Prompt tune with minimal setting raise exception

[hyeoungho]

Hello,

I am testing auto prompt tuning as instructed in the document: https://microsoft.github.io/graphrag/posts/prompt_tuning/auto_prompt_tuning/

However, I got this weird exception and don't understand why this is an issue. Any ideas?

(.venv) hybae@---:/mnt/e/Work/GraphRagNews$ python -m graphrag.prompt_tune --root /mnt/e/Work/GraphRagNews/test --config settings.yaml --no-entity-types --output custom_prompt

Traceback (most recent call last):
  File "<frozen runpy>", line 198, in _run_module_as_main
  File "<frozen runpy>", line 88, in _run_code
  File "/mnt/e/Work/GraphRagNews/.venv/lib/python3.11/site-packages/graphrag/prompt_tune/__main__.py", line 128, in <module>
    loop.run_until_complete(
  File "/usr/lib/python3.11/asyncio/base_events.py", line 654, in run_until_complete
    return future.result()
           ^^^^^^^^^^^^^^^
  File "/mnt/e/Work/GraphRagNews/.venv/lib/python3.11/site-packages/graphrag/prompt_tune/cli.py", line 57, in prompt_tune
    graph_config = load_config(root_path, config)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/e/Work/GraphRagNews/.venv/lib/python3.11/site-packages/graphrag/config/load_config.py", line 52, in load_config
    resolve_path(
  File "/mnt/e/Work/GraphRagNews/.venv/lib/python3.11/site-packages/graphrag/config/resolve_path.py", line 152, in resolve_path
    return _resolve_timestamp_path(path_to_resolve, pattern_or_timestamp_value)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/e/Work/GraphRagNews/.venv/lib/python3.11/site-packages/graphrag/config/resolve_path.py", line 117, in _resolve_timestamp_path
    return _resolve_timestamp_path_with_dir(path, pattern_or_timestamp_value)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/e/Work/GraphRagNews/.venv/lib/python3.11/site-packages/graphrag/config/resolve_path.py", line 70, in _resolve_timestamp_path_with_dir
    raise ValueError(msg)
ValueError: Parent directory /mnt/e/Work/GraphRagNews/test/output does not exist or is not a directory.

When I initialized the root directory, I have only 'prompt' folder generated. Don't understand why 'output' folder is required. Then I manually made 'output' folder, which generates a new exception:

Traceback (most recent call last):
  File "<frozen runpy>", line 198, in _run_module_as_main
  File "<frozen runpy>", line 88, in _run_code
  File "/mnt/e/Work/GraphRagNews/.venv/lib/python3.11/site-packages/graphrag/prompt_tune/__main__.py", line 128, in <module>
    loop.run_until_complete(
  File "/usr/lib/python3.11/asyncio/base_events.py", line 654, in run_until_complete
    return future.result()
           ^^^^^^^^^^^^^^^
  File "/mnt/e/Work/GraphRagNews/.venv/lib/python3.11/site-packages/graphrag/prompt_tune/cli.py", line 57, in prompt_tune
    graph_config = load_config(root_path, config)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/e/Work/GraphRagNews/.venv/lib/python3.11/site-packages/graphrag/config/load_config.py", line 52, in load_config
    resolve_path(
  File "/mnt/e/Work/GraphRagNews/.venv/lib/python3.11/site-packages/graphrag/config/resolve_path.py", line 152, in resolve_path
    return _resolve_timestamp_path(path_to_resolve, pattern_or_timestamp_value)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/e/Work/GraphRagNews/.venv/lib/python3.11/site-packages/graphrag/config/resolve_path.py", line 117, in _resolve_timestamp_path
    return _resolve_timestamp_path_with_dir(path, pattern_or_timestamp_value)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/mnt/e/Work/GraphRagNews/.venv/lib/python3.11/site-packages/graphrag/config/resolve_path.py", line 78, in _resolve_timestamp_path_with_dir
    raise ValueError(msg)
ValueError: No timestamp directories found in /mnt/e/Work/GraphRagNews/test/output that match ^\d{8}-\d{6}$

From this exception message, the code expected a folder name with timestamp expression like the one generated during indexing process. But again, I don't understand why that is required. It doesn't make sense to me.

[hyeoungho]

Found a solution and raised #1149

[KumarAditya98]

Why doesn't the above URL exist?

[Tom-Boscher]

Curious to see which solution you found

[hyeoungho]

Updated the link to the PR. I changed the prompt_tune/cli.py to provide run_id in load_config so that it can properly initialize folders.

[hyeoungho]

Looks like new load_config does not require run_id. I just closed the PR.