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.qwen_image

Qwen-Image diffusion architecture for image generation.

QwenImageArchConfig

class max.pipelines.architectures.qwen_image.QwenImageArchConfig(*, pipeline_config, quantization_encoding=None)

source

Bases: ArchConfig

Pipeline-level config for QwenImage (implements ArchConfig; no KV cache).

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

QwenImageConfig

class max.pipelines.architectures.qwen_image.QwenImageConfig(*, config_file=None, section_name=None, patch_size=2, in_channels=64, out_channels=None, num_layers=60, attention_head_dim=128, num_attention_heads=24, joint_attention_dim=3584, guidance_embeds=False, axes_dims_rope=(16, 56, 56), rope_theta=10000, zero_cond_t=False, eps=1e-06, dtype=bfloat16, device=<factory>)

source

Bases: QwenImageConfigBase

Parameters:

• config_file (str | None)
• section_name (str | None)
• patch_size (int)
• in_channels (int)
• out_channels (int | None)
• num_layers (int)
• attention_head_dim (int)
• num_attention_heads (int)
• joint_attention_dim (int)
• guidance_embeds (bool)
• axes_dims_rope (tuple[int, ...])
• rope_theta (int)
• zero_cond_t (bool)
• eps (float)
• dtype (DType)
• device (DeviceRef)

generate()

static generate(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:

QwenImageConfigBase

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].

QwenImageConfigBase

class max.pipelines.architectures.qwen_image.QwenImageConfigBase(*, config_file=None, section_name=None, patch_size=2, in_channels=64, out_channels=None, num_layers=60, attention_head_dim=128, num_attention_heads=24, joint_attention_dim=3584, guidance_embeds=False, axes_dims_rope=(16, 56, 56), rope_theta=10000, zero_cond_t=False, eps=1e-06, dtype=bfloat16, device=<factory>)

source

Bases: MAXModelConfigBase

Parameters:

• config_file (str | None)
• section_name (str | None)
• patch_size (int)
• in_channels (int)
• out_channels (int | None)
• num_layers (int)
• attention_head_dim (int)
• num_attention_heads (int)
• joint_attention_dim (int)
• guidance_embeds (bool)
• axes_dims_rope (tuple[int, ...])
• rope_theta (int)
• zero_cond_t (bool)
• eps (float)
• dtype (DType)
• device (DeviceRef)

attention_head_dim

attention_head_dim: int

source

axes_dims_rope

axes_dims_rope: tuple[int, ...]

source

device

device: DeviceRef

source

dtype

dtype: DType

source

eps

eps: float

source

guidance_embeds

guidance_embeds: bool

source

in_channels

in_channels: int

source

joint_attention_dim

joint_attention_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].

num_attention_heads

num_attention_heads: int

source

num_layers

num_layers: int

source

out_channels

out_channels: int | None

source

patch_size

patch_size: int

source

rope_theta

rope_theta: int

source

zero_cond_t

zero_cond_t: bool

source

QwenImageTransformerModel

class max.pipelines.architectures.qwen_image.QwenImageTransformerModel(config, encoding, devices, weights, session)

source

Bases: ComponentModel

Parameters:

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

load_model()

load_model()

source

Load and return a runtime model instance.

Return type:

Callable[[…], Any]