-
With the introduction of [graphql-java-async](https://github.com/graphql-java/graphql-java/pull/283) planned, it would be great if one could choose between generating blocking and non-blocking code.
…
-
Hi @lucidrains - thanks for your awesome work here. Great stuff as always.
I recently came across Google's new SingSong paper (https://arxiv.org/pdf/2301.12662.pdf), in which they adapt AudioLM fo…
smcio updated
2 months ago
-
When I input the following code:
```
import nlpaug.augmenter.word as naw
text = 'The quick brown fox jumps over the lazy dog .'
aug = naw.BackTranslationAug(
from_model_name='facebook/w…
-
I tried to test GPTQ's PPL metrics on the opt model via opt.py. The PPL metrics of the opt model are normal with the use of fake
quantization. However, when I try to place the opt_pack before the o…
-
Beautiful work as usual, thanks for this implementation.
I'm curious if you tried using this for a regression task? I have tried using TimeSFormer without success yet, I know the signal is there b…
-
Describe the bug
Running an error
Log output
***** Evaluating perplexity, Epoch 0/1 *****
Traceback (most recent call last):
File "main.py", line 345, in
main()
File "main.py", line 306, in m…
-
If it's running for a while, it will eventually crash due to a memory leak somewhere. It seems that the garbage collector can't delete things.
-
moving discussion from https://github.com/facebookresearch/audiocraft/issues/480 to here, as its specific to this fork.
-
Let me ask if train model with full nuscenes data, what is your configuration need, I have problem with too much RAM when training, my RAM is 13GB, I use colab or kaggle
Thank you very much
-
### 请提出你的问题 Please ask your question
paddle开启分布式训练时,如tp、pp均为2时,我想进行性能分析,查看板卡占用信息确实可以看到4张卡都被占用了,但是打印device时永远都是0卡,这怎样查看所有卡的信息并进行分析呢?