OpenAI GPT2¶

Overview¶

OpenAI GPT-2 model was proposed in Language Models are Unsupervised Multitask Learners by Alec Radford*, Jeffrey Wu*, Rewon Child, David Luan, Dario Amodei** and Ilya Sutskever**. It’s a causal (unidirectional) transformer pre-trained using language modeling on a very large corpus of ~40 GB of text data.

The abstract from the paper is the following:

GPT-2 is a large transformer-based language model with 1.5 billion parameters, trained on a dataset[1] of 8 million web pages. GPT-2 is trained with a simple objective: predict the next word, given all of the previous words within some text. The diversity of the dataset causes this simple goal to contain naturally occurring demonstrations of many tasks across diverse domains. GPT-2 is a direct scale-up of GPT, with more than 10X the parameters and trained on more than 10X the amount of data.

Tips:

GPT-2 is a model with absolute position embeddings so it’s usually advised to pad the inputs on the right rather than the left.
GPT-2 was trained with a causal language modeling (CLM) objective and is therefore powerful at predicting the next token in a sequence. Leveraging this feature allows GPT-2 to generate syntactically coherent text as it can be observed in the run_generation.py example script.
The PyTorch models can take the past as input, which is the previously computed key/value attention pairs. Using this past value prevents the model from re-computing pre-computed values in the context of text generation. See reusing the past in generative models for more information on the usage of this argument.

Write With Transformer is a webapp created and hosted by Hugging Face showcasing the generative capabilities of several models. GPT-2 is one of them and is available in five different sizes: small, medium, large, xl and a distilled version of the small checkpoint: distilgpt-2.

The original code can be found here.

GPT2Config¶

class transformers.GPT2Config(vocab_size=50257, n_positions=1024, n_ctx=1024, n_embd=768, n_layer=12, n_head=12, activation_function='gelu_new', resid_pdrop=0.1, embd_pdrop=0.1, attn_pdrop=0.1, layer_norm_epsilon=1e-05, initializer_range=0.02, summary_type='cls_index', summary_use_proj=True, summary_activation=None, summary_proj_to_labels=True, summary_first_dropout=0.1, bos_token_id=50256, eos_token_id=50256, **kwargs)[source]¶

This is the configuration class to store the configuration of a GPT2Model. It is used to instantiate an GPT-2 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 GPT-2 small architecture.

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

Parameters

vocab_size (int, optional, defaults to 50257) – Vocabulary size of the GPT-2 model. Defines the different tokens that can be represented by the inputs_ids passed to the forward method of GPT2Model.
n_positions (int, optional, defaults to 1024) – 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).
n_ctx (int, optional, defaults to 1024) – Dimensionality of the causal mask (usually same as n_positions).
n_embd (int, optional, defaults to 768) – Dimensionality of the embeddings and hidden states.
n_layer (int, optional, defaults to 12) – Number of hidden layers in the Transformer encoder.
n_head (int, optional, defaults to 12) – Number of attention heads for each attention layer in the Transformer encoder.
activation_function (str, optional, defaults to ‘gelu’) – Activation function selected in the list [“relu”, “swish”, “gelu”, “tanh”, “gelu_new”].
resid_pdrop (float, optional, defaults to 0.1) – The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.
embd_pdrop (int, optional, defaults to 0.1) – The dropout ratio for the embeddings.
attn_pdrop (float, optional, defaults to 0.1) – The dropout ratio for the attention.
layer_norm_epsilon (float, optional, defaults to 1e-5) – The epsilon to use in the layer normalization layers
initializer_range (float, optional, defaults to 16) – The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
summary_type (string, optional, defaults to “cls_index”) –
Argument used when doing sequence summary. Used in for the multiple choice head in GPT2DoubleHeadsModel. Is one of the following options:
- ’last’ => take the last token hidden state (like XLNet)
- ’first’ => take the first token hidden state (like Bert)
- ’mean’ => take the mean of all tokens hidden states
- ’cls_index’ => supply a Tensor of classification token position (GPT/GPT-2)
- ’attn’ => Not implemented now, use multi-head attention
summary_use_proj (boolean, optional, defaults to True) – Argument used when doing sequence summary. Used in for the multiple choice head in GPT2DoubleHeadsModel. Add a projection after the vector extraction
summary_activation (string or None, optional, defaults to None) – Argument used when doing sequence summary. Used in for the multiple choice head in GPT2DoubleHeadsModel. ‘tanh’ => add a tanh activation to the output, Other => no activation.
summary_proj_to_labels (boolean, optional, defaults to True) – Argument used when doing sequence summary. Used in for the multiple choice head in GPT2DoubleHeadsModel. If True, the projection outputs to config.num_labels classes (otherwise to hidden_size). Default: False.
summary_first_dropout (float, optional, defaults to 0.1) – Argument used when doing sequence summary. Used in for the multiple choice head in GPT2DoubleHeadsModel. Add a dropout before the projection and activation

Example:

>>> from transformers import GPT2Model, GPT2Config

>>> # Initializing a GPT2 configuration
>>> configuration = GPT2Config()

>>> # Initializing a model from the configuration
>>> model = GPT2Model(configuration)

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

GPT2Tokenizer¶

GPT-2 BPE tokenizer, using byte-level Byte-Pair-Encoding.

