遇见数据集

temp0726

收藏
魔搭社区2026-08-02 更新2026-08-02 收录
官方服务:

资源简介:

# LLM Grounding Eval Package Run commands from this directory so the relative dataset paths in `configs/llm_eval_models.json` resolve correctly. Smoke test without API calls: ```powershell python3.10 scripts\run_llm_grounding_eval.py --config configs\llm_eval_models.json --dry-run --dataset all --model gpt_5_5 ``` Example API run: ```powershell $env:OPENAI_API_KEY="sk-449ebe707c9e4866bc45c7e1c072025a" python3.10 scripts\run_llm_grounding_eval.py --config configs\llm_eval_models.json --model gpt_5_5 --dataset all --limit 5 --resume ``` macOS / zsh: ```zsh export OPENAI_API_KEY="<YOUR_API_KEY>" python3.10 scripts/run_llm_grounding_eval.py --config configs/llm_eval_models.json --model gpt_5_5 --dataset all --limit 5 --resume ``` If you are running from the repository root instead of this package directory, use: ```zsh export OPENAI_API_KEY="<YOUR_API_KEY>" cd llm_grounding_eval python3.10 scripts/run_llm_grounding_eval.py --config configs/llm_eval_models.json --model gpt_5_5 --dataset all --limit 5 --resume ``` export MATRIX_API_KEY="sk-449ebe707c9e4866bc45c7e1c072025a" rm -rf outputs/llm_grounding_eval/gpt_5_5 python3.10 scripts/run_llm_grounding_eval.py \ --config configs/llm_eval_models.json \ --model gpt_5_5 \ --dataset all \ --limit 1 \ --debug export MATRIX_API_KEY="sk-449ebe707c9e4866bc45c7e1c072025a" python3.10 scripts/run_llm_grounding_eval.py \ --config configs/llm_eval_models.json \ --model qwen3_vl_plus \ --dataset all \ --limit 1 \ --debug export MATRIX_API_KEY="sk-449ebe707c9e4866bc45c7e1c072025a" ./batch_llm_eval.sh full gpt_5_4 python3.10 scripts/run_llm_grounding_eval.py \ --config configs/llm_eval_models.json \ --model claude_opus_4_8 \ --dataset all \ --limit 1 \ --debug export MATRIX_API_KEY=‘sk-00d60555930e471f931fa15180060a32’ ./batch_llm_eval.sh full gpt_5_4 Outputs are written to `outputs/llm_grounding_eval/<model_name>/` inside this package directory. export MATRIX_API_KEY="sk-449ebe707c9e4866bc45c7e1c072025a" nohup bash scripts/run_1000_vlm.sh \ > logs/run_1000_claude_opus_4_7.nohup.log 2>&1 & 9X4LUOxs4A8iRi2xJnKQGFJEckYGbYUR cd llm_grounding_eval export MATRIX_API_KEY="9X4LUOxs4A8iRi2xJnKQGFJEckYGbYUR" bash scripts/full_llm_evalopen_1000_macos.sh

提供机构:
maas
创建时间:
2026-07-26
二维码
社区交流群
二维码
科研交流群
商业服务