-
`libtorch` is easy to use. However, the size of its shared libraries is large, see below.
![80a](https://user-images.githubusercontent.com/5284924/177512166-780505e6-21b5-4ece-96cb-6e2553ca4ae7.png)
…
-
Hi,
When will you support modified beam search with LM shallow fusion and LODR in Chinese corpus, such as aishell?
-
I'm trying to run the RNNT training benchmark, but the Docker build results in a GPG error because of an unsigned repository. Is NVIDIA still supporting Ubuntu 18 with this repository?
```
training/…
-
# 왜?
논문을 스터디 & 발표함에 있어, 논문을 어떻게 읽고 이해할 것인지에 대한 "틀"을 확실하게 정의하고 진행하는 것이 좋을 것.
- 논문의 단순 번역이 아닌 이해가 우리의 목표.
- 논문을 이해하는 통일된 틀을 정의할 수 있다면, 그것은 스터디원들 사이의 소통에도 큰 도움이 될 것.
- 여러번의 시행착오를 통해 그런 틀을 정의할 수 …
-
有的时候想说,一长段的文字,一句按一次的这样上屏,好像很影响思路。建议在长按的时候,如果说话若有停顿,可以把停顿之前的内容先上屏。就像讯飞输入法一样。
-
opening this issue to track ideas for how to port Halide to nim.
This could be either using halide in Nim via:
* A1: a wrapper (eg using nimterop) or
* A2: re-implementing some ideas in nim directl…
-
Hi,
I'm currently trying to train a transducer model using rf. I use the `torch_amp="bfloat16"` option from previous setups. In the predictor I use a `rf.LayerNorm` followed by `rf.LSTM`. I think thi…
-
- [x] Install [warp-transducer](https://github.com/HawkAaron/warp-transducer) in Dockerfile - see https://github.com/ryanleary/mlperf-rnnt-ref/issues/1#issuecomment-559186425.
- [x] Correct `forward…
samgd updated
4 years ago
-
torch 1.8.0+cu111
cuda 11.1
```
`(base) jovyan@liubiao-0:~/apex$ pip install -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./
/opt/cond…
-
python 3.7.11
nvcc: 11.1, V11.1.74
Build cuda_11.1.TC455_06.29069683_0
pytorch:1.8.0+cu111
cudnn:8005
RTX3090
(python37) ldf@ldf-virtual-machine:~/.jupyter/qinghao/PanGu/panguAlpha_pytorch/ape…