For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).

Python module

max.pipelines.architectures.ideogram4

Ideogram 4 flow-matching text-to-image architecture.

Ideogram4ArchConfig

class max.pipelines.architectures.ideogram4.Ideogram4ArchConfig(*, pipeline_config: 'PipelineConfig', quantization_encoding: 'SupportedEncoding | None' = None)

source

Bases: ArchConfig

Parameters:

• pipeline_config (PipelineConfig)
• quantization_encoding (SupportedEncoding | None)

DEFAULT_ENCODING

DEFAULT_ENCODING: ClassVar[SupportedEncoding] = 'bfloat16'

source

SUPPORTED_ENCODINGS

SUPPORTED_ENCODINGS: ClassVar[set[SupportedEncoding]] = {'bfloat16'}

source

calculate_max_seq_len()

classmethod calculate_max_seq_len(huggingface_config, model_config)

source

Returns the resolved maximum sequence length.

Bounds or defaults the user’s model_config.max_length with the model’s own limits. Construction runs this once and stores the result on model_config.max_length; memory planning may lower it further, but only on the memory plan.

Parameters:

• huggingface_config (AutoConfig) – The HuggingFace config to read model bounds from.
• model_config (MAXModelConfig) – The model config whose max_length carries the user’s setting.

Return type:

int

get_max_seq_len()

get_max_seq_len()

source

Returns the effective maximum sequence length for the model.

For configs that store a deployment length, this is the value initialize received; for metadata-only configs it derives from the checkpoint.

Return type:

int

initialize()

classmethod initialize(pipeline_config, model_config=None, *, max_seq_len)

source

Initialize the config from a PipelineConfig.

Parameters:

• pipeline_config (PipelineConfig) – The pipeline configuration.
• model_config (MAXModelConfig | None) – The model configuration to read from. When None (the default), pipeline_config.model is used. Pass an explicit config (e.g. pipeline_config.draft_model) to initialize the arch config for a different model.
• max_seq_len (int) – The effective maximum sequence length to store on the config. The value is received, never derived here: the pipeline model passes the memory plan’s VRAM-clamped length, while memory planning (which runs before a plan exists) passes the construction-resolved model_config.max_length. Configs whose sequence length is pure model metadata (e.g. diffusion components) ignore it.

Return type:

Self

pipeline_config

pipeline_config: PipelineConfig

source

quantization_encoding

quantization_encoding: SupportedEncoding | None = None

source

Ideogram4Config

class max.pipelines.architectures.ideogram4.Ideogram4Config(*, config_file=None, section_name=None, emb_dim=4608, num_layers=34, num_heads=18, intermediate_size=12288, adaln_dim=512, norm_eps=1e-05, in_channels=128, llm_features_dim=53248, rope_theta=5000000.0, mrope_section=(24, 20, 20), dtype=bfloat16, device=<factory>)

source

Bases: MAXModelConfigBase

Architecture parameters for Ideogram4Transformer2DModel.

Defaults mirror ideogram-ai/ideogram-4-fp8 transformer/config.json and the reference ideogram4.modeling_ideogram4.Ideogram4Config.

Parameters:

• config_file (str | None)
• section_name (str | None)
• emb_dim (int)
• num_layers (int)
• num_heads (int)
• intermediate_size (int)
• adaln_dim (int)
• norm_eps (float)
• in_channels (int)
• llm_features_dim (int)
• rope_theta (float)
• mrope_section (tuple[int, ...])
• dtype (DType)
• device (DeviceRef)

adaln_dim

adaln_dim: int

source

device

device: DeviceRef

source

dtype

dtype: DType

source

emb_dim

emb_dim: int

source

head_dim

property head_dim: int

source

in_channels

in_channels: int

source

initialize_from_config()

classmethod initialize_from_config(config_dict, encoding, devices)

source

Parameters:

• config_dict (dict[str, Any])
• encoding (Literal['float32', 'float16', 'bfloat16', 'q4_k', 'q4_0', 'q6_k', 'float8_e4m3fn', 'float4_e2m1fnx2', 'float6_e2m3fn', 'gptq'])
• devices (list[Device])

Return type:

Self

intermediate_size

intermediate_size: int

source

llm_features_dim

llm_features_dim: int

source

model_config

model_config: ClassVar[ConfigDict] = {'arbitrary_types_allowed': True, 'extra': 'forbid', 'strict': False}

source

Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

mrope_section

mrope_section: tuple[int, ...]

source

norm_eps

norm_eps: float

source

num_heads

num_heads: int

source

num_layers

num_layers: int

source

rope_theta

rope_theta: float

source

Ideogram4Transformer2DModel

class max.pipelines.architectures.ideogram4.Ideogram4Transformer2DModel(config)

source

Bases: Module

Ideogram 4 flow-matching transformer producing velocity predictions.

Parameters:

config (Ideogram4Config)

forward()

forward(*args)

source

Defines the computation performed by the module.

Users must override this method in their subclass to define the module’s computation.

Parameters:

• *args (Tensor) – Positional arguments for the computation.
• **kwargs – Keyword arguments for the computation.

Returns:

The result of applying the module to the input.

Raises:

NotImplementedError – If the subclass does not override this method.

Return type:

tuple[Tensor, …]

input_types()

input_types()

source

Return type:

tuple[TensorType, …]

Ideogram4TransformerModel

class max.pipelines.architectures.ideogram4.Ideogram4TransformerModel(config, encoding, devices, weights)

source

Bases: ComponentModel

Loads + adapts one Ideogram 4 DiT checkpoint (cond or uncond branch).

Parameters:

• config (dict[str, Any])
• encoding (SupportedEncoding)
• devices (list[Device])
• weights (Weights)

load_model()

load_model()

source

Adapt the checkpoint into a bf16 state dict (no standalone compile).

The actual graph compilation happens once in the pipeline, where both branches are fused into a single denoise-step graph.

Return type:

None