Amit Moryossef
Amit Moryossef
This change partially addresses #770. I performed a crude find and replace: `margin-left` is `margin-inline-start` and `margin-right` is `margin-inline-end`. This might need to be done to other files, such as...
### Expected Behavior Before translating to Hebrew, wanted to check if RTL is supported. It can be better supported. Following #388 ### Current Behavior small popover: - buttons are not...
In tests, or when wanting to just debug something quickly, the process hangs on `import evaluate`. I benchmarked it on two devices: ```bash python -c "import time; start = time.time();...
# What does this PR do? ### 1 / Problem * `find_executable_batch_size()` currently **halves** the batch after every OOM. *Example:* request **128**, real limit **100** → old loop jumps straight...
When creating an algorithm with lstlisting, the numbers go outside of the margins: One must add `,xleftmargin=10pt` to get `aclpubcheck` to pass. I think this can be modified in the...
It seems like 2D attention works with flash attention, but not 4D attention masks. In this example, I have the same inputs except for the attention masks. The 4D attention...
### Feature request Support [MASK] token prediction for models such as ModernBERT ### Motivation This repository seems like an extremely fast way to run encoder only language models such as...
It would be very helpful if the repository included an official Dockerfile that captures all required dependencies for VACE and its preprocessing stack. Right now, VACE depends on multiple external...
Full reproduction: https://github.com/sign/image-latent-transformer/blob/main/examples/benchmark_image_encoder.py I am trying to find a model to encode lots of small images per batch. I started by using Huggingface Transformers, and creating a `ViTModel`: ```py ViTModel.from_pretrained("WinKawaks/vit-tiny-patch16-224")...