# MegatronBERT¶

## Overview¶

The MegatronBERT model was proposed in Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism by Mohammad Shoeybi, Mostofa Patwary, Raul Puri, Patrick LeGresley, Jared Casper and Bryan Catanzaro.

The abstract from the paper is the following:

Recent work in language modeling demonstrates that training large transformer models advances the state of the art in Natural Language Processing applications. However, very large models can be quite difficult to train due to memory constraints. In this work, we present our techniques for training very large transformer models and implement a simple, efficient intra-layer model parallel approach that enables training transformer models with billions of parameters. Our approach does not require a new compiler or library changes, is orthogonal and complimentary to pipeline model parallelism, and can be fully implemented with the insertion of a few communication operations in native PyTorch. We illustrate this approach by converging transformer based models up to 8.3 billion parameters using 512 GPUs. We sustain 15.1 PetaFLOPs across the entire application with 76% scaling efficiency when compared to a strong single GPU baseline that sustains 39 TeraFLOPs, which is 30% of peak FLOPs. To demonstrate that large language models can further advance the state of the art (SOTA), we train an 8.3 billion parameter transformer language model similar to GPT-2 and a 3.9 billion parameter model similar to BERT. We show that careful attention to the placement of layer normalization in BERT-like models is critical to achieving increased performance as the model size grows. Using the GPT-2 model we achieve SOTA results on the WikiText103 (10.8 compared to SOTA perplexity of 15.8) and LAMBADA (66.5% compared to SOTA accuracy of 63.2%) datasets. Our BERT model achieves SOTA results on the RACE dataset (90.9% compared to SOTA accuracy of 89.4%).

Tips:

We have provided pretrained BERT-345M checkpoints for use to evaluate or finetuning downstream tasks.

To access these checkpoints, first sign up for and setup the NVIDIA GPU Cloud (NGC) Registry CLI. Further documentation for downloading models can be found in the NGC documentation.

BERT-345M-uncased:

.. code-block:: bash


wget –content-disposition https://api.ngc.nvidia.com/v2/models/nvidia/megatron_bert_345m/versions/v0.1_uncased/zip -O megatron_bert_345m_v0_1_uncased.zip

BERT-345M-cased:

.. code-block:: bash


wget –content-disposition https://api.ngc.nvidia.com/v2/models/nvidia/megatron_bert_345m/versions/v0.1_cased/zip -O megatron_bert_345m_v0_1_cased.zip

Once you have obtained the checkpoints from NVIDIA GPU Cloud (NGC), you have to convert them to a format that will easily be loaded by Hugging Face Transformers and our port of the BERT code.

The following commands allow you to do the conversion. We assume that the folder models/megatron_bert contains megatron_bert_345m_v0_1_{cased, uncased}.zip and that the commands are run from inside that folder:

.. code-block:: bash


python3 $PATH_TO_TRANSFORMERS/models/megatron_bert/convert_megatron_bert_checkpoint.py megatron_bert_345m_v0_1_uncased.zip python3$PATH_TO_TRANSFORMERS/models/megatron_bert/convert_megatron_bert_checkpoint.py megatron_bert_345m_v0_1_cased.zip


This model was contributed by jdemouth. The original code can be found here. That repository contains a multi-GPU and multi-node implementation of the Megatron Language models. In particular, it contains a hybrid model parallel approach using “tensor parallel” and “pipeline parallel” techniques.

## MegatronBertConfig¶

class transformers.MegatronBertConfig(vocab_size=29056, hidden_size=1024, num_hidden_layers=24, num_attention_heads=16, intermediate_size=4096, hidden_act='gelu', hidden_dropout_prob=0.1, attention_probs_dropout_prob=0.1, max_position_embeddings=512, type_vocab_size=2, initializer_range=0.02, layer_norm_eps=1e-12, pad_token_id=0, position_embedding_type='absolute', use_cache=True, **kwargs)[source]

This is the configuration class to store the configuration of a MegatronBertModel. It is used to instantiate a MEGATRON_BERT 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 MEGATRON_BERT megatron-bert-uncased-345m 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 29056) – Vocabulary size of the MEGATRON_BERT model. Defines the number of different tokens that can be represented by the inputs_ids passed when calling MegatronBertModel.

• hidden_size (int, optional, defaults to 1024) – Dimensionality of the encoder layers and the pooler layer.

• num_hidden_layers (int, optional, defaults to 24) – Number of hidden layers in the Transformer encoder.

• num_attention_heads (int, optional, defaults to 16) – Number of attention heads for each attention layer in the Transformer encoder.

