v4.2.0

• Add num_batches and trt_static_shape parameters.
• Remove trt_int8, trt_int8_sample_step and trt_int8_batch_size parameters.
• Improve performance by using separate streams for transfering tensors between CPU and GPU.