# Bart¶

DISCLAIMER: If you see something strange, file a Github Issue and assign @sshleifer

## Overview¶

The Bart model was proposed by Mike Lewis, Yinhan Liu, Naman Goyal, Marjan Ghazvininejad, Abdelrahman Mohamed, Omer Levy, Ves Stoyanov and Luke Zettlemoyer on 29 Oct, 2019. According to the abstract,

• Bart uses a standard seq2seq/machine translation architecture with a bidirectional encoder (like BERT) and a left-to-right decoder (like GPT).

• The pretraining task involves randomly shuffling the order of the original sentences and a novel in-filling scheme, where spans of text are replaced with a single mask token.

• BART is particularly effective when fine tuned for text generation but also works well for comprehension tasks. It matches the performance of RoBERTa with comparable training resources on GLUE and SQuAD, achieves new state-of-the-art results on a range of abstractive dialogue, question answering, and summarization tasks, with gains of up to 6 ROUGE.

The Authors’ code can be found here

Implementation Notes:

• Bart doesn’t use token_type_ids for sequence classification. Use BartTokenizer.encode to get the proper splitting.

• The forward pass of BartModel will create decoder inputs (using the helper function transformers.modeling_bart._prepare_bart_decoder_inputs) if they are not passed. This is different than some other modeling APIs.

• Model predictions are intended to be identical to the original implementation. This only works, however, if the string you pass to fairseq.encode starts with a space.

• BartForConditionalGeneration.generate should be used for conditional generation tasks like summarization, see the example in that docstrings

• Models that load the "facebook/bart-large-cnn" weights will not have a mask_token_id, or be able to perform mask filling tasks.

## BartConfig¶

class transformers.BartConfig(activation_dropout=0.0, extra_pos_embeddings=2, activation_function='gelu', vocab_size=50265, d_model=1024, encoder_ffn_dim=4096, encoder_layers=12, encoder_attention_heads=16, decoder_ffn_dim=4096, decoder_layers=12, decoder_attention_heads=16, encoder_layerdrop=0.0, decoder_layerdrop=0.0, attention_dropout=0.0, dropout=0.1, max_position_embeddings=1024, init_std=0.02, classifier_dropout=0.0, num_labels=3, is_encoder_decoder=True, pad_token_id=1, bos_token_id=0, eos_token_id=2, normalize_before=False, add_final_layer_norm=False, scale_embedding=False, normalize_embedding=True, static_position_embeddings=False, add_bias_logits=False, **common_kwargs)[source]

Configuration class for Bart. Parameters are renamed from the fairseq implementation

is_valid_mbart() → bool[source]

Is the configuration aligned with the MBART paper.

## BartTokenizer¶

class transformers.BartTokenizer(vocab_file, merges_file, errors='replace', bos_token='<s>', eos_token='</s>', sep_token='</s>', cls_token='<s>', unk_token='<unk>', pad_token='<pad>', mask_token='<mask>', add_prefix_space=False, **kwargs)[source]

## BartModel¶

class transformers.BartModel(config: transformers.configuration_bart.BartConfig)[source]

The bare BART Model outputting raw hidden-states without any specific head on top.

This model is a PyTorch torch.nn.Module sub-class. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matters related to general usage and behavior.

Parameters

config (BartConfig) – 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.

forward(input_ids, attention_mask=None, decoder_input_ids=None, encoder_outputs: Optional[Tuple] = None, decoder_attention_mask=None, decoder_cached_states=None, use_cache=None, output_attentions=None, output_hidden_states=None)[source]

The BartModel forward method, overrides the __call__() special method.

Note

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.

Parameters
• input_ids (torch.LongTensor of shape (batch_size, sequence_length)) – Indices of input sequence tokens in the vocabulary. Use BartTokenizer.encode to produce them. Padding will be ignored by default should you provide it. Indices can be obtained using transformers.BartTokenizer.encode(text).

• attention_mask (torch.Tensor of shape (batch_size, sequence_length), optional, defaults to None) – Mask to avoid performing attention on padding token indices in input_ids. Mask values selected in [0, 1]: 1 for tokens that are NOT MASKED, 0 for MASKED tokens.

