DiffusionConfig

class tfts.models.diffusion.DiffusionConfig(hidden_size: int = 64, num_layers: int = 3, num_attention_heads: int = 8, attention_probs_dropout_prob: float = 0.1, hidden_dropout_prob: float = 0.1, ffn_intermediate_size: int = 256, max_position_embeddings: int = 512, initializer_range: float = 0.02, layer_norm_eps: float = 1e-12, pad_token_id: int = 0, num_diffusion_steps: int = 1000, beta_start: float = 0.0001, beta_end: float = 0.02, **kwargs)

Bases: BaseConfig

Initializes the configuration for the Diffusion model with the specified parameters.

Parameters:

• hidden_size – Size of each attention head.

• num_layers – The number of stacked transformer layers.

• num_attention_heads – The number of attention heads.

• attention_probs_dropout_prob – Dropout rate for attention probabilities.

• hidden_dropout_prob – Dropout rate for hidden layers.

• ffn_intermediate_size – Size of the intermediate layer in the feed-forward network.

• max_position_embeddings – Maximum sequence length for positional embeddings.

• initializer_range – Standard deviation for weight initialization.

• layer_norm_eps – Epsilon for layer normalization.

• pad_token_id – ID for padding token.

• num_diffusion_steps – Number of diffusion steps.

• beta_start – Starting noise level.

• beta_end – Ending noise level.

Inherited-members:

Methods

from_dict(config_dict)
from_json(json_file)
from_pretrained(pretrained_model_name_or_path)
save_pretrained(save_directory)
to_dict()
to_json(json_file)
update(config_dict)

Attributes

attribute_map
model_type