-
I have created the following PR to add support for Thai language.
I am in the process of creating a dataset to train the model but would love a PR review of the code first to make sure I am on the ri…
-
Hi, Thanks for your open-source code, I'm trying to reappear your code.
Although I complete the preprocess of nuscenes, I failed in training stage due to the unknown version of python library.
Coul…
-
I've encountered an error compiling a PyTorch extension using **`nvcc`**. Specifically, I'm on PyTorch **`2.0.1+cu117`** and CUDA **`11.8.0`**, and I've tried this on both **`A100`** and **`A40`** car…
-
I'm looking for a simple use case -- or test data conversions, to help define the base transformer objects:
- types like dates, numbers, addresses, phone-numbers
- codecs for various representations o…
-
So this may seem like a minor gripe, but the specification is currently very clear on the “Missing and non-applicable values MUST be coded as n/a” [overt caps in the original]. I am currently battli…
tsalo updated
4 years ago
-
I encountered multiple issues while trying to perform full fine-tuning of the LLaMA 3 8B model with DeepSpeed with A100-80GB x 2.
As a result, I decided to follow the DeepSpeed tutorial on Huggingfa…
qmin2 updated
2 weeks ago
-
I love the benchmarks by @faroit on audio read speeds. I too have tried a number of different libraries, and have generally been using scipy.io.wavfile.read() for its speed whenever possible, but sw…
-
Many trackers that produce skeleton information also produce contour information, yet WCON provides no way to share that information across trackers.
Additionally, some skeleton-finding routines pr…
-
Hi ! First of all thank you for your work and for answering the questions of everybody.
I would like to train my dataset and I made it to the Scannet format :
Own Dataset
|- train
|- …
-
**Guidelines**
*We are very excited to meet you at the 2020 OHBM Brainhack 🎉* *To submit a project, you need to be an attendee of the 2020 OHBM Brainhack. We ask you to register first over [here](h…