This tokenizer has been trained to treat spaces like parts of the tokens (a bit like sentencepiece) so a word will be encoded differently whether it is at the beginning of the sentence (without space) or not:

>>> from transformers import GPT2Tokenizer
>>> tokenizer = GPT2Tokenizer.from_pretrained("gpt2")
>>> tokenizer("Hello world")['input_ids']
[15496, 995]
>>> tokenizer(" Hello world")['input_ids']
[18435, 995]

You can get around that behavior by passing add_prefix_space=True when instantiating this tokenizer or when you call it on some text, but since the model was not pretrained this way, it might yield a decrease in performance.

Note

When used with is_pretokenized=True, this tokenizer will add a space before each word (even the first one).

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

Parameters

vocab_file (str) – Path to the vocabulary file.
merges_file (str) – Path to the merges file.
errors (str, optional, defaults to “replace”) – Paradigm to follow when decoding bytes to UTF-8. See bytes.decode for more information.
unk_token (string, optional, defaults to <|endoftext|>) – 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.
bos_token (string, optional, defaults to <|endoftext|>) – The beginning of sequence token.
eos_token (string, optional, defaults to <|endoftext|>) – The end of sequence token.

save_vocabulary(save_directory)[source]¶

Save the vocabulary and special tokens file to a directory.

Parameters: save_directory (str) – The directory in which to save the vocabulary.
Returns: Paths to the files saved.
Return type: Tuple(str)

GPT2TokenizerFast¶

Constructs a “Fast” GPT-2 BPE tokenizer (backed by HuggingFace’s tokenizers library), using byte-level Byte-Pair-Encoding.

This tokenizer has been trained to treat spaces like parts of the tokens (a bit like sentencepiece) so a word will be encoded differently whether it is at the beginning of the sentence (without space) or not:

>>> from transformers import GPT2TokenizerFast
>>> tokenizer = GPT2TokenizerFast.from_pretrained("gpt2")
>>> tokenizer("Hello world")['input_ids']
[15496, 995]
>>> tokenizer(" Hello world")['input_ids']
[18435, 995]

You can get around that behavior by passing add_prefix_space=True when instantiating this tokenizer or when you call it on some text, but since the model was not pretrained this way, it might yield a decrease in performance.

Note

When used with is_pretokenized=True, this tokenizer needs to be instantiated with add_prefix_space=True.

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

Parameters

vocab_file (str) – Path to the vocabulary file.
merges_file (str) – Path to the merges file.
errors (str, optional, defaults to “replace”) – Paradigm to follow when decoding bytes to UTF-8. See bytes.decode for more information.
unk_token (string, optional, defaults to <|endoftext|>) – 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.
bos_token (string, optional, defaults to <|endoftext|>) – The beginning of sequence token.
eos_token (string, optional, defaults to <|endoftext|>) – The end of sequence token.
add_prefix_space (bool, optional, defaults to False) – Whether to add a leading space to the first word. This allows to treat the leading word just as any other word. (GPT2 tokenizer detect beginning of words by the preceeding space)
trim_offsets (bool, optional, defaults to True) – Whether the post processing step should trim offsets to avoid including whitespaces.

GPT2Model¶

class transformers.GPT2Model(config)[source]¶

The bare GPT2 Model transformer 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 matter related to general usage and behavior.

Parameters: config (GPT2Config) – 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=None, past=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, use_cache=None, output_attentions=None, output_hidden_states=None)[source]¶

The GPT2Model 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, input_ids_length)) –
input_ids_length = sequence_length if past is None else past[0].shape[-2] (sequence_length of input past key value states). Indices of input sequence tokens in the vocabulary.

If past is used, only input_ids that do not have their past calculated should be passed as input_ids.

Indices can be obtained using transformers.GPT2Tokenizer. See transformers.PreTrainedTokenizer.encode() and transformers.PreTrainedTokenizer.__call__() for details.

