Hugging Face's logo

Transformers documentation

GraniteMoeSWA

You are viewing main version, which requires installation from source. If you'd like regular pip install, checkout the latest stable version (v5.14.0).
Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started

This model was contributed to Hugging Face Transformers on 2026-07-29.

FlashAttention Tensor parallelism

GraniteMoeSWA

GraniteMoeSWA combines the mixture-of-experts (MoE) architecture of GraniteMoeShared with the sliding-window attention and learnable attention sinks of GraniteSWA:

  • Mixture of experts. Each block routes every token to a subset of experts (num_experts_per_tok of num_local_experts). Optional shared experts are supported but disabled by default (shared_intermediate_size=0); set it to a positive value to enable them.
  • Per-layer sliding window attention. Each layer is either "full_attention" or "sliding_attention" (configured by layer_types). By default every fourth layer (i % 4 == 0) keeps full attention and the rest attend only to the most recent sliding_window tokens.
  • Learnable per-head attention sinks. Each head learns a scalar sink that rescales its attention output by sigmoid(logsumexp(attn_logits) - sink), equivalent to appending a single extra learnable logit to the softmax denominator (the attention-sink mechanism used by GPT-OSS).

SDPA is not supported because the attention sink cannot be expressed through torch.nn.functional.scaled_dot_product_attention. Supported backends are:

  • Training + inference: "eager", "flex_attention" (preferred for training)
  • Inference: "flash_attention_3" (via vLLM FA3 ‘hub’ kernel — also the fallback when FlashAttention-3 is not installed but kernels is), "flash_attention_4"

The example below demonstrates how to generate text with Pipeline or the AutoModelForCausalLM class.

Pipeline
AutoModel
from transformers import pipeline


pipe = pipeline(
    task="text-generation",
    model="ibm-granite/granite-swash-3b-a600m",
)
pipe("Explain quantum computing in simple terms", max_new_tokens=50)

GraniteMoeSWAConfig

class transformers.GraniteMoeSWAConfig

< >

( transformers_version: str | None = Nonearchitectures: list[str] | None = Noneoutput_hidden_states: bool | None = Falsereturn_dict: bool | None = Truedtype: typing.Union[str, ForwardRef('torch.dtype'), NoneType] = Nonechunk_size_feed_forward: int = 0is_encoder_decoder: bool = Falseid2label: dict[int, str] | dict[str, str] | None = Nonelabel2id: dict[str, int] | dict[str, str] | None = Noneproblem_type: typing.Optional[typing.Literal['regression', 'single_label_classification', 'multi_label_classification']] = Nonevocab_size: int = 32000hidden_size: int = 4096intermediate_size: int = 11008num_hidden_layers: int = 32num_attention_heads: int = 32num_key_value_heads: int | None = Nonehidden_act: str = 'silu'max_position_embeddings: int = 2048initializer_range: float = 0.02rms_norm_eps: float = 1e-06use_cache: bool = Truepad_token_id: int | None = Nonebos_token_id: int | None = 1eos_token_id: int | list[int] | None = 2tie_word_embeddings: bool = Falserope_parameters: transformers.modeling_rope_utils.RopeParameters | dict | None = Noneattention_bias: bool = Falseattention_dropout: float | int | None = 0.0embedding_multiplier: float | int | None = 1.0logits_scaling: float | int | None = 1.0residual_multiplier: float | int | None = 1.0attention_multiplier: float | int | None = 1.0num_local_experts: int | None = 8num_experts_per_tok: int | None = 2output_router_logits: bool | None = Falserouter_aux_loss_coef: float | None = 0.001shared_intermediate_size: int = 0sliding_window: int | None = 128layer_types: list[str] | None = Nonelayer_rope_theta: list[float | int] | None = None )

