Measure execution times of BERT models using WebGPU and WASM
Export Hugging Face models to ONNX
Measure over-refusal in LLMs using OR-Bench
Track, rank and evaluate open LLMs and chatbots
Display LLM benchmark leaderboard and info
Benchmark LLMs in accuracy and translation across languages
Explore and benchmark visual document retrieval models
Create and upload a Hugging Face model card
Evaluate code generation with diverse feedback types
Optimize and train foundation models using IBM's FMS
Evaluate reward models for math reasoning
Browse and submit LLM evaluations
Calculate memory needed to train AI models
WebGPU Embedding Benchmark is a tool designed to measure the execution times of BERT models using WebGPU and WebAssembly (WASM). It helps developers and researchers evaluate the performance of embedding models in web-based environments, leveraging modern graphics technologies for accelerated computations.
• WebGPU Acceleration: Leverages WebGPU for hardware-accelerated computations. • WASM Execution: Utilizes WebAssembly for efficient model inference. • Detailed Timing Measurements: Provides precise execution time metrics for model inference. • Cross-Platform Compatibility: Runs on modern web browsers supporting WebGPU. • Model Optimization Insights: Offers benchmarks to guide model optimization strategies. • Performance Comparison: Enables comparison of performance across different hardware setups.
What does WebGPU Embedding Benchmark measure?
It measures the execution time of BERT models using WebGPU and WASM, providing insights into performance bottlenecks.
Which browsers support WebGPU?
As of now, browsers like Chrome, Firefox, and Edge provide experimental or full support for WebGPU.
Why is WebGPU combined with WASM for this benchmark?
WebGPU offers hardware acceleration, while WASM provides efficient computation, making them a powerful combination for high-performance web-based model inference.