Dinghao Zhou

Results 152 comments of Dinghao Zhou

runtime memory leak

> Is there any solution to fix it? Maybe related issues: https://github.com/pytorch/pytorch/issues/27971#issuecomment-543067718

Training a non-streaming model

closed due to not activating for long time

Shorter decoding segments when using server/x86

reopen this issue if there are still problems

tfio.audio.AudioIOTensor is not thread safe

Any update on this?

不支持自带密码的转存链接

支持了吗

Watchdog caught collective operation timeout

may some oom occurs in training， ```bash num workers * gpus

[WIP] Text Injection ASR

Are there any relevant references? thx

tfio.audio.AudioIOTensor VS tf.audio.decode_wav

Any update on this?

[transformer] Add moe_noisy_gate

贴class下边 好奇完整的epoch跑完会咋样 这个作用是加速收敛呢 还是最终效果也会变好

[transformer] Add moe_noisy_gate

后边会支持模型并行， moe这里需要特殊的处理， 看到了这个 参考下 截个图放这里 ref：https://zhuanlan.zhihu.com/p/681154742