Parameters

  • vocab_size (int, optional, defaults to 32000) — Vocabulary size of the model. Defines the number of different tokens that can be represented by the input_ids.
  • hidden_size (int, optional, defaults to 4096) — Dimension of the hidden representations.
  • intermediate_size (int, optional, defaults to 11008) — Dimension of the MLP representations.
  • num_hidden_layers (int, optional, defaults to 32) — Number of hidden layers in the Transformer decoder.
  • num_attention_heads (int, optional, defaults to 32) — Number of attention heads for each attention layer in the Transformer decoder.
  • num_key_value_heads (int, optional) — This is the number of key_value heads that should be used to implement Grouped Query Attention. If num_key_value_heads=num_attention_heads, the model will use Multi Head Attention (MHA), if num_key_value_heads=1 the model will use Multi Query Attention (MQA) otherwise GQA is used. When converting a multi-head checkpoint to a GQA checkpoint, each group key and value head should be constructed by meanpooling all the original heads within that group. For more details, check out this paper. If it is not specified, will default to num_attention_heads.
  • hidden_act (str, optional, defaults to silu) — The non-linear activation function (function or string) in the decoder. For example, "gelu", "relu", "silu", etc.
  • max_position_embeddings (int, optional, defaults to 2048) — The maximum sequence length that this model might ever be used with.
  • initializer_range (float, optional, defaults to 0.02) — The standard deviation of the truncated_normal_initializer for initializing all weight matrices.
  • rms_norm_eps (float, optional, defaults to 1e-06) — The epsilon used by the rms normalization layers.
  • 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 or when the model is a decoder-only generative model.
  • pad_token_id (int, optional) — Token id used for padding in the vocabulary.
  • bos_token_id (int, optional, defaults to 1) — Token id used for beginning-of-stream in the vocabulary.
  • eos_token_id (Union[int, list[int]], optional, defaults to 2) — Token id used for end-of-stream in the vocabulary.
  • tie_word_embeddings (bool, optional, defaults to False) — Whether to tie weight embeddings according to model’s tied_weights_keys mapping.
  • rope_parameters (Union[~modeling_rope_utils.RopeParameters, dict], optional) — Dictionary containing the configuration parameters for the RoPE embeddings. The dictionary should contain a value for rope_theta and optionally parameters used for scaling in case you want to use RoPE with longer max_position_embeddings.
  • attention_bias (bool, optional, defaults to False) — Whether to use a bias in the query, key, value and output projection layers during self-attention.
  • attention_dropout (Union[float, int], optional, defaults to 0.0) — The dropout ratio for the attention probabilities.
  • embedding_multiplier (Union[float, int], optional, defaults to 1.0) — Scaling factor applied to the word embeddings. Used to scale the embeddings relative to the hidden size.
  • logits_scaling (Union[float, int], optional, defaults to 1.0) — Scaling factor applied to the output logits before computing the probability distribution.
  • residual_multiplier (Union[float, int], optional, defaults to 1.0) — Scaling factor applied to the residual connections.
  • attention_multiplier (Union[float, int], optional, defaults to 1.0) — Scaling factor applied to the attention weights.
  • num_local_experts (int, optional, defaults to 8) — Number of local experts on each device. num_experts should be divisible by num_local_experts.
  • num_experts_per_tok (int, optional, defaults to 2) — Number of experts to route each token to. This is the top-k value for the token-choice routing.
  • output_router_logits (bool, optional, defaults to False) — Whether or not the router logits should be returned by the model. Enabling this will also allow the model to output the auxiliary loss, including load balancing loss and router z-loss.
  • router_aux_loss_coef (float, optional, defaults to 0.001) — Auxiliary load balancing loss coefficient. Used to penalize uneven expert routing in MoE models.
  • shared_intermediate_size (int, optional, defaults to 0) — intermediate size for shared experts. Defaults to 0, which disables the shared experts.
  • sliding_window (int, optional, defaults to 128) — Size of the sliding attention window used by layers whose layer_types entry is "sliding_attention".
  • layer_types (list[str], optional) — Per-layer attention type, each either "full_attention" or "sliding_attention". When None, every fourth layer (i % 4 == 0) uses full attention and the rest use sliding window attention.
  • layer_rope_theta (list[float], optional) — Per-layer RoPE base (theta) frequency. 0 sets NoPE (no positional embedding) for that layer. Overrides global rope_parameters["rope_theta"], which is only used when this list is not provided or specified (layer_rope_theta = None).

