| #!/usr/bin/env bash |
| |
| |
| |
| |
| |
| |
| set -euo pipefail |
|
|
| ROOT="$(cd "$(dirname "$0")/.." && pwd)" |
| DATA="$ROOT/data" |
| GH_RAW="https://raw.githubusercontent.com" |
|
|
| fetch() { |
| mkdir -p "$(dirname "$2")" |
| if [ -s "$2" ]; then echo " [skip] $2 (exists)"; return; fi |
| echo " [get ] $2" |
| curl -fsSL --retry 3 "$1" -o "$2" |
| } |
|
|
| echo "== BFCL v4 (Berkeley Function-Call Leaderboard, Apache-2.0) ==" |
| BFCL="$GH_RAW/ShishirPatil/gorilla/main/berkeley-function-call-leaderboard/bfcl_eval/data" |
| for f in BFCL_v4_simple_python.json BFCL_v4_multiple.json BFCL_v4_parallel.json; do |
| fetch "$BFCL/$f" "$DATA/bfcl/$f" |
| done |
|
|
| echo "== Seal-Tools (Apache-2.0) ==" |
| SEAL="$GH_RAW/fairyshine/Seal-Tools/master/Seal-Tools_Dataset" |
| fetch "$SEAL/tool.jsonl" "$DATA/sealtools/tool.jsonl" |
| fetch "$SEAL/test_in_domain.jsonl" "$DATA/sealtools/test_in_domain.jsonl" |
|
|
| echo "== ToolAlpaca (Apache-2.0) ==" |
| TA="$GH_RAW/tangqiaoyu/ToolAlpaca/main/data" |
| fetch "$TA/eval_simulated.json" "$DATA/toolalpaca/eval_simulated.json" |
| fetch "$TA/eval_real.json" "$DATA/toolalpaca/eval_real.json" |
|
|
| echo "== API-Bank (MIT; HF dataset liminghao1630/API-Bank) ==" |
| AB="https://huggingface.co/datasets/liminghao1630/API-Bank/resolve/main/test-data" |
| fetch "$AB/level-1-api.json" "$DATA/apibank/level-1-api.json" |
| fetch "$AB/level-2-api.json" "$DATA/apibank/level-2-api.json" |
|
|
| echo |
| echo "Done. Two datasets need a manual step (only for the corresponding experiments):" |
| echo " * ToolBench -> data/toolbench/test_instruction/G{1,2,3}_*.json" |
| echo " from the official OpenBMB/ToolBench Google Drive release." |
| echo " * tau2-bench -> data/tau2/tools_{airline,retail,telecom}.json" |
| echo " generated from github.com/sierra-research/tau2-bench;" |
| echo " see data/README.md." |
|
|