• encoder_outputs (tuple(tuple(torch.FloatTensor), optional, defaults to None) – Tuple consists of (last_hidden_state, optional: hidden_states, optional: attentions) last_hidden_state of shape (batch_size, sequence_length, hidden_size), optional, defaults to None) is a sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention of the decoder.

• decoder_input_ids (torch.LongTensor of shape (batch_size, target_sequence_length), optional, defaults to None) – Provide for translation and summarization training. By default, the model will create this tensor by shifting the input_ids right, following the paper.

• decoder_attention_mask (torch.BoolTensor of shape (batch_size, tgt_seq_len), optional, defaults to None) – Default behavior: generate a tensor that ignores pad tokens in decoder_input_ids. Causal mask will also be used by default. If you want to change padding behavior, you should read _prepare_decoder_inputs() and modify. See diagram 1 in the paper for more info on the default strategy

• output_attentions (bool, optional, defaults to None) – If set to True, the attentions tensors of all attention layers are returned. See attentions under returned tensors for more detail.

Example:

>>> from transformers import BartTokenizer, BartModel
>>> import torch

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

>>> last_hidden_states = outputs[0]  # The last hidden-state is the first element of the output tuple

transformers.modeling_bart._prepare_bart_decoder_inputs(config, input_ids, decoder_input_ids=None, decoder_padding_mask=None, causal_mask_dtype=torch.float32)[source]

Prepare masks that ignore padding tokens in the decoder and a causal mask for the decoder if none are provided. This mimics the default behavior in fairseq. To override it pass in masks. Note: this is not called during generation

## BartForSequenceClassification¶

class transformers.BartForSequenceClassification(config: transformers.configuration_bart.BartConfig, **kwargs)[source]

Bart model with a sequence classification/head on top (a linear layer on top of the pooled output) e.g. for GLUE tasks.

This model is a PyTorch torch.nn.Module sub-class. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matters related to general usage and behavior.

Parameters

config (BartConfig) – 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.

forward(input_ids, attention_mask=None, encoder_outputs=None, decoder_input_ids=None, decoder_attention_mask=None, labels=None, output_attentions=None, output_hidden_states=None, use_cache=None)[source]

The BartForSequenceClassification forward method, overrides the __call__() special method.

Note

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.

Parameters
• input_ids (torch.LongTensor of shape (batch_size, sequence_length)) – Indices of input sequence tokens in the vocabulary. Use BartTokenizer.encode to produce them. Padding will be ignored by default should you provide it. Indices can be obtained using transformers.BartTokenizer.encode(text).

• attention_mask (torch.Tensor of shape (batch_size, sequence_length), optional, defaults to None) – Mask to avoid performing attention on padding token indices in input_ids. Mask values selected in [0, 1]: 1 for tokens that are NOT MASKED, 0 for MASKED tokens.

• encoder_outputs (tuple(tuple(torch.FloatTensor), optional, defaults to None) – Tuple consists of (last_hidden_state, optional: hidden_states, optional: attentions) last_hidden_state of shape (batch_size, sequence_length, hidden_size), optional, defaults to None) is a sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention of the decoder.

• decoder_input_ids (torch.LongTensor of shape (batch_size, target_sequence_length), optional, defaults to None) – Provide for translation and summarization training. By default, the model will create this tensor by shifting the input_ids right, following the paper.

• decoder_attention_mask (torch.BoolTensor of shape (batch_size, tgt_seq_len), optional, defaults to None) – Default behavior: generate a tensor that ignores pad tokens in decoder_input_ids. Causal mask will also be used by default. If you want to change padding behavior, you should read _prepare_decoder_inputs() and modify. See diagram 1 in the paper for more info on the default strategy

• output_attentions (bool, optional, defaults to None) – If set to True, the attentions tensors of all attention layers are returned. See attentions under returned tensors for more detail.

• labels (torch.LongTensor of shape (batch_size,), optional, defaults to None) – Labels for computing the sequence classification/regression loss. Indices should be in [0, ..., config.num_labels - 1]. If config.num_labels > 1 a classification loss is computed (Cross-Entropy).

Returns

loss (torch.FloatTensor of shape (1,), optional, returned when label is provided):

Classification loss (cross entropy)

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 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 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.

