labels)
>>> loss = outputs.loss
>>> logits = outputs.logits

>>> # inference
>>> input_ids = tokenizer(
...     "summarize: studies have shown that owning a dog is good for you", return_tensors="pt"
... ).input_ids  # Batch size 1
>>> outputs = model.generate(input_ids)
>>> print(tokenizer.decode(outputs[0], skip_special_tokens=True))
>>> # studies have shown that owning a dog is good for you.

#### parallelize

< >

( device_map = None )

Parameters

• device_map (Dict[int, list], optional, defaults to None) — A dictionary that maps attention modules to devices. Note that the embedding module and LMHead are always automatically mapped to the first device (for esoteric reasons). That means that the first device should have fewer attention modules mapped to it than other devices. For reference, the mt5 models have the following number of attention modules:

• mt5-small: 6
• mt5-base: 12
• mt5-large: 24
• mt5-xl: 24
• mt5-xxl: 24

This is an experimental feature and is a subject to change at a moment’s notice.

Uses a device map to distribute attention modules of the model across several devices. If no device map is given, it will evenly distribute blocks across all devices.

Example:

# Here is an example of a device map on a machine with 4 GPUs using mt5-xl, which has a total of 24 attention modules:
model = MT5ForConditionalGeneration.from_pretrained("mt5-xl")
device_map = {
0: [0, 1, 2],
1: [3, 4, 5, 6, 7, 8, 9],
2: [10, 11, 12, 13, 14, 15, 16],
3: [17, 18, 19, 20, 21, 22, 23],
}
model.parallelize(device_map)

## MT5EncoderModel

### class transformers.MT5EncoderModel

< >

( config: MT5Config )

Parameters

• config (MT5Config) — 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 MT5 Model transformer outputting encoder’s raw hidden-states without any specific head on top.

The MT5 model was proposed in Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer by Colin Raffel, Noam Shazeer, Adam Roberts, Katherine Lee, Sharan Narang, Michael Matena, Yanqi Zhou, Wei Li, Peter J. Liu. It’s an encoder decoder transformer pre-trained in a text-to-text denoising generative setting.

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.

Examples:

>>> from transformers import MT5EncoderModel, AutoTokenizer

>>> article = "UN Offizier sagt, dass weiter verhandelt werden muss in Syrien."
>>> input_ids = tokenizer(article, return_tensors="pt").input_ids
>>> outputs = model(input_ids)
>>> hidden_state = outputs.last_hidden_state

#### deparallelize

< >

( )

Moves the model to cpu from a model parallel state.

Example:

# On a 4 GPU machine with mt5-xl:
model = MT5ForConditionalGeneration.from_pretrained("Mt5-xl")
device_map = {
0: [0, 1, 2],
1: [3, 4, 5, 6, 7, 8, 9],
2: [10, 11, 12, 13, 14, 15, 16],
3: [17, 18, 19, 20, 21, 22, 23],
}
model.parallelize(device_map)  # Splits the model across several devices
model.deparallelize()  # Put the model back on cpu and cleans memory by calling torch.cuda.empty_cache()

#### forward

< >

( input_ids: typing.Optional[torch.LongTensor] = None attention_mask: typing.Optional[torch.FloatTensor] = None head_mask: typing.Optional[torch.FloatTensor] = None inputs_embeds: typing.Optional[torch.FloatTensor] = None output_attentions: typing.Optional[bool] = None output_hidden_states: typing.Optional[bool] = None return_dict: typing.Optional[bool] = None ) transformers.modeling_outputs.BaseModelOutput or tuple(torch.FloatTensor)

Parameters

• input_ids (torch.LongTensor of shape (batch_size, sequence_length)) — Indices of input sequence tokens in the vocabulary. MT5 is a model with relative position embeddings so you should be able to pad the inputs on both the right and the left.

Indices can be obtained using AutoTokenizer. See PreTrainedTokenizer.encode() and PreTrainedTokenizer.call() for detail.

To know more on how to prepare input_ids for pretraining take a look a MT5 Training.

• 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.

• 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.
• 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.modeling_outputs.BaseModelOutput or tuple(torch.FloatTensor)

A transformers.modeling_outputs.BaseModelOutput 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 (MT5Config) and inputs.

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

• 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, if the model has an embedding layer, + 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 optional 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 MT5EncoderModel 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 AutoTokenizer, MT5EncoderModel

>>> tokenizer = AutoTokenizer.from_pretrained("mt5-small")
>>> model = MT5EncoderModel.from_pretrained("mt5-small")
>>> input_ids = tokenizer(
...     "Studies have been shown that owning a dog is good for you", return_tensors="pt"
... ).input_ids  # Batch size 1
>>> outputs = model(input_ids=input_ids)
>>> last_hidden_states = outputs.last_hidden_state

#### parallelize

< >

( device_map = None )

Parameters

• device_map (Dict[int, list], optional, defaults to None) — A dictionary that maps attention modules to devices. Note that the embedding module and LMHead are always automatically mapped to the first device (for esoteric reasons). That means that the first device should have fewer attention modules mapped to it than other devices. For reference, the mt5 models have the following number of attention modules:

• mt5-small: 6
• mt5-base: 12
• mt5-large: 24
• mt5-xl: 24
• mt5-xxl: 24

This is an experimental feature and is a subject to change at a moment’s notice.

