ML::GGUF::NomicBertMoE(B)
Constructors
new(gguf : GGUFFile, backend : B)
SourceClass methods
Instance methods
backend
Sourcedim
SourceBatch embed: pre-tokenize all texts, then sequential GPU forward passes (true parallel batching requires attention masking — future work)
Debug: embed with per-layer GPU vs CPU comparison
ffn_dim
Sourcehead_dim
Sourcemax_seq_len
Sourcemoe_every_n
Sourcen_experts
Sourcen_experts_used
Sourcen_heads
Sourcen_layers
Sourcerope_theta
SourceTokenize text → token IDs (delegates to SentencePiece unigram tokenizer)
vocab_size
Source