|
Download progress/github/SpecForge/docs/benchmarks/dashboard.md from FasterDFlash/Hanrui: direct link, hf CLI and curl.
- Browser
- Download file 618 Bytes
-
https://huggingface.co/FasterDFlash/Hanrui/resolve/main/progress/github/SpecForge/docs/benchmarks/dashboard.md
- Command line
-
hf download hf://FasterDFlash/Hanrui/progress/github/SpecForge/docs/benchmarks/dashboard.md
-
curl -L -o dashboard.md https://huggingface.co/FasterDFlash/Hanrui/resolve/main/progress/github/SpecForge/docs/benchmarks/dashboard.md
618 Bytes
| # Interactive Benchmark Dashboard | |
| View the interactive benchmark dashboard to explore SpecForge performance results: | |
| **[🚀 Spec Bundle](../dashboard/index.html)** | |
| The dashboard displays the following key metrics: | |
| - **Acceptance Length**: Average number of tokens accepted per speculation step | |
| - **Throughput**: Output tokens generated per second (tokens/s) | |
| - **Speedup**: Performance improvement ratio over baseline | |
| ## Benchmark Datasets | |
| View results across multiple benchmarks: | |
| - MTBench | |
| - HumanEval | |
| - GSM8K | |
| - Math500 | |
| --- | |
| If the dashboard doesn't load, please ensure JavaScript is enabled in your browser. | |