Method init_kv_cache

qwen2_model.py:290–300

(
        self,
        max_batch_size: int,
        max_seq_len: int,
        device: torch.device,
        dtype: torch.dtype,
    )

        return output

    def init_kv_cache(
        self,
        max_batch_size: int,
        max_seq_len: int,
        device: torch.device,
        dtype: torch.dtype,
    ):
        for layer in self.layers:
            layer.self_attn.init_kv_cache(
                max_batch_size, max_seq_len, dtype=dtype, device=device
            )

    def del_kv_cache(self):
        for layer in self.layers: