DeepASR performance tuning code #8673
Conversation
reyoung
requested review from
jacquesqiao,
JiayiFeng,
QiJune,
tonyyang-svail,
helinwang,
kuke and
dzhwinter
| @@ -0,0 +1,21 @@ | |||
| backports.functools-lru-cache==1.5 | |||
There was a problem hiding this comment.
The directory name DeepASR should be lower-cased, deep_asr.
| @@ -0,0 +1,13 @@ | |||
| #!/bin/bash | |||
There was a problem hiding this comment.
Given that run_on_1card.sh is almost exactly the same as run_on_4cards.sh, it seems reasonable to merge them into one file with a command line parameter.
| @@ -0,0 +1,13 @@ | |||
| #!/bin/bash | |||
| export PYTHONUNBUFFERED=1 | |||
| export CUDA_VISIBLE_DEVICES=7 | |||
There was a problem hiding this comment.
Is the magic number 7 due to that the computer you are using has 8 GPU cards? Is it possible to make it less "magical" by, say, calling nvidia-smi or some other commands to retrieve the number of GPUs on the system?
| @@ -0,0 +1,13 @@ | |||
| #!/bin/bash | |||
| export PYTHONUNBUFFERED=1 | |||
| export CUDA_VISIBLE_DEVICES=4,5,6,7 | |||
There was a problem hiding this comment.
Is it possible to make the magic numbers 4,5,6,7 less magical?
|
Since you haven't replied for a long time, we have closed this issue/pr. |
No description provided.