Return type

tuple(torch.FloatTensor) comprising various elements depending on the configuration (BartConfig) and inputs

Example:

>>> from transformers import BartTokenizer, BartForSequenceClassification
>>> 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, logits = outputs[:2]


class transformers.BartForQuestionAnswering(config)[source]

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

This model is a PyTorch torch.nn.Module sub-class. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matters related to general usage and behavior.

Parameters

config (BartConfig) – 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.

forward(input_ids, attention_mask=None, encoder_outputs=None, decoder_input_ids=None, decoder_attention_mask=None, start_positions=None, end_positions=None, output_attentions=None, output_hidden_states=None, use_cache=None)[source]

The BartForQuestionAnswering forward method, overrides the __call__() special method.

Note

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.

Parameters
• input_ids (torch.LongTensor of shape (batch_size, sequence_length)) – Indices of input sequence tokens in the vocabulary. Use BartTokenizer.encode to produce them. Padding will be ignored by default should you provide it. Indices can be obtained using transformers.BartTokenizer.encode(text).

• attention_mask (torch.Tensor of shape (batch_size, sequence_length), optional, defaults to None) – Mask to avoid performing attention on padding token indices in input_ids. Mask values selected in [0, 1]: 1 for tokens that are NOT MASKED, 0 for MASKED tokens.

• encoder_outputs (tuple(tuple(torch.FloatTensor), optional, defaults to None) – Tuple consists of (last_hidden_state, optional: hidden_states, optional: attentions) last_hidden_state of shape (batch_size, sequence_length, hidden_size), optional, defaults to None) is a sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention of the decoder.

• decoder_input_ids (torch.LongTensor of shape (batch_size, target_sequence_length), optional, defaults to None) – Provide for translation and summarization training. By default, the model will create this tensor by shifting the input_ids right, following the paper.

• decoder_attention_mask (torch.BoolTensor of shape (batch_size, tgt_seq_len), optional, defaults to None) – Default behavior: generate a tensor that ignores pad tokens in decoder_input_ids. Causal mask will also be used by default. If you want to change padding behavior, you should read _prepare_decoder_inputs() and modify. See diagram 1 in the paper for more info on the default strategy

• output_attentions (bool, optional, defaults to None) – If set to True, the attentions tensors of all attention layers are returned. See attentions under returned tensors for more detail.

• start_positions (torch.LongTensor of shape (batch_size,), optional, defaults to None) – 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, defaults to None) – 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

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_scores (torch.FloatTensor of shape (batch_size, sequence_length,)):

Span-start scores (before SoftMax).

end_scores (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.

Return type

tuple(torch.FloatTensor) comprising various elements depending on the configuration (BartConfig) and inputs

Example:

>>> from transformers import BartTokenizer, BartForQuestionAnswering
>>> import torch

>>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
>>> start_positions = torch.tensor([1])
>>> end_positions = torch.tensor([3])

>>> outputs = model(**inputs, start_positions=start_positions, end_positions=end_positions)
>>> loss, start_scores, end_scores = outputs[:3]


## BartForConditionalGeneration¶

class transformers.BartForConditionalGeneration(config: transformers.configuration_bart.BartConfig)[source]

The BART Model with a language modeling head. Can be used for summarization.

This model is a PyTorch torch.nn.Module sub-class. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matters related to general usage and behavior.

Parameters

config (BartConfig) – 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.

forward(input_ids, attention_mask=None, encoder_outputs=None, decoder_input_ids=None, decoder_attention_mask=None, decoder_cached_states=None, labels=None, use_cache=None, output_attentions=None, output_hidden_states=None, **unused)[source]

The BartForConditionalGeneration forward method, overrides the __call__() special method.

Note

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.

Parameters
• input_ids (torch.LongTensor of shape (batch_size, sequence_length)) – Indices of input sequence tokens in the vocabulary. Use BartTokenizer.encode to produce them. Padding will be ignored by default should you provide it. Indices can be obtained using transformers.BartTokenizer.encode(text).

