Use dtype="auto" instead of float16 and add embedding sanity check

Pin torch<2.7 for CUDA 12.0 compat on HF Spaces A100

Return dummy ImageProcessingMixin stub instead of None

Skip image processor loading for text-only embedding models

Add torchvision dep to fix image processor error for gemma3_text models

Fix image processor error for gemma3_text-based models (e.g. harrier)

Add microsoft to trusted publishers

Embedding Gemma

Support Cohere

trust nvidia

Float 16, and better polling

Background jobs and polling

Move to jobs-based execution

Refactor to use gr.Progress API and upgrade to Gradio 5

Add heartbeat mechanism to prevent SSE timeout during model loading

Fix: Remove unsupported time_limit parameter

Enable queue and set 30-minute timeout for long-running evaluations

Leaderboard default

Initialize comparison plot on page load

Set Leaderboard tab as default

Improve leaderboard loading with better error handling and force refresh

Merge remote Space repository

Add HF dataset integration and leaderboard

Setup HF space

initial commit

Merge branch 'main' of https://github.com/EliezerIsrael/RabHebBench

Uploade dataset to HF

Delete space_README.md

Update README to streamline corpus and model info

Initial Commit