# GPT-J¶

## Overview¶

The GPT-J model was released in the kingoflolz/mesh-transformer-jax repository by Ben Wang and Aran Komatsuzaki. It is a GPT-2-like causal language model trained on the Pile dataset.

This model was contributed by Stella Biderman.

Tips:

• To load GPT-J in float32 one would need at least 2x model size CPU RAM: 1x for initial weights and another 1x to load the checkpoint. So for GPT-J it would take at least 48GB of CPU RAM to just load the model. To reduce the CPU RAM usage there are a few options. The torch_dtype argument can be used to initialize the model in half-precision. And the low_cpu_mem_usage argument can be used to keep the RAM usage to 1x. There is also a fp16 branch which stores the fp16 weights, which could be used to further minimize the RAM usage. Combining all this it should take roughly 12.1GB of CPU RAM to load the model.

>>> from transformers import GPTJForCausalLM
>>> import torch

>>> model = GPTJForCausalLM.from_pretrained("EleutherAI/gpt-j-6B", revision="float16", torch_dtype=torch.float16, low_cpu_mem_usage=True)

• The model should fit on 16GB GPU for inference. For training/fine-tuning it would take much more GPU RAM. Adam optimizer for example makes four copies of the model: model, gradients, average and squared average of the gradients. So it would need at least 4x model size GPU memory, even with mixed precision as gradient updates are in fp32. This is not including the activations and data batches, which would again require some more GPU RAM. So one should explore solutions such as DeepSpeed, to train/fine-tune the model. Another option is to use the original codebase to train/fine-tune the model on TPU and then convert the model to Transformers format for inference. Instructions for that could be found here

• Although the embedding matrix has a size of 50400, only 50257 entries are used by the GPT-2 tokenizer. These extra tokens are added for the sake of efficiency on TPUs. To avoid the mis-match between embedding matrix size and vocab size, the tokenizer for GPT-J contains 143 extra tokens <|extratoken_1|>... <|extratoken_143|>, so the vocab_size of tokenizer also becomes 50400.

### Generation¶

The generate() method can be used to generate text using GPT-J model.

>>> from transformers import AutoModelForCausalLM, AutoTokenizer
>>> model = AutoModelForCausalLM.from_pretrained("EleutherAI/gpt-j-6B")
>>> tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-j-6B")

>>> prompt = "In a shocking finding, scientists discovered a herd of unicorns living in a remote, " \
...          "previously unexplored valley, in the Andes Mountains. Even more surprising to the " \
...          "researchers was the fact that the unicorns spoke perfect English."

>>> input_ids = tokenizer(prompt, return_tensors="pt").input_ids

>>> gen_tokens = model.generate(input_ids, do_sample=True, temperature=0.9, max_length=100,)
>>> gen_text = tokenizer.batch_decode(gen_tokens)[0]


…or in float16 precision:

>>> from transformers import GPTJForCausalLM, AutoTokenizer
>>> import torch

>>> model = GPTJForCausalLM.from_pretrained("EleutherAI/gpt-j-6B", torch_dtype=torch.float16)
>>> tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-j-6B")

>>> prompt = "In a shocking finding, scientists discovered a herd of unicorns living in a remote, " \
...          "previously unexplored valley, in the Andes Mountains. Even more surprising to the " \
...          "researchers was the fact that the unicorns spoke perfect English."

>>> input_ids = tokenizer(prompt, return_tensors="pt").input_ids

>>> gen_tokens = model.generate(input_ids, do_sample=True, temperature=0.9, max_length=100,)
>>> gen_text = tokenizer.batch_decode(gen_tokens)[0]


## GPTJConfig¶

class transformers.GPTJConfig(vocab_size=50400, n_positions=2048, n_ctx=2048, n_embd=4096, n_layer=28, n_head=16, rotary_dim=64, n_inner=None, activation_function='gelu_new', resid_pdrop=0.0, embd_pdrop=0.0, attn_pdrop=0.0, layer_norm_epsilon=1e-05, initializer_range=0.02, scale_attn_weights=True, use_cache=True, bos_token_id=50256, eos_token_id=50256, **kwargs)[source]

This is the configuration class to store the configuration of a GPTJModel. It is used to instantiate a GPT-J 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-J gpt-j-6B 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 50400) – Vocabulary size of the GPT-J model. Defines the number of different tokens that can be represented by the inputs_ids passed when calling GPTJModel.

• n_positions (int, optional, defaults to 2048) – 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 2048) – Dimensionality of the causal mask (usually same as n_positions).

• n_embd (int, optional, defaults to 4096) – Dimensionality of the embeddings and hidden states.

• n_layer (int, optional, defaults to 28) – Number of hidden layers in the Transformer encoder.

