Method kv_cache_clear

llama_cpp/_internals.py:292–296

(self)

        return llama_cpp.llama_pooling_type(self.ctx)

    def kv_cache_clear(self):
        # Embedding models with non-causal attention may not allocate memory.
        if self.memory is None:
            return
        llama_cpp.llama_memory_clear(self.memory, True)

    def kv_cache_seq_rm(self, seq_id: int, p0: int, p1: int) -> bool:
        assert self.memory is not None, "Memory is not initialized"