drbh
drbh
This PR is a work in progress and add support for Idefics3 in TGI. opening for transparency and feedback. This implementation uses the `AutoProcessor/Idefics3Processor` that will be added when this...
This PR adds a new image tag to include both the latest version and sha tag on images. This is helpful in understanding which version the latest changes derive from
This PR resolves some small issues with qwen2-vl. 1. doubles the size of `WARMUP_IMAGE_BASE64` from 20x20px to 40x40px (meets qwens minimal requirement without hacky fix) 2. removes hacky fix to...
This PR is a work in progress that explores adding support for video inputs with Qwen2-VL. Thank you @mfarre for getting this effort started. TODOS - [X] suport `video_url`s -...
This PR improve mapping of tool ids to responses and enables tool_call_ids to be included in templates Changes: 1. maps input `tool.id` to `tool_calls[0].id` for chosen tool if `tool.id` is...
This PR reopens #2982 to run CI
This PR adds a new `HuggingFaceProvider` to enable integration with [Inference Endpoints](https://endpoints.huggingface.co/)
This PR adds a new github workflow that aims to improve the release workflow. The workflow requires manually triggering and checks that the intended release semver version matches the version...
This PR adds the [fasterp](https://github.com/drbh/fasterp) tool as a new module. Since fasterp exposes the same interface as fastp the module is very similar. Please let me know if any changes...
This article explores three different ways to compute MoE and focuses on the "how" of MoE