Jong Wook Kim
Jong Wook Kim
It'll depend on the platform and the [start method](https://docs.python.org/3/library/multiprocessing.html#contexts-and-start-methods), but in general you shouldn't expect CUDA-related code to work across the process boundary. Initialize the model in the process that...
Hi, see https://github.com/openai/CLIP/blob/main/data/prompts.md#cifar100 where you can now find the class names and prompts for ensembling zero-shot predictions for CIFAR100.
There could be other things remaining on the VRAM that we can't fully control, such as optimized CUDA kernels. If you absolutely need to release the VRAM, I'd recommend running...
This should work after reinstalling the package from the latest commit. As an alternative, downloading the checkpoint from the URLs in [clip.py](https://github.com/openai/CLIP/blob/main/clip/clip.py) and loading its path with `clip.load` would also...
This is unfortunately true for most of CUDA programs. The runtime is free to select different implementations (for matrix multiplication, etc.) depending on the batch size, even between subsequent calls,...
You're correct that the argmax operation takes the representation at the EOT position. There's nothing inherently wrong with taking the average along the sequence dimension, but taking representation at the...
I am traveling, but let me try in the next 24 hours or so. I never knew that there can be WAV files with
It hangs with the Cython version as well, ``` In [1]: import pyrubberband.rubberband as pyrb In [2]: import numpy as np In [3]: a=np.random.randn(10000) In [4]: a Out[4]: array([ 0.31215944,...
This sounds great! I was also wondering how fast it'd be if [Triton's flash attention](https://github.com/openai/triton/blob/master/python/tutorials/06-fused-attention.py) was integrated, but unfortunately it's A100 only. Implementation-wise, I think we could subclass the class...
Hi Chenxi, thanks for building the demo page. I've been trying to keep this repo minimal with the core components only and directing third-party integrations to the [show and tell](https://github.com/openai/whisper/discussions/categories/show-and-tell)...