easydel.modules.minimax_text_v1.modeling_minimax_text_01

easydel.modules.minimax_text_v1.modeling_minimax_text_01#

class easydel.modules.minimax_text_v1.modeling_minimax_text_01.GLU(*args: Any, **kwargs: Any)[source]#

Gated Linear Unit for MiniMax models.

class easydel.modules.minimax_text_v1.modeling_minimax_text_01.MiniMaxText01Attention(*args: Any, **kwargs: Any)[source]#

Multi-head attention layer for MiniMax Text models.

class easydel.modules.minimax_text_v1.modeling_minimax_text_01.MiniMaxText01BlockSparseTop2MLP(*args: Any, **kwargs: Any)[source]#

Sparse top-2 MLP expert for MiniMax Text MoE layers.

class easydel.modules.minimax_text_v1.modeling_minimax_text_01.MiniMaxText01DecoderLayer(*args: Any, **kwargs: Any)[source]#

Decoder layer for MiniMax Text models.

class easydel.modules.minimax_text_v1.modeling_minimax_text_01.MiniMaxText01ForCausalLM(*args: Any, **kwargs: Any)[source]#

MiniMax Text model with language modeling head for causal generation.

get_decoder()[source]#: Returns the decoder part of the model’s graph definition.

get_encoder()[source]#: Returns the encoder part of the model’s graph definition. Decoder-Only models don’t have an encoder.

class easydel.modules.minimax_text_v1.modeling_minimax_text_01.MiniMaxText01LightningAttention(*args: Any, **kwargs: Any)[source]#

Lightning attention mechanism for MiniMax Text models.

class easydel.modules.minimax_text_v1.modeling_minimax_text_01.MiniMaxText01MLP(*args: Any, **kwargs: Any)[source]#

Feedforward network for MiniMax Text models.

class easydel.modules.minimax_text_v1.modeling_minimax_text_01.MiniMaxText01Model(*args: Any, **kwargs: Any)[source]#

Base MiniMax Text model with embeddings and decoder layers.

get_decoder()[source]#: Returns the decoder part of the model’s graph definition.

get_encoder()[source]#: Returns the encoder part of the model’s graph definition. Decoder-Only models don’t have an encoder.

get_lm_head()[source]#: Returns the language model head of the module. Base Models don’t have a Language Model Head.

class easydel.modules.minimax_text_v1.modeling_minimax_text_01.MiniMaxText01SparseMoeBlock(*args: Any, **kwargs: Any)[source]#

Mixture of Experts block for MiniMax Text models.

easydel.modules.minimax_text_v1.modeling_minimax_text_01.compute_slops(nhd)[source]#: Generate ALiBi-style slopes for the provided number of attention heads.

easydel.modules.minimax_text_v1.modeling_minimax_text_01.get_activation_fn(activation)[source]#: Map activation name to a callable used inside MiniMax feed-forward blocks.