WeightNameResolver

Resolves module parameter paths (from the DSL module tree) to weight tensor names (from model files like GGUF or SafeTensors).

The module tree produces paths like: "MLP/blk.0/attn/attn.q_proj.weight"

Different model formats use different tensor naming conventions: GGUF: "blk.0.attn_q.weight" SafeTensors: "model.layers.0.self_attn.q_proj.weight"

Implementations of this interface translate between the two.

Functions

abstract fun resolve(modulePath: String, paramName: String): String?

Resolve a module parameter path to the tensor name in the model file.