• attention_mask (torch.Tensor of shape (batch_size, sequence_length), optional, defaults to None) – Mask to avoid performing attention on padding token indices in input_ids. Mask values selected in [0, 1]: 1 for tokens that are NOT MASKED, 0 for MASKED tokens.

• encoder_outputs (tuple(tuple(torch.FloatTensor), optional, defaults to None) – Tuple consists of (last_hidden_state, optional: hidden_states, optional: attentions) last_hidden_state of shape (batch_size, sequence_length, hidden_size), optional, defaults to None) is a sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention of the decoder.

• decoder_input_ids (torch.LongTensor of shape (batch_size, target_sequence_length), optional, defaults to None) – Provide for translation and summarization training. By default, the model will create this tensor by shifting the input_ids right, following the paper.

• decoder_attention_mask (torch.BoolTensor of shape (batch_size, tgt_seq_len), optional, defaults to None) – Default behavior: generate a tensor that ignores pad tokens in decoder_input_ids. Causal mask will also be used by default. If you want to change padding behavior, you should read _prepare_decoder_inputs() and modify. See diagram 1 in the paper for more info on the default strategy

• output_attentions (bool, optional, defaults to None) – If set to True, the attentions tensors of all attention layers are returned. See attentions under returned tensors for more detail.

• labels (torch.LongTensor of shape (batch_size, sequence_length), optional, defaults to None) – Labels for computing the masked language modeling loss. Indices should either be in [0, ..., config.vocab_size] or -100 (see input_ids docstring). Tokens with indices set to -100 are ignored (masked), the loss is only computed for the tokens with labels in [0, ..., config.vocab_size].

Returns

masked_lm_loss (optional, returned when labels is provided) torch.FloatTensor of shape (1,):

prediction_scores (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.

Return type

tuple(torch.FloatTensor) comprising various elements depending on the configuration (RobertaConfig) and inputs

Conditional generation example:

# Mask filling only works for bart-large
from transformers import BartTokenizer, BartForConditionalGeneration
TXT = "My friends are <mask> but they eat too many carbs."

input_ids = tokenizer([TXT], return_tensors='pt')['input_ids']
logits = model(input_ids)[0]

values, predictions = probs.topk(5)

tokenizer.decode(predictions).split()
# ['good', 'great', 'all', 'really', 'very']


Summarization example:

from transformers import BartTokenizer, BartForConditionalGeneration, BartConfig

# see examples/summarization/bart/run_eval.py for a longer example

ARTICLE_TO_SUMMARIZE = "My friends are cool but they eat too many carbs."
inputs = tokenizer([ARTICLE_TO_SUMMARIZE], max_length=1024, return_tensors='pt')

# Generate Summary
summary_ids = model.generate(inputs['input_ids'], num_beams=4, max_length=5, early_stopping=True)
print([tokenizer.decode(g, skip_special_tokens=True, clean_up_tokenization_spaces=False) for g in summary_ids])

generate(input_ids: Optional[torch.LongTensor] = None, max_length: Optional[int] = None, min_length: Optional[int] = None, do_sample: Optional[bool] = None, early_stopping: Optional[bool] = None, num_beams: Optional[int] = None, temperature: Optional[float] = None, top_k: Optional[int] = None, top_p: Optional[float] = None, repetition_penalty: Optional[float] = None, bad_words_ids: Optional[Iterable[int]] = None, bos_token_id: Optional[int] = None, pad_token_id: Optional[int] = None, eos_token_id: Optional[int] = None, length_penalty: Optional[float] = None, no_repeat_ngram_size: Optional[int] = None, num_return_sequences: Optional[int] = None, attention_mask: Optional[torch.LongTensor] = None, decoder_start_token_id: Optional[int] = None, use_cache: Optional[bool] = None, **model_specific_kwargs) → torch.LongTensor

Generates sequences for models with a LM head. The method currently supports greedy decoding, beam-search decoding, sampling with temperature, sampling with top-k or nucleus sampling.

Parameters
• input_ids – (optional) torch.LongTensor of shape (batch_size, sequence_length) The sequence used as a prompt for the generation. If None the method initializes it as an empty torch.LongTensor of shape (1,).

• max_length – (optional) int The max length of the sequence to be generated. Between min_length and infinity. Default to 20.

