czy36mengfei

thanks，When gpu memory is sufficient, no matter how big the batchsize is, each reasoning time should not make much difference. The batch size becomes larger, the number of inferences becomes...

I made the following modifications to support different batch sizes. ``` def make_animation_batch(source_image, driving_video, generator, kp_detector, relative=True, adapt_movement_scale=True, cpu=False): with torch.no_grad(): predictions = [] source = torch.tensor( source_image[np.newaxis].astype(np.float32)).permute(0, 3, 1,...

My gpu is adequate.After running, I see that the gpu has several GB left.

学长好， cmake输出如下 -- The CXX compiler identification is GNU 5.4.0 -- Check for working C compiler: /usr/bin/cc -- Check for working C compiler: /usr/bin/cc -- works -- Detecting C compiler...

> @czy36mengfei 看一下python命令默认是哪个环境？tensorflow，TF_LIB都没找到 十分感谢您的解答。我发现CMakeLists里调用的是python3.5而不是默认的python，我把它改成python就可以了。 不过在跑test_op.py测试的时候出现了新的问题： *** Begin stack trace *** tensorflow::CurrentStackTrace() std::_Function_handler::_M_invoke(std::_Any_data const&, long long&&, std::_Any_data const&) tensorflow::thread::ThreadPool::Impl::ParallelFor(long long, long long, std::function) tensorflow::thread::ThreadPool::ParallelFor(long long, long long, std::function) tensorflow::Shard(int, tensorflow::thread::ThreadPool*, long...

请问按您的实验，batch_size的范围是多少，为多少的时候最好？batch_size小训练起来太慢了～ @LeeSureman

@LeeSureman 谢谢。 所以您里面的设置是您的代码中得分最高的batch_size？即ontonote，weibo时batch_size=10得分比batch_size为1高，resume的batch_size为1时得分最高？ `if args.dataset == 'ontonote': args.batch = 10 args.lr = 0.045 elif args.dataset == 'resume': args.batch = 1 args.lr = 0.015 elif args.dataset == 'weibo': args.batch = 10...

> 是的 好的，谢谢。

需要执行： pip install tfds-nightly==1.0.2.dev201904090105 来安装tfds，官方教程给的翻译数据集只有这个版本的tfds可以用

 你那边说的不支持是什么意思，功能无效吗？发一下具体代码。 我这边可以正常运行