This is the configuration class to store the configuration of a GraniteMoeSWAModel. It is used to instantiate a Granitemoe Swa 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 ibm-granite/granite-swash-3b-a600m

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

>>> from transformers import GraniteMoeSWAModel, GraniteMoeSWAConfig

>>> # Initializing a GraniteMoeSWA configuration
>>> configuration = GraniteMoeSWAConfig()

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

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

GraniteMoeSWAModel

class transformers.GraniteMoeSWAModel

< >

( config: GraniteMoeSWAConfig )

Parameters

  • config (GraniteMoeSWAConfig) — Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the from_pretrained() method to load the model weights.

The bare Granitemoe Swa Model 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.

forward

< >

( input_ids: typing.Optional[torch.LongTensor] = Noneattention_mask: typing.Optional[torch.Tensor] = Noneposition_ids: typing.Optional[torch.LongTensor] = Nonepast_key_values: transformers.cache_utils.Cache | None = Noneinputs_embeds: typing.Optional[torch.FloatTensor] = Noneuse_cache: bool | None = None**kwargs: Unpack ) → MoeModelOutputWithPast or tuple(torch.FloatTensor)

Parameters

  • input_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of input sequence tokens in the vocabulary. Padding will be ignored by default.

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

    What are input IDs?

  • attention_mask (torch.Tensor 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.

    What are attention masks?

  • 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?

  • past_key_values (~cache_utils.Cache, optional) — Pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention blocks) that can be used to speed up sequential decoding. This typically consists in the past_key_values returned by the model at a previous stage of decoding, when use_cache=True or config.use_cache=True.

    Only Cache instance is allowed as input, see our kv cache guide. If no past_key_values are passed, DynamicCache will be initialized by default.

    The model will output the same cache format that is fed as input.

    If past_key_values are used, the user is expected to input only unprocessed input_ids (those that don’t have their past key value states given to this model) of shape (batch_size, unprocessed_length) instead of all input_ids of shape (batch_size, sequence_length).

  • inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) — Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the model’s internal embedding lookup matrix.
  • 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

MoeModelOutputWithPast or tuple(torch.FloatTensor)

A MoeModelOutputWithPast 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 (GraniteMoeSWAConfig) and inputs.

The GraniteMoeSWAModel forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

  • 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.

  • past_key_values (Cache, optional, returned when use_cache=True is passed or when config.use_cache=True) — It is a Cache instance. For more details, see our kv cache guide.

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

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

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

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

  • router_logits (tuple(torch.FloatTensor), optional, returned when output_router_probs=True and config.add_router_probs=True is passed or when config.output_router_probs=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, sequence_length, num_experts).

    Raw router logtis (post-softmax) that are computed by MoE routers, these terms are used to compute the auxiliary loss for Mixture of Experts models.

GraniteMoeSWAForCausalLM

class transformers.GraniteMoeSWAForCausalLM

< >

( config: GraniteMoeSWAConfig )

Parameters

  • config (GraniteMoeSWAConfig) — 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 Granitemoe Swa Model for causal language modeling.

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.

forward

< >

( input_ids: typing.Optional[torch.LongTensor] = Noneattention_mask: typing.Optional[torch.Tensor] = Noneposition_ids: typing.Optional[torch.LongTensor] = Nonepast_key_values: transformers.cache_utils.Cache | None = Noneinputs_embeds: typing.Optional[torch.FloatTensor] = Nonelabels: typing.Optional[torch.LongTensor] = Noneoutput_router_logits: bool | None = Nonelogits_to_keep: typing.Union[int, torch.Tensor] = 0**kwargs ) → MoeCausalLMOutputWithPast or tuple(torch.FloatTensor)

Parameters

  • input_ids (torch.LongTensor of shape (batch_size, sequence_length), optional) — Indices of input sequence tokens in the vocabulary. Padding will be ignored by default.