What are input IDs?
past (List[torch.FloatTensor] of length config.n_layers) – Contains pre-computed hidden-states (key and values in the attention blocks) as computed by the model (see past output below). Can be used to speed up sequential decoding. The input_ids which have their past given to this model should not be passed as input_ids as they have already been computed.
attention_mask (torch.FloatTensor of shape (batch_size, sequence_length), optional, defaults to None) –
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.

What are attention masks?
token_type_ids (torch.LongTensor of shape (batch_size, input_ids_length), optional, defaults to None) – input_ids_length = sequence_length if `past is None else 1 Segment token indices to indicate first and second portions of the inputs. Indices are selected in [0, 1]: 0 corresponds to a sentence A token, 1 corresponds to a sentence B token What are token type IDs?
position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional, defaults to None) –
Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

What are position IDs?
head_mask (torch.FloatTensor of shape (num_heads,) or (num_layers, num_heads), optional, defaults to None) – Mask to nullify selected heads of the self-attention modules. Mask values selected in [0, 1]: 1 indicates the head is not masked, 0 indicates the head is masked.
inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional, defaults to None) – 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. If past is used, optionally only the last inputs_embeds have to be input (see past).
use_cache (bool) – If use_cache is True, past key value states are returned and can be used to speed up decoding (see past). Defaults to True.
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.

Returns

last_hidden_state (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size)):

Sequence of hidden-states at the last layer of the model. If past is used only the last hidden-state of the sequences of shape (batch_size, 1, hidden_size) is output.

past (List[torch.FloatTensor] of length config.n_layers with each tensor of shape (2, batch_size, num_heads, sequence_length, embed_size_per_head)):

Contains pre-computed hidden-states (key and values in the attention blocks). Can be used (see past 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 + 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 (GPT2Config) and inputs

Example:

>>> from transformers import GPT2Tokenizer, GPT2Model
>>> import torch

>>> tokenizer = GPT2Tokenizer.from_pretrained('gpt2')
>>> model = GPT2Model.from_pretrained('gpt2')

>>> 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

get_input_embeddings()[source]¶

Returns the model’s input embeddings.

Returns: A torch module mapping vocabulary to hidden states.
Return type: nn.Module

set_input_embeddings(new_embeddings)[source]¶

Set model’s input embeddings

Parameters: value (nn.Module) – A module mapping vocabulary to hidden states.

GPT2LMHeadModel¶

class transformers.GPT2LMHeadModel(config)[source]¶

The GPT2 Model transformer with a language modeling head on top (linear layer with weights tied to the input embeddings).

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 matter related to general usage and behavior.

Parameters: config (GPT2Config) – 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=None, past=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, labels=None, use_cache=None, output_attentions=None, output_hidden_states=None)[source]¶

The GPT2LMHeadModel 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, input_ids_length)) –
input_ids_length = sequence_length if past is None else past[0].shape[-2] (sequence_length of input past key value states). Indices of input sequence tokens in the vocabulary.

If past is used, only input_ids that do not have their past calculated should be passed as input_ids.

Indices can be obtained using transformers.GPT2Tokenizer. See transformers.PreTrainedTokenizer.encode() and transformers.PreTrainedTokenizer.__call__() for details.

What are input IDs?
past (List[torch.FloatTensor] of length config.n_layers) – Contains pre-computed hidden-states (key and values in the attention blocks) as computed by the model (see past output below). Can be used to speed up sequential decoding. The input_ids which have their past given to this model should not be passed as input_ids as they have already been computed.
attention_mask (torch.FloatTensor of shape (batch_size, sequence_length), optional, defaults to None) –
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.

What are attention masks?
token_type_ids (torch.LongTensor of shape (batch_size, input_ids_length), optional, defaults to None) –
input_ids_length = sequence_length if `past is None else 1 Segment token indices to indicate first and second portions of the inputs. Indices are selected in [0, 1]: 0 corresponds to a sentence A token, 1 corresponds to a sentence B token What are token type IDs?
position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional, defaults to None) –
Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

What are position IDs?
head_mask (torch.FloatTensor of shape (num_heads,) or (num_layers, num_heads), optional, defaults to None) – Mask to nullify selected heads of the self-attention modules. Mask values selected in [0, 1]: 1 indicates the head is not masked, 0 indicates the head is masked.
inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional, defaults to None) – 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. If past is used, optionally only the last inputs_embeds have to be input (see past).
use_cache (bool) – If use_cache is True, past key value states are returned and can be used to speed up decoding (see past). Defaults to True.
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 language modeling. Note that the labels are shifted inside the model, i.e. you can set labels = input_ids Indices are selected in [-100, 0, ..., config.vocab_size] All labels set to -100 are ignored (masked), the loss is only computed for labels in [0, ..., config.vocab_size]

Returns

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

Language modeling loss.

