|
Download code/README.md from badr7/rapidchat-data: direct link, hf CLI and curl.
- Browser
- Download file 834 Bytes
-
https://huggingface.co/datasets/badr7/rapidchat-data/resolve/main/code/README.md
- Command line
-
hf download hf://datasets/badr7/rapidchat-data/code/README.md
-
curl -L -o README.md https://huggingface.co/datasets/badr7/rapidchat-data/resolve/main/code/README.md
834 Bytes
| # rapidchat code | |
| | file | what it does | | |
| |---|---| | |
| | `vc_bench.py` | builds the re-skinned airline world and the round-1 scenario families | | |
| | `vc_bench_r2.py`, `vc_bench_r3.py`, `vc_bench_r3b.py` | round-2 and round-3 scenario families | | |
| | `validate_vc_bench.py`, `validate_r2.py`, `validate_r3.py` | replay every scenario's gold actions with the real tau2 airline tools and re-derive each answer | | |
| | `oracle_vc_bench.py`, `mutation_test.py` | sanity checks: an oracle agent must pass, mutated answers must fail | | |
| | `vc_install.py` | registers the practice world as a tau2 domain (`vc_airline`) | | |
| | `vc_sft.py` | turns passing tau2 conversations into chat-format training examples | | |
| | `train_lora.py` | LoRA fine-tune (rank 32) with a chunked, checkpointed loss; resumable | | |
| | `merge.py` | merges an adapter into the base weights | | |