• intermediate_size (int, optional, defaults to 4096) – Dimensionality of the “intermediate” (often named feed-forward) layer in the Transformer encoder.

• hidden_act (str or Callable, optional, defaults to "gelu") – The non-linear activation function (function or string) in the encoder and pooler. If string, "gelu", "relu", "silu" and "gelu_new" are supported.

• hidden_dropout_prob (float, optional, defaults to 0.1) – The dropout probability for all fully connected layers in the embeddings, encoder, and pooler.

• attention_probs_dropout_prob (float, optional, defaults to 0.1) – The dropout ratio for the attention probabilities.

• max_position_embeddings (int, optional, defaults to 512) – 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).

• type_vocab_size (int, optional, defaults to 2) – The vocabulary size of the token_type_ids passed when calling MegatronBertModel.

• initializer_range (float, optional, defaults to 0.02) – The standard deviation of the truncated_normal_initializer for initializing all weight matrices.

• layer_norm_eps (float, optional, defaults to 1e-12) – The epsilon used by the layer normalization layers.

• position_embedding_type (str, optional, defaults to "absolute") – Type of position embedding. Choose one of "absolute", "relative_key", "relative_key_query". For positional embeddings use "absolute". For more information on "relative_key", please refer to Self-Attention with Relative Position Representations (Shaw et al.). For more information on "relative_key_query", please refer to Method 4 in Improve Transformer Models with Better Relative Position Embeddings (Huang et al.).

• use_cache (bool, optional, defaults to True) – Whether or not the model should return the last key/values attentions (not used by all models). Only relevant if config.is_decoder=True.

Examples:

>>> from transformers import MegatronBertModel, MegatronBertConfig

>>> # Initializing a MEGATRON_BERT bert-base-uncased style configuration
>>> configuration = MegatronBertConfig()

>>> # Initializing a model from the bert-base-uncased style configuration
>>> model = MegatronBertModel(configuration)

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


## MegatronBertModel¶

class transformers.MegatronBertModel(config, add_pooling_layer=True)[source]

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

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.

Parameters

config (MegatronBertConfig) – 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 model can behave as an encoder (with only self-attention) as well as a decoder, in which case a layer of cross-attention is added between the self-attention layers, following the architecture described in Attention is all you need by Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser and Illia Polosukhin.

To behave as an decoder the model needs to be initialized with the is_decoder argument of the configuration set to True. To be used in a Seq2Seq model, the model needs to initialized with both is_decoder argument and add_cross_attention set to True; an encoder_hidden_states is then expected as an input to the forward pass.

forward(input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, encoder_hidden_states=None, encoder_attention_mask=None, past_key_values=None, use_cache=None, output_attentions=None, output_hidden_states=None, return_dict=None)[source]

The MegatronBertModel 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.

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

What are input IDs?

• 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.

• token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) –

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) –

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) –

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.

• encoder_hidden_states (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. Used in the cross-attention if the model is configured as a decoder.

• encoder_attention_mask (torch.FloatTensor of shape (batch_size, sequence_length), optional) –

Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in the cross-attention if the model is configured as a decoder. Mask values selected in [0, 1]:

• 1 for tokens that are not masked,

• 0 for tokens that are masked.

• 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).

• 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).

Returns

A BaseModelOutputWithPoolingAndCrossAttentions 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 (MegatronBertConfig) 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.

• pooler_output (torch.FloatTensor of shape (batch_size, hidden_size)) – Last layer hidden-state of the first token of the sequence (classification token) after further processing through the layers used for the auxiliary pretraining task. E.g. for BERT-family of models, this returns the classification token after processing through a linear layer and a tanh activation function. The linear layer weights are trained from the next sentence prediction (classification) objective during pretraining.

• 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.

• cross_attentions (tuple(torch.FloatTensor), optional, returned when output_attentions=True and config.add_cross_attention=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.

• 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 optionally if config.is_encoder_decoder=True 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 optionally if config.is_encoder_decoder=True in the cross-attention blocks) that can be used (see past_key_values input) to speed up sequential decoding.

Return type

BaseModelOutputWithPoolingAndCrossAttentions or tuple(torch.FloatTensor)

Example:

>>> from transformers import BertTokenizer, MegatronBertModel
>>> import torch

>>> tokenizer = BertTokenizer.from_pretrained('nvidia/megatron-bert-cased-345m')
>>> model = MegatronBertModel.from_pretrained('nvidia/megatron-bert-cased-345m')

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

>>> last_hidden_states = outputs.last_hidden_state


class transformers.MegatronBertForMaskedLM(config)[source]

MegatronBert Model with a language modeling head on top.