Uses a device map to distribute attention modules of the model across several devices. If no device map is given, it will evenly distribute blocks across all devices.

Example:

# Here is an example of a device map on a machine with 4 GPUs using mt5-xl, which has a total of 24 attention modules:
model = MT5ForConditionalGeneration.from_pretrained("mt5-xl")
device_map = {
0: [0, 1, 2],
1: [3, 4, 5, 6, 7, 8, 9],
2: [10, 11, 12, 13, 14, 15, 16],
3: [17, 18, 19, 20, 21, 22, 23],
}
model.parallelize(device_map)

## TFMT5Model

### class transformers.TFMT5Model

< >

( *args **kwargs )

This class overrides TFT5Model. Please check the superclass for the appropriate documentation alongside usage examples.

Examples:

>>> from transformers import TFMT5Model, AutoTokenizer

>>> article = "UN Offizier sagt, dass weiter verhandelt werden muss in Syrien."
>>> summary = "Weiter Verhandlung in Syrien."
>>> inputs = tokenizer(article, return_tensors="tf")
>>> labels = tokenizer(text_target=summary, return_tensors="tf")

>>> outputs = model(input_ids=inputs["input_ids"], decoder_input_ids=labels["input_ids"])
>>> hidden_states = outputs.last_hidden_state

## TFMT5ForConditionalGeneration

### class transformers.TFMT5ForConditionalGeneration

< >

( *args **kwargs )

This class overrides TFT5ForConditionalGeneration. Please check the superclass for the appropriate documentation alongside usage examples.

Examples:

>>> from transformers import TFMT5ForConditionalGeneration, AutoTokenizer

>>> article = "UN Offizier sagt, dass weiter verhandelt werden muss in Syrien."
>>> summary = "Weiter Verhandlung in Syrien."
>>> inputs = tokenizer(article, text_target=summary, return_tensors="tf")

>>> outputs = model(**inputs)
>>> loss = outputs.loss

## TFMT5EncoderModel

### class transformers.TFMT5EncoderModel

< >

( *args **kwargs )

This class overrides TFT5EncoderModel. Please check the superclass for the appropriate documentation alongside usage examples.

Examples:

>>> from transformers import TFMT5EncoderModel, AutoTokenizer

>>> article = "UN Offizier sagt, dass weiter verhandelt werden muss in Syrien."
>>> input_ids = tokenizer(article, return_tensors="tf").input_ids
>>> outputs = model(input_ids)
>>> hidden_state = outputs.last_hidden_state

## FlaxMT5Model

### class transformers.FlaxMT5Model

< >

( config: T5Config input_shape: typing.Tuple[int] = (1, 1) seed: int = 0 dtype: dtype = <class 'jax.numpy.float32'> _do_init: bool = True gradient_checkpointing: bool = False **kwargs )

This class overrides FlaxT5Model. Please check the superclass for the appropriate documentation alongside usage examples.

Examples:

>>> from transformers import FlaxMT5Model, AutoTokenizer

>>> article = "UN Offizier sagt, dass weiter verhandelt werden muss in Syrien."
>>> summary = "Weiter Verhandlung in Syrien."
>>> inputs = tokenizer(article, return_tensors="np")

>>> decoder_input_ids = tokenizer(text_target=summary, return_tensors="np").input_ids

>>> outputs = model(input_ids=inputs["input_ids"], decoder_input_ids=decoder_input_ids)
>>> hidden_states = outputs.last_hidden_state

## FlaxMT5ForConditionalGeneration

### class transformers.FlaxMT5ForConditionalGeneration

< >

( config: T5Config input_shape: typing.Tuple[int] = (1, 1) seed: int = 0 dtype: dtype = <class 'jax.numpy.float32'> _do_init: bool = True gradient_checkpointing: bool = False **kwargs )

This class overrides FlaxT5ForConditionalGeneration. Please check the superclass for the appropriate documentation alongside usage examples.

Examples:

>>> from transformers import FlaxMT5ForConditionalGeneration, AutoTokenizer

>>> article = "UN Offizier sagt, dass weiter verhandelt werden muss in Syrien."
>>> summary = "Weiter Verhandlung in Syrien."
>>> inputs = tokenizer(article, return_tensors="np")

>>> decoder_input_ids = tokenizer(text_target=summary, return_tensors="np").input_ids

>>> outputs = model(**inputs, decoder_input_ids=decoder_input_ids)
>>> logits = outputs.logits

## FlaxMT5EncoderModel

### class transformers.FlaxMT5EncoderModel

< >

( config: T5Config input_shape: typing.Tuple[int] = (1, 1) seed: int = 0 dtype: dtype = <class 'jax.numpy.float32'> _do_init: bool = True gradient_checkpointing: bool = False **kwargs )

This class overrides FlaxT5EncoderModel. Please check the superclass for the appropriate documentation alongside usage examples.

Examples:

>>> from transformers import FlaxT5EncoderModel, AutoTokenizer

>>> hidden_states = outputs.last_hidden_state