Struct Glm5NextWeightLoader 

pub struct Glm5NextWeightLoader;

Trait Implementations

impl ModelWeightLoader for Glm5NextWeightLoader

fn binds_vision_encoder(&self) -> bool

2026-09-25: True only when metrale_config::glm_vision_enabled() (METRALE_GLM_VISION). The GLM-5.3 config parser reads the same function before it parses vision_config; this method takes no config, so the gate is the environment.

fn load_vision_encoder( &self, store: &WeightStore, config: &ModelConfig, gpu: &dyn GpuBackend, ) -> Result<Option<VisionTower>>

2026-09-25: Bind the model.visual.* tower; see crate::weight_loader::glm5_next_vision.

fn defer_predicate(&self, config: &ModelConfig) -> Option<DeferHook>

2026-09-25: Keep the MTP layer’s BF16 routed-expert weights off the device (is_full_width_mtp_expert); bind_expert quantises them from disk. The layer is num_hidden_layers. A checkpoint whose MTP experts are U8 defers nothing.

fn supports_tp(&self) -> bool

2026-09-25: DSA, KDA and the MLP all shard under TP (see glm5_next_load.rs); routed experts are also split by EP (local_expert_range).

fn load_mtp_weights( &self, _store: &WeightStore, _config: &ModelConfig, _gpu: &dyn GpuBackend, ) -> Result<Option<MtpWeights>>

2026-09-25: None: the GLM-5.3 MTP layer is loaded by glm5_next_mtp::load_glm5next_mtp_module, not as MtpWeights.

fn prune_after_load( &self, store: &mut WeightStore, config: &ModelConfig, gpu: &dyn GpuBackend, ) -> Result<()>

2026-09-25: Free the store tensors is_reuploaded matches, then those is_quantized_expert_weight matches.

fn load_layers( &self, store: &WeightStore, config: &ModelConfig, gpu: &dyn GpuBackend, _layer_kv_dtypes: &[KvCacheDtype], ) -> Result<Vec<Box<dyn TransformerLayer>>>

2026-09-25: Load all transformer layers from the weight store.

fn load_embedding( &self, store: &WeightStore, _config: &ModelConfig, _gpu: &dyn GpuBackend, ) -> Result<DenseWeight>

fn load_final_norm( &self, store: &WeightStore, _config: &ModelConfig, _gpu: &dyn GpuBackend, ) -> Result<DenseWeight>

2026-09-25: Load the final RMSNorm weight used before the LM head.

fn load_lm_head( &self, store: &WeightStore, _config: &ModelConfig, _gpu: &dyn GpuBackend, ) -> Result<DenseWeight>

fn precision_schedule(&self, _config: &ModelConfig) -> PrecisionSchedule

2026-09-25: Per-(layer, role) weight precision schedule. The default is PrecisionSchedule::default(), in which every role is Dtype::Inherit.

fn load_ngram_embedding( &self, _store: &WeightStore, _config: &ModelConfig, _gpu: &dyn GpuBackend, _max_tokens: usize, ) -> Result<Option<NgramEmbedding>>

2026-09-25: Build the n-gram embedding, for an architecture that fuses hashed n-gram lookups into the input embedding (the LongCat loader). The default None leaves the plain embed_tokens gather in place.

fn load_mtp_weights_multi( &self, store: &WeightStore, config: &ModelConfig, gpu: &dyn GpuBackend, ) -> Result<Vec<MtpWeights>>

2026-09-25: Load every MTP module; factory::build calls this one. The default wraps load_mtp_weights (zero or one module). The MiniMax and Step3p7 loaders override it.

fn kv_layer_dims(&self, _config: &ModelConfig) -> Vec<(usize, usize)>

2026-09-25: Per-attention-layer (num_kv_heads, head_dim), indexed like layer_kv_dtypes, for models whose attention geometry differs per layer (the Gemma-4 loader). The default is empty: every layer uses the global dims.

fn load_dflash_weights( &self, _drafter_store: &WeightStore, _config: &ModelConfig, _gpu: &dyn GpuBackend, _tp_size: usize, ) -> Result<Option<DflashWeights>>

2026-09-25: Load DFlash drafter weights from the drafter’s own WeightStore. The default returns None, and no loader overrides it: factory::build loads the drafter with the free function dflash_loader::load_dflash_weights.

fn load_lora_adapters( &self, adapters: &[LoraAdapterInput<'_>], config: &ModelConfig, gpu: &dyn GpuBackend, max_loras: usize, max_lora_rank: usize, ) -> Result<Option<LoraWeights>>

2026-09-25: Load startup-static PEFT LoRA adapters into the LoRA pool, through metrale_model_layers::lora::load_lora_adapters_multi. factory::build calls it before the buffer arena and the KV budget, so the pool counts as used memory.

fn reads_expert_table_plan(&self) -> bool

2026-10-02: Does this loader read the serve’s MoE expert-table decision (metrale_model_layers::layers::MoeExpertTables)? The serve plans and publishes it before load only for a loader that does.