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.

Parameters

config (MegatronBertConfig) – 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, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, encoder_hidden_states=None, encoder_attention_mask=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None)[source]

The MegatronBertForMaskedLM 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.

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

What are input IDs?

• 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.

• token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) –

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) –

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) –

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.

• labels (torch.LongTensor of shape (batch_size, sequence_length), optional) – Labels for computing the masked language modeling loss. Indices should be in [-100, 0, ..., config.vocab_size] (see input_ids docstring) Tokens with indices set to -100 are ignored (masked), the loss is only computed for the tokens with labels in [0, ..., config.vocab_size]

Returns

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

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

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

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

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

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

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

Return type

MaskedLMOutput or tuple(torch.FloatTensor)

Example:

>>> from transformers import BertTokenizer, MegatronBertForMaskedLM
>>> import torch

>>> tokenizer = BertTokenizer.from_pretrained('nvidia/megatron-bert-cased-345m')

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

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


## MegatronBertForCausalLM¶

class transformers.MegatronBertForCausalLM(config)[source]

MegatronBert Model with a language modeling head on top for CLM fine-tuning.

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.

Parameters

config (MegatronBertConfig) – 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, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, encoder_hidden_states=None, encoder_attention_mask=None, labels=None, past_key_values=None, use_cache=None, output_attentions=None, output_hidden_states=None, return_dict=None)[source]

The MegatronBertForCausalLM 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.

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

What are input IDs?

• 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.

• token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) –

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) –

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) –

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.

• encoder_hidden_states (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. Used in the cross-attention if the model is configured as a decoder.

• encoder_attention_mask (torch.FloatTensor of shape (batch_size, sequence_length), optional) –

Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in the cross-attention if the model is configured as a decoder. Mask values selected in [0, 1]:

• 1 for tokens that are not masked,

• 0 for tokens that are masked.

• labels (torch.LongTensor of shape (batch_size, sequence_length), optional) – Labels for computing the left-to-right language modeling loss (next word prediction). Indices should be in [-100, 0, ..., config.vocab_size] (see input_ids docstring) Tokens with indices set to -100 are ignored (masked), the loss is only computed for the tokens with labels n [0, ..., config.vocab_size]

• 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).

• 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).

Returns

A CausalLMOutputWithCrossAttentions 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 (MegatronBertConfig) and inputs.

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

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

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

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

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

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

• 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).

Cross attentions weights after the attention softmax, used to compute the weighted average in the cross-attention heads.

• past_key_values (tuple(tuple(torch.FloatTensor)), optional, returned when use_cache=True is passed or when config.use_cache=True) – Tuple of torch.FloatTensor tuples of length config.n_layers, with each tuple containing the cached key, value states of the self-attention and the cross-attention layers if model is used in encoder-decoder setting. Only relevant if config.is_decoder = True.

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

Example:

>>> from transformers import BertTokenizer, MegatronBertForCausalLM, MegatronBertConfig
>>> import torch

>>> tokenizer = BertTokenizer.from_pretrained('nvidia/megatron-bert-cased-345m')

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

>>> prediction_logits = outputs.logits


Return type

CausalLMOutputWithCrossAttentions or tuple(torch.FloatTensor)

## MegatronBertForNextSentencePrediction¶

class transformers.MegatronBertForNextSentencePrediction(config)[source]

MegatronBert Model with a next sentence prediction (classification) head on top.

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.

Parameters

config (MegatronBertConfig) – 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, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None, **kwargs)[source]

The MegatronBertForNextSentencePrediction 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.

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

What are input IDs?

• 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.

• token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) –

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) –

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) –

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.

• labels (torch.LongTensor of shape (batch_size,), optional) –

Labels for computing the next sequence prediction (classification) loss. Input should be a sequence pair (see input_ids docstring). Indices should be in [0, 1]:

• 0 indicates sequence B is a continuation of sequence A,

• 1 indicates sequence B is a random sequence.

Returns

A NextSentencePredictorOutput 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 (MegatronBertConfig) and inputs.

• loss (torch.FloatTensor of shape (1,), optional, returned when next_sentence_label is provided) – Next sequence prediction (classification) loss.

• logits (torch.FloatTensor of shape (batch_size, 2)) – Prediction scores of the next sequence prediction (classification) head (scores of True/False continuation 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.

