Datasets:
Download web/streamora.example/tests/eval_config.json from Qwen/RecreationBench: direct link, hf CLI and curl.
- Browser
- Download file 868 Bytes
-
https://huggingface.co/datasets/Qwen/RecreationBench/resolve/main/web/streamora.example/tests/eval_config.json
- Command line
-
hf download hf://datasets/Qwen/RecreationBench/web/streamora.example/tests/eval_config.json
-
curl -L -o eval_config.json https://huggingface.co/datasets/Qwen/RecreationBench/resolve/main/web/streamora.example/tests/eval_config.json
868 Bytes
| { | |
| "scoring_mode": "tests_only", | |
| "test_weights": { | |
| "functional": 0.5, | |
| "visual": 0.5, | |
| "structural": 0.0, | |
| "quality": 0.0 | |
| }, | |
| "vlm_judge": { | |
| "enabled": true, | |
| "mode": "assertion", | |
| "max_concurrency": 8, | |
| "temperature": 0.0, | |
| "_runtime": "model / backend / base_url / api_key must be supplied at run time. They are deliberately absent here: they are properties of whoever runs the judge, not of this dataset. NOTE that the harness silently falls back to a default model if none is supplied — pass one explicitly." | |
| }, | |
| "screenshot_resolutions": [ | |
| { | |
| "width": 1920, | |
| "height": 1080, | |
| "name": "desktop" | |
| }, | |
| { | |
| "width": 375, | |
| "height": 812, | |
| "name": "mobile" | |
| } | |
| ], | |
| "screenshot_wait": { | |
| "wait_until": "networkidle", | |
| "extra_wait_ms": 3000, | |
| "scroll_for_lazy_load": true | |
| } | |
| } | |