Alessandro Cauduro

Building from sources for Mac Apple Silicon (ARM)

2

I was able to build from sources as per the instructions without any error on my mac Apple Silicon. But when I tried to run the tests, I got the...

Recommend way to have multiple visualizations on the same globe.

4

I want to create a globe with 3 different visualisations that are selected by the user to be viewed one at a time. But I want for this to be...

timed metadata support?

2

Is there any support for the media server to inject metadata into the stream, similar to what AWS IVS does? https://docs.aws.amazon.com/ivs/latest/userguide/metadata.html This is import to sync streaming with the user...

[Bug]: When loading two models into the same GPU with Docker, the second model requires more memory allocation than the first

1

### Your current environment How I`m loading the models: ``` docker run \ --runtime nvidia \ -e VLLM_USE_V1=1 \ --gpus 0 \ --ipc=host \ -v "${HF_HOME}:/root/.cache/huggingface" \ vllm/vllm-openai:latest \ --model...

Change Docker Entrypoint to vllm.entrypoints.openai.api_server + Blackwell support

3

To make it easier to transition from plain vanilla vLLM to vLLM with LMCache with Docker, it would be nice to use the same entrypoint as vLLM docker, which uses...

Support for a newer Pytorch / Blackwell

1

Tried running the project with RTX5090, but it doesn't work because Blackwell requires Pytorch 2.8 with CUDA 12.9 support. I tried upgrading torch, but then I had issues with the...

I've tried to generate TTS in Brazillian Portuguese with speaches.ai and compare with the same created by https://github.com/thewh1teagle/kokoro-onnx Audio generated via the Speaches.AI API sounds like a non native speaker...