gemma3n

type AltUp struct {
	CorrectionScale       ml.Tensor   `gguf:"altup_correct_scale.weight"`
	PredictionCoefficient *nn.Linear  `gguf:"altup_predict_coef"`
	CorrectionCoefficient *nn.Linear  `gguf:"altup_correct_coef"`
	Router                *nn.Linear  `gguf:"altup_router"`
	RouterNorm            *nn.RMSNorm `gguf:"altup_router_norm"`
}

type Laurel struct {
	LinearLeft     *nn.Linear  `gguf:"laurel_l"`
	LinearRight    *nn.Linear  `gguf:"laurel_r"`
	PostLaurelNorm *nn.RMSNorm `gguf:"laurel_post_norm"`
}

type Model struct {
	model.Base
	model.SentencePiece

	*TextModel
}

type PerLayerProjector struct {
	TokenEmbedding *TextScaledWordEmbedding `gguf:"per_layer_token_embd"`
	Projector      *nn.Linear               `gguf:"per_layer_model_proj"`
	Norm           *nn.RMSNorm              `gguf:"per_layer_proj_norm"`
}

type TextAttention struct {
	Query     *nn.Linear  `gguf:"attn_q"`
	QueryNorm *nn.RMSNorm `gguf:"attn_q_norm"`
	Key       *nn.Linear  `gguf:"attn_k"`
	KeyNorm   *nn.RMSNorm `gguf:"attn_k_norm"`
	Value     *nn.Linear  `gguf:"attn_v"`
	Output    *nn.Linear  `gguf:"attn_output"`
}

type TextLayer struct {
	*AltUp
	*Laurel

	AttentionNorm     *nn.RMSNorm `gguf:"attn_norm"`
	Attention         *TextAttention
	PostAttentionNorm *nn.RMSNorm `gguf:"post_attention_norm"`

	MLPNorm     *nn.RMSNorm `gguf:"ffn_norm"`
	MLP         *TextMLP
	PostMLPNorm *nn.RMSNorm `gguf:"post_ffw_norm"`

	PerLayerInputGate  *nn.Linear  `gguf:"inp_gate"`
	PerLayerProjection *nn.Linear  `gguf:"proj"`
	PostPerLayerNorm   *nn.RMSNorm `gguf:"post_norm"`
}

type TextMLP struct {
	Gate *nn.Linear `gguf:"ffn_gate"`
	Up   *nn.Linear `gguf:"ffn_up"`
	Down *nn.Linear `gguf:"ffn_down"`
}

type TextModel struct {
	TokenEmbedding *TextScaledWordEmbedding `gguf:"token_embd"`

	*PerLayerProjector

	AltupEmbd   *nn.Linear `gguf:"altup_proj"`
	AltupUnembd *nn.Linear `gguf:"altup_unembd_proj"`

	TextLayers []TextLayer `gguf:"blk"`
	OutputNorm *nn.RMSNorm `gguf:"output_norm"`
	Output     *nn.Linear  `gguf:"output,alt:token_embd"`

	TextOptions
}

type TextOptions struct {
	// contains filtered or unexported fields
}

type TextScaledWordEmbedding struct {
	*nn.Embedding
}