Example:

>>> from transformers import BertTokenizer, MegatronBertForNextSentencePrediction
>>> import torch

>>> tokenizer = BertTokenizer.from_pretrained('nvidia/megatron-bert-cased-345m')
>>> model = MegatronBertForNextSentencePrediction.from_pretrained('nvidia/megatron-bert-cased-345m')

>>> prompt = "In Italy, pizza served in formal settings, such as at a restaurant, is presented unsliced."
>>> next_sentence = "The sky is blue due to the shorter wavelength of blue light."
>>> encoding = tokenizer(prompt, next_sentence, return_tensors='pt')

>>> outputs = model(**encoding, labels=torch.LongTensor([1]))
>>> logits = outputs.logits
>>> assert logits[0, 0] < logits[0, 1] # next sentence was random


Return type

NextSentencePredictorOutput or tuple(torch.FloatTensor)

## MegatronBertForPreTraining¶

class transformers.MegatronBertForPreTraining(config, add_binary_head=True)[source]

MegatronBert Model with two heads on top as done during the pretraining: a masked language modeling head and a next sentence prediction (classification) head.

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.

Parameters

config (MegatronBertConfig) – 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, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, labels=None, next_sentence_label=None, output_attentions=None, output_hidden_states=None, return_dict=None)[source]

The MegatronBertForPreTraining 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.

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

What are input IDs?

• 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.

• token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) –

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) –

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) –

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.

• labels (torch.LongTensor of shape (batch_size, sequence_length), optional) – Labels for computing the masked language modeling loss. Indices should be in [-100, 0, ..., config.vocab_size] (see input_ids docstring) Tokens with indices set to -100 are ignored (masked), the loss is only computed for the tokens with labels in [0, ..., config.vocab_size]

• next_sentence_label (torch.LongTensor of shape (batch_size,), optional) –

Labels for computing the next sequence prediction (classification) loss. Input should be a sequence pair (see input_ids docstring) Indices should be in [0, 1]:

• 0 indicates sequence B is a continuation of sequence A,

• 1 indicates sequence B is a random sequence.

• kwargs (Dict[str, any], optional, defaults to {}) – Used to hide legacy arguments that have been deprecated.

Returns

A MegatronBertForPreTrainingOutput 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 (MegatronBertConfig) and inputs.

• loss (optional, returned when labels is provided, torch.FloatTensor of shape (1,)) – Total loss as the sum of the masked language modeling loss and the next sequence prediction (classification) loss.

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

• seq_relationship_logits (torch.FloatTensor of shape (batch_size, 2)) – Prediction scores of the next sequence prediction (classification) head (scores of True/False continuation 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.

Example:

>>> from transformers import BertTokenizer, MegatronBertForPreTraining
>>> import torch

>>> tokenizer = BertTokenizer.from_pretrained('nvidia/megatron-bert-cased-345m')
>>> model = MegatronBertForPreTraining.from_pretrained('nvidia/megatron-bert-cased-345m')

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

>>> prediction_logits = outputs.prediction_logits
>>> seq_relationship_logits = outputs.seq_relationship_logits


Return type

MegatronBertForPreTrainingOutput or tuple(torch.FloatTensor)

## MegatronBertForSequenceClassification¶

class transformers.MegatronBertForSequenceClassification(config)[source]

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

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.

Parameters

config (MegatronBertConfig) – 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, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None)[source]

The MegatronBertForSequenceClassification 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.

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

What are input IDs?

• 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.

• token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) –

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) –

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) –

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.

• labels (torch.LongTensor of shape (batch_size,), optional) – Labels for computing the sequence classification/regression loss. Indices should be in [0, ..., config.num_labels - 1]. If config.num_labels == 1 a regression loss is computed (Mean-Square loss), If config.num_labels > 1 a classification loss is computed (Cross-Entropy).

Returns

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

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

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

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

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

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

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

Return type

SequenceClassifierOutput or tuple(torch.FloatTensor)

Example:

>>> from transformers import BertTokenizer, MegatronBertForSequenceClassification
>>> import torch

>>> tokenizer = BertTokenizer.from_pretrained('nvidia/megatron-bert-cased-345m')
>>> model = MegatronBertForSequenceClassification.from_pretrained('nvidia/megatron-bert-cased-345m')

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


## MegatronBertForMultipleChoice¶

class transformers.MegatronBertForMultipleChoice(config)[source]

MegatronBert Model with a multiple choice classification head on top (a linear layer on top of the pooled output and a softmax) e.g. for RocStories/SWAG tasks.

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.

Parameters

config (MegatronBertConfig) – 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, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None)[source]

The MegatronBertForMultipleChoice 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, num_choices, sequence_length)) –

Indices of input sequence tokens in the vocabulary.