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

    What are input IDs?

  • attention_mask (torch.Tensor 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.

    What are attention masks?

  • 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?

  • past_key_values (~cache_utils.Cache, optional) — Pre-computed hidden-states (key and values in the self-attention blocks and in the cross-attention blocks) that can be used to speed up sequential decoding. This typically consists in the past_key_values returned by the model at a previous stage of decoding, when use_cache=True or config.use_cache=True.

    Only Cache instance is allowed as input, see our kv cache guide. If no past_key_values are passed, DynamicCache will be initialized by default.

    The model will output the same cache format that is fed as input.

    If past_key_values are used, the user is expected to input only unprocessed input_ids (those that don’t have their past key value states given to this model) of shape (batch_size, unprocessed_length) instead of all input_ids of shape (batch_size, sequence_length).

  • inputs_embeds (torch.FloatTensor of shape (batch_size, sequence_length, hidden_size), optional) — Optionally, instead of passing input_ids you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert input_ids indices into associated vectors than the model’s internal embedding lookup matrix.
  • labels (torch.LongTensor of shape (batch_size, sequence_length), optional) — Labels for computing the masked language modeling loss. Indices should either be in [0, ..., config.vocab_size] or -100 (see input_ids docstring). Tokens with indices set to -100 are ignored (masked), the loss is only computed for the tokens with labels in [0, ..., config.vocab_size].
  • output_router_logits (bool, optional) — Whether or not to return the logits of all the routers. They are useful for computing the router loss, and should not be returned during inference.
  • logits_to_keep (Union[int, torch.Tensor], optional, defaults to 0) — If an int, compute logits for the last logits_to_keep tokens. If 0, calculate logits for all input_ids (special case). Only last token logits are needed for generation, and calculating them only for that token can save memory, which becomes pretty significant for long sequences or large vocabulary size. If a torch.Tensor, must be 1D corresponding to the indices to keep in the sequence length dimension. This is useful when using packed tensor format (single dimension for batch and sequence length).

Returns

MoeCausalLMOutputWithPast or tuple(torch.FloatTensor)

A MoeCausalLMOutputWithPast 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 (None) and inputs.

The GraniteMoeSWAForCausalLM forward method, overrides the __call__ special method.

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the pre and post processing steps while the latter silently ignores them.

  • 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).

  • aux_loss (torch.FloatTensor, optional, returned when labels is provided) — aux_loss for the sparse modules.

  • router_logits (tuple(torch.FloatTensor), optional, returned when output_router_probs=True and config.add_router_probs=True is passed or when config.output_router_probs=True) — Tuple of torch.FloatTensor (one for each layer) of shape (batch_size, sequence_length, num_experts).

    Raw router logtis (post-softmax) that are computed by MoE routers, these terms are used to compute the auxiliary loss for Mixture of Experts models.

  • past_key_values (Cache, optional, returned when use_cache=True is passed or when config.use_cache=True) — It is a Cache instance. For more details, see our kv cache guide.

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

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

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

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

Example:

>>> from transformers import AutoTokenizer, GraniteMoeSWAForCausalLM

>>> model = GraniteMoeSWAForCausalLM.from_pretrained("ibm/PowerMoE-3b")
>>> tokenizer = AutoTokenizer.from_pretrained("ibm/PowerMoE-3b")

>>> prompt = "Hey, are you conscious? Can you talk to me?"
>>> inputs = tokenizer(prompt, return_tensors="pt")

>>> # Generate
>>> generate_ids = model.generate(inputs.input_ids, max_length=30)
>>> tokenizer.batch_decode(generate_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)[0]
"Hey, are you conscious? Can you talk to me?\nI'm not conscious, but I can talk to you."
Update on GitHub

←GraniteMoeShared GraniteSWA→