-
Hi there!
I stumbled upon the very promising ContextNet which you already implemented in NeMo.
Before I start to train the model by myself, I was wondering how soon we can expect NVIDIA to upload…
-
### 1、故障描述(Description)
运行video_demo推理demo.mp4生成demo_gt.mp4没有问题,推理自己的视频就有问题。
报错:`RuntimeError: Sizes of tensors must match except in dimension 2. Got 46 and 45 (The offending index is 0)`
### 2、复…
-
Hi, I'm currently working on move Kaldi to the Android. After I compiled the Kaldi-andriod and run my ivector case on Android, it's 30 times slower than Linux.
The ivector running speed on Linux i…
-
How can I get a space(blank) it self in tokenizer vocab?
I am trying to finetune the stt_en_conformer_transducer_large model with grapheme in korean. I changed some code to load labels with chardata…
-
您好,
我复现了hdv2和hdv3模型,但是和您提供的结果总有一些差距。
配置超参数: weight_decay=1e-4, learing rate:3e-4 *mul ,vgg loss,还有 #146 中提到的数据增广方式,patch size 256,训练300epoch。
我看您在hdv2和v3提供的版本中,模型结构都没有变化,它们都有什么区别呢?除了我…
-
Hi, I'm trying to train conformer for the Lithuanian language.
I'm training on `NVIDIA Tesla P100` and I keep getting OOM after some time.
Using all default values from `examples/conformer/confi…
-
Hello,
I have been using Quartznet 15x5 model for some time and recently I have tried to use the sentencepieces model trained on my data to encode the text and finetune the network with this setup.…
-
/home/kyq/anaconda3/lib/python3.7/site-packages/torch/include/c10/util/Exception.h:314:13: note: declared here
inline void deprecated_AT_CHECK() {}
^
In file included from /home/kyq/a…
-
Hi,
I've just pulled the last source code from the repository. The speed of training has changed from 2s/batch to 1s/batch but memory goes up to 5GB and freezes my pc with only 8GB of RAM.
I cha…
-
1、"We recommend generating optical flow labels on 2X size images for better labels. ”
question: Shouldn't the output size of optical flow network be the same as the input image?and if I get optical…