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

What are input IDs?

• attention_mask (torch.FloatTensor of shape (batch_size, num_choices, 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.

• token_type_ids (torch.LongTensor of shape (batch_size, num_choices, sequence_length), optional) –

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, num_choices, sequence_length), optional) –

Indices of positions of each input sequence tokens in the position embeddings. Selected in the range [0, config.max_position_embeddings - 1].

What are position IDs?

• 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, num_choices, 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.

• labels (torch.LongTensor of shape (batch_size,), optional) – Labels for computing the multiple choice classification loss. Indices should be in [0, ..., num_choices-1] where num_choices is the size of the second dimension of the input tensors. (See input_ids above)

Returns

A MultipleChoiceModelOutput 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 (MegatronBertConfig) and inputs.

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

• logits (torch.FloatTensor of shape (batch_size, num_choices)) – num_choices is the second dimension of the input tensors. (see input_ids above).

Classification 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

MultipleChoiceModelOutput or tuple(torch.FloatTensor)

Example:

>>> from transformers import BertTokenizer, MegatronBertForMultipleChoice
>>> import torch

>>> tokenizer = BertTokenizer.from_pretrained('nvidia/megatron-bert-cased-345m')
>>> model = MegatronBertForMultipleChoice.from_pretrained('nvidia/megatron-bert-cased-345m')

>>> prompt = "In Italy, pizza served in formal settings, such as at a restaurant, is presented unsliced."
>>> choice0 = "It is eaten with a fork and a knife."
>>> choice1 = "It is eaten while held in the hand."
>>> labels = torch.tensor(0).unsqueeze(0)  # choice0 is correct (according to Wikipedia ;)), batch size 1

>>> encoding = tokenizer([prompt, prompt], [choice0, choice1], return_tensors='pt', padding=True)
>>> outputs = model(**{k: v.unsqueeze(0) for k,v in encoding.items()}, labels=labels)  # batch size is 1

>>> # the linear classifier still needs to be trained
>>> loss = outputs.loss
>>> logits = outputs.logits


## MegatronBertForTokenClassification¶

class transformers.MegatronBertForTokenClassification(config)[source]

MegatronBert Model with a token classification head on top (a linear layer on top of the hidden-states output) e.g. for Named-Entity-Recognition (NER) tasks.

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.

Parameters

config (MegatronBertConfig) – 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, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None)[source]

The MegatronBertForTokenClassification 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.

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

What are input IDs?

• 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.

• token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) –

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) –

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) –

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.

• labels (torch.LongTensor of shape (batch_size, sequence_length), optional) – Labels for computing the token classification loss. Indices should be in [0, ..., config.num_labels - 1].

Returns

A TokenClassifierOutput 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 (MegatronBertConfig) and inputs.

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

• logits (torch.FloatTensor of shape (batch_size, sequence_length, config.num_labels)) – Classification 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

TokenClassifierOutput or tuple(torch.FloatTensor)

Example:

>>> from transformers import BertTokenizer, MegatronBertForTokenClassification
>>> import torch

>>> tokenizer = BertTokenizer.from_pretrained('nvidia/megatron-bert-cased-345m')
>>> model = MegatronBertForTokenClassification.from_pretrained('nvidia/megatron-bert-cased-345m')

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

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


class transformers.MegatronBertForQuestionAnswering(config)[source]

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

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.

Parameters

config (MegatronBertConfig) – 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, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, start_positions=None, end_positions=None, output_attentions=None, output_hidden_states=None, return_dict=None)[source]

The MegatronBertForQuestionAnswering 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.

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

What are input IDs?

• 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.

• token_type_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) –

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) –

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) –

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.

• start_positions (torch.LongTensor of shape (batch_size,), optional) – Labels for position (index) of the start of the labelled span for computing the token classification loss. Positions are clamped to the length of the sequence (sequence_length). Position outside of the sequence are not taken into account for computing the loss.

• end_positions (torch.LongTensor of shape (batch_size,), optional) – Labels for position (index) of the end of the labelled span for computing the token classification loss. Positions are clamped to the length of the sequence (sequence_length). Position outside of the sequence are not taken into account for computing the loss.

Returns

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

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

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

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

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

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

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

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

Return type

QuestionAnsweringModelOutput or tuple(torch.FloatTensor)

Example:

>>> from transformers import BertTokenizer, MegatronBertForQuestionAnswering
>>> import torch

>>> tokenizer = BertTokenizer.from_pretrained('nvidia/megatron-bert-cased-345m')