• min_length – (optional) int The min length of the sequence to be generated. Between 0 and infinity. Default to 0.

• do_sample – (optional) bool If set to False greedy decoding is used. Otherwise sampling is used. Defaults to False as defined in configuration_utils.PretrainedConfig.

• early_stopping – (optional) bool if set to True beam search is stopped when at least num_beams sentences finished per batch. Defaults to False as defined in configuration_utils.PretrainedConfig.

• num_beams – (optional) int Number of beams for beam search. Must be between 1 and infinity. 1 means no beam search. Default to 1.

• temperature – (optional) float The value used to module the next token probabilities. Must be strictly positive. Default to 1.0.

• top_k – (optional) int The number of highest probability vocabulary tokens to keep for top-k-filtering. Between 1 and infinity. Default to 50.

• top_p – (optional) float The cumulative probability of parameter highest probability vocabulary tokens to keep for nucleus sampling. Must be between 0 and 1. Default to 1.

• repetition_penalty – (optional) float The parameter for repetition penalty. Between 1.0 and infinity. 1.0 means no penalty. Default to 1.0.

• pad_token_id – (optional) int Padding token. Default to specicic model pad_token_id or None if it does not exist.

• bos_token_id – (optional) int BOS token. Defaults to bos_token_id as defined in the models config.

• eos_token_id – (optional) int EOS token. Defaults to eos_token_id as defined in the models config.

• length_penalty – (optional) float Exponential penalty to the length. Default to 1.

• no_repeat_ngram_size – (optional) int If set to int > 0, all ngrams of size no_repeat_ngram_size can only occur once.

• bad_words_ids – (optional) list of lists of int bad_words_ids contains tokens that are not allowed to be generated. In order to get the tokens of the words that should not appear in the generated text, use tokenizer.encode(bad_word, add_prefix_space=True).

• num_return_sequences – (optional) int The number of independently computed returned sequences for each element in the batch. Default to 1.

torch.LongTensor of same shape as input_ids Mask to avoid performing attention on padding token indices. Mask values selected in [0, 1]: 1 for tokens that are NOT MASKED, 0 for MASKED tokens. Defaults to None.

• decoder_start_token_id=None – (optional) int If an encoder-decoder model starts decoding with a different token than BOS. Defaults to None and is changed to BOS later.

• use_cache – (optional) bool If use_cache is True, past key values are used to speed up decoding if applicable to model. Defaults to True.

• model_specific_kwargs – (optional) dict Additional model specific kwargs will be forwarded to the forward function of the model.

Returns

torch.LongTensor of shape (batch_size * num_return_sequences, sequence_length)

sequence_length is either equal to max_length or shorter if all batches finished early due to the eos_token_id

Return type

output

Examples:

tokenizer = AutoTokenizer.from_pretrained('distilgpt2')   # Initialize tokenizer
outputs = model.generate(max_length=40)  # do greedy decoding
print('Generated: {}'.format(tokenizer.decode(outputs[0], skip_special_tokens=True)))

tokenizer = AutoTokenizer.from_pretrained('openai-gpt')   # Initialize tokenizer
input_context = 'The dog'
input_ids = tokenizer.encode(input_context, return_tensors='pt')  # encode input context
outputs = model.generate(input_ids=input_ids, num_beams=5, num_return_sequences=3, temperature=1.5)  # generate 3 independent sequences using beam search decoding (5 beams) with sampling from initial context 'The dog'
for i in range(3): #  3 output sequences were generated
print('Generated {}: {}'.format(i, tokenizer.decode(outputs[i], skip_special_tokens=True)))

tokenizer = AutoTokenizer.from_pretrained('distilgpt2')   # Initialize tokenizer
input_context = 'The dog'
input_ids = tokenizer.encode(input_context, return_tensors='pt')  # encode input context
outputs = model.generate(input_ids=input_ids, max_length=40, temperature=0.7, num_return_sequences=3)  # 3 generate sequences using by sampling
for i in range(3): #  3 output sequences were generated
print('Generated {}: {}'.format(i, tokenizer.decode(outputs[i], skip_special_tokens=True)))

tokenizer = AutoTokenizer.from_pretrained('ctrl')   # Initialize tokenizer