In this work, we propose a variable-rate scheme for deep video compression, which can achieve continuously variable rate by a single model. The key idea is to use the R-D tradeoff parameter \(\lambda\) as the conditional parameter to control the bitrate. The scheme is developed on DVC, which jointly learns motion estimation, motion compression, motion compensation, and residual compression functions. In this framework, the motion and residual compression auto-encoders are critical for the rate adaptation because they generate the final bitstream directly. Inspired by the recent work on deep variable-rate image compression \cite{choi2019variable}, we propose to use the conditional auto-encoders, which are deeply modulated by \(\lambda\) via scaling-networks, to achieve the basic rate adaptation. However, since other complicated modules, i.e., the motion estimation and motion compensation, also affect the final bitrate indirectly, the basic rate adaptation still has a certain compression performance loss compared with the fixed-rate models. To address this, we propose to add the R-D tradeoff parameter map (\(\lambda\) map) to the inputs of the two modules as a conditional map. Finally, we use a multi-rate-distortion loss function together with a step-by-step training strategy to optimize the entire scheme. The experiments show that the proposed scheme achieves continuously variable rate by a single model with almost the same compression efficiency as multiple fixed-rate models. The additional parameters and computation of our model are negligible when compared with a single fixed-rate model.