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

• rotary_dim (int, optional, defaults to 64) – Number of dimensions in the embedding that Rotary Position Embedding is applied to.

• n_inner (int, optional, defaults to None) – Dimensionality of the inner feed-forward layers. None will set it to 4 times n_embd

• activation_function (str, optional, defaults to "gelu_new") – Activation function, to be selected in the list ["relu", "silu", "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 0.02) – The standard deviation of the truncated_normal_initializer for initializing all weight matrices.

• scale_attn_weights (bool, optional, defaults to True) – Scale attention weights by dividing by sqrt(hidden_size).

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

Example:

>>> from transformers import GPTJModel, GPTJConfig

>>> # Initializing a GPT-J 6B configuration
>>> configuration = GPTJConfig()

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

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


## GPTJModel¶

class transformers.GPTJModel(config)[source]

The bare GPT-J 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 (GPTJConfig) – 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_key_values=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, return_dict=None)[source]

The GPTJModel 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 transformers.GPTJTokenizer. 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.n_positions - 1].

What are position IDs?

• head_mask (torch.FloatTensor of shape (num_attention_heads,) or (n_layer, num_attention_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, n_ctx), optional) – Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the model’s internal embedding lookup matrix.

• output_attentions (bool, optional) – Whether or not to return the attentions tensors of all attention layers. See attentions under returned tensors for more detail.

• output_hidden_states (bool, optional) – Whether or not to return the hidden states of all layers. See hidden_states under returned tensors for more detail.

• return_dict (bool, optional) – Whether or not to return a ModelOutput instead of a plain tuple.

Returns

A BaseModelOutputWithPast 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 (GPTJConfig) 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.

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

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

• 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

BaseModelOutputWithPast or tuple(torch.FloatTensor)

Example:

>>> from transformers import GPT2Tokenizer, GPTJModel
>>> import torch

>>> tokenizer = GPT2Tokenizer.from_pretrained('EleutherAI/gpt-j-6B')
>>> model = GPTJModel.from_pretrained('EleutherAI/gpt-j-6B')

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

>>> last_hidden_states = outputs.last_hidden_state


## GPTJForCausalLM¶

class transformers.GPTJForCausalLM(config)[source]

The GPT-J 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 (GPTJConfig) – 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_key_values=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, return_dict=None)[source]

The GPTJForCausalLM 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 transformers.GPTJTokenizer. 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.n_positions - 1].

What are position IDs?

• head_mask (torch.FloatTensor of shape (num_attention_heads,) or (n_layer, num_attention_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, n_ctx), 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 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

A CausalLMOutputWithPast 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 (GPTJConfig) 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).

• 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))

Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see past_key_values 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

CausalLMOutputWithPast or tuple(torch.FloatTensor)

Example:

>>> import torch
>>> from transformers import GPT2Tokenizer, GPTJForCausalLM

>>> tokenizer = GPT2Tokenizer.from_pretrained('EleutherAI/gpt-j-6B')
>>> model = GPTJForCausalLM.from_pretrained('EleutherAI/gpt-j-6B')

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


## GPTJForSequenceClassification¶

class transformers.GPTJForSequenceClassification(config)[source]

The GPT-J Model transformer with a sequence classification head on top (linear layer).

GPTJForSequenceClassification uses the last token in order to do the classification, as other causal models (e.g. GPT, GPT-2, GPT-Neo) do.

Since it does classification on the last token, it requires to know the position of the last token. If a pad_token_id is defined in the configuration, it finds the last token that is not a padding token in each row. If no pad_token_id is defined, it simply takes the last value in each row of the batch. Since it cannot guess the padding tokens when inputs_embeds are passed instead of input_ids, it does the same (take the last value in each row of the batch).

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 (GPTJConfig) – 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_key_values=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, return_dict=None)[source]

The GPTJForSequenceClassification 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 transformers.GPTJTokenizer. 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.n_positions - 1].

What are position IDs?

• head_mask (torch.FloatTensor of shape (num_attention_heads,) or (n_layer, num_attention_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, n_ctx), 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 SequenceClassifierOutputWithPast 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 (GPTJConfig) 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).

• 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))

Contains pre-computed hidden-states (key and values in the self-attention blocks) that can be used (see past_key_values 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

SequenceClassifierOutputWithPast or tuple(torch.FloatTensor)

Example:

>>> from transformers import GPT2Tokenizer, GPTJForSequenceClassification
>>> import torch

>>> tokenizer = GPT2Tokenizer.from_pretrained('EleutherAI/gpt-j-6B')
>>> model = GPTJForSequenceClassification.from_pretrained('EleutherAI/gpt-j-6B')

>>> 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