Method num_layers

tensorrt_llm/runtime/generation.py:1175–1178

(self)

    @property
    def num_layers(self):
        assert self._model_config.num_layers % self.mapping.pp_size == 0, \
            f"num_layers {self._model_config.num_layers} must be a multiple of pipeline parallelism size {self.mapping.pp_size}"
        return self._model_config.num_layers // self.mapping.pp_size

    @property
    def first_layer(self):