Skip to content

fix(wren-ai-service): column pruning step in retrieval pipeline and correct code for sql generation in evaluation #1225

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
paopa merged 3 commits into from
Jan 24, 2025
Merged

fix(wren-ai-service): column pruning step in retrieval pipeline and correct code for sql generation in evaluation #1225

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
2df42ef
fix: correct the retrieval behavior for column pruning
paopa Jan 24, 2025
118daa0
fix: predict process for ask missing sql reasoning parameter
paopa Jan 24, 2025
b11d578
chore: column pruning disable by default for openai llm from launcher
paopa Jan 24, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 7 additions & 2 deletions wren-ai-service/eval/pipelines.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -177,11 +177,14 @@ def __init__(
):
super().__init__(meta)

_indexing = indexing.DBSchema(
_db_schema_indexing = indexing.DBSchema(
**pipe_components["db_schema_indexing"],
column_batch_size=settings.column_indexing_batch_size,
)
deploy_model(mdl, _indexing)
_table_description_indexing = indexing.TableDescription(
**pipe_components["table_description_indexing"],
)
deploy_model(mdl, [_db_schema_indexing, _table_description_indexing])

self._retrieval = retrieval.Retrieval(
**pipe_components["db_schema_retrieval"],
Expand Down Expand Up @@ -247,6 +250,7 @@ async def _process(self, prediction: dict, document: list, **_) -> dict:
samples=prediction["samples"],
has_calculated_field=prediction.get("has_calculated_field", False),
has_metric=prediction.get("has_metric", False),
sql_generation_reasoning=prediction.get("reasoning", ""),
)

prediction["actual_output"] = actual_output
Expand Down Expand Up @@ -337,6 +341,7 @@ async def _process(self, prediction: dict, **_) -> dict:
sql_samples=[],
has_calculated_field=has_calculated_field,
has_metric=has_metric,
sql_generation_reasoning=prediction.get("reasoning", ""),
)

prediction["actual_output"] = actual_output
Expand Down
8 changes: 4 additions & 4 deletions wren-ai-service/src/pipelines/retrieval/retrieval.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -268,9 +268,8 @@ def prompt(
"db_schemas token count is greater than 100,000, so we will prune columns"
)
db_schemas = [
ddl
build_table_ddl(construct_db_schema)[0]
for construct_db_schema in construct_db_schemas
for ddl, _ in build_table_ddl(construct_db_schema)
]

if history:
Expand Down Expand Up @@ -417,8 +416,9 @@ def __init__(

# for the first time, we need to load the encodings
_model = llm_provider.get_model()
if "gpt-4o" in _model or "gpt-4o-mini" in _model:
allow_using_db_schemas_without_pruning = True
if allow_using_db_schemas_without_pruning and (
"gpt-4o" in _model or "gpt-4o-mini" in _model
):
_encoding = tiktoken.get_encoding("o200k_base")
else:
_encoding = tiktoken.get_encoding("cl100k_base")
Expand Down
Loading