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).

past (List[torch.FloatTensor] of length config.n_layers with each tensor of shape (2, batch_size, num_heads, sequence_length, embed_size_per_head)):

Contains pre-computed hidden-states (key and values in the attention blocks). Can be used (see past 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 + 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 (GPT2Config) and inputs

Example:

>>> import torch
>>> from transformers import GPT2Tokenizer, GPT2LMHeadModel

>>> tokenizer = GPT2Tokenizer.from_pretrained('gpt2')
>>> model = GPT2LMHeadModel.from_pretrained('gpt2')

>>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt")
>>> outputs = model(**inputs, labels=inputs["input_ids"])
>>> loss, logits = outputs[:2]

get_output_embeddings()[source]¶

Returns the model’s output embeddings.

Returns: A torch module mapping hidden states to vocabulary.
Return type: nn.Module

GPT2DoubleHeadsModel¶

class transformers.GPT2DoubleHeadsModel(config)[source]¶

The GPT2 Model transformer with a language modeling and a multiple-choice classification head on top e.g. for RocStories/SWAG tasks. The two heads are two linear layers. The language modeling head has its weights tied to the input embeddings, the classification head takes as input the input of a specified classification token index in the input sequence).

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 matter related to general usage and behavior.

Parameters: config (GPT2Config) – 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=None, past=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, mc_token_ids=None, labels=None, mc_labels=None, use_cache=None, output_attentions=None, output_hidden_states=None, **kwargs)[source]¶

The GPT2DoubleHeadsModel 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, input_ids_length)) –
input_ids_length = sequence_length if past is None else past[0].shape[-2] (sequence_length of input past key value states). Indices of input sequence tokens in the vocabulary.

If past is used, only input_ids that do not have their past calculated should be passed as input_ids.

Indices can be obtained using transformers.GPT2Tokenizer. See transformers.PreTrainedTokenizer.encode() and transformers.PreTrainedTokenizer.__call__() for details.

What are input IDs?
past (List[torch.FloatTensor] of length config.n_layers) – Contains pre-computed hidden-states (key and values in the attention blocks) as computed by the model (see past output below). Can be used to speed up sequential decoding. The input_ids which have their past given to this model should not be passed as input_ids as they have already been computed.
attention_mask (torch.FloatTensor of shape (batch_size, sequence_length), optional, defaults to None) –
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.

What are attention masks?
token_type_ids (torch.LongTensor of shape (batch_size, input_ids_length), optional, defaults to None) –
input_ids_length = sequence_length if `past is None else 1 Segment token indices to indicate first and second portions of the inputs. Indices are selected in [0, 1]: 0 corresponds to a sentence A token, 1 corresponds to a sentence B token What are token type IDs?
position_ids (torch.LongTensor of shape (batch_size, sequence_length), optional, defaults to None) –
Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

What are position IDs?
head_mask (torch.FloatTensor of shape (num_heads,) or (num_layers, num_heads), optional, defaults to None) – Mask to nullify selected heads of the self-attention modules. Mask values selected in [0, 1]: 1 indicates the head is not masked, 0 indicates the head is masked.
inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional, defaults to None) – 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. If past is used, optionally only the last inputs_embeds have to be input (see past).
use_cache (bool) – If use_cache is True, past key value states are returned and can be used to speed up decoding (see past). Defaults to True.
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.
mc_token_ids (torch.LongTensor of shape (batch_size, num_choices), optional, default to index of the last token of the input) – Index of the classification token in each input sequence. Selected in the range [0, input_ids.size(-1) - 1[.
labels (torch.LongTensor of shape (batch_size, sequence_length), optional, defaults to None) – Labels for language modeling. Note that the labels are shifted inside the model, i.e. you can set labels = input_ids Indices are selected in [-1, 0, ..., config.vocab_size] All labels set to -100 are ignored (masked), the loss is only computed for labels in [0, ..., config.vocab_size]
mc_labels (torch.LongTensor of shape (batch_size), optional, defaults to None) – Labels for computing the multiple choice classification loss. Indices should be in [0, ..., num_choices] where num_choices is the size of the second dimension of the input tensors. (see input_ids above)
kwargs (Dict[str, any], optional, defaults to {}) – Used to hide legacy arguments that have been deprecated.

Returns

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

Language modeling loss.

mc_loss (torch.FloatTensor of shape (1,), optional, returned when mc_labels is provided):

Multiple choice classification loss.

lm_prediction_scores (torch.FloatTensor of shape (batch_size, num_choices, sequence_length, config.vocab_size)):

Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax).

mc_prediction_scores (torch.FloatTensor of shape (batch_size, num_choices)):

Prediction scores of the multiple choice classification head (scores for each choice before SoftMax).