YG-Soft/YG_FT
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat(data-process): 支持思维链输出类型

Browse Source
This commit is contained in:
caoxiaozhu
2026-07-27 13:08:44 +08:00
parent ecafb7eb13
commit de2e8952b5
14 changed files with 443 additions and 46 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
backend/tests/test_data_process_api.py
View File

@@ -1405,6 +1405,40 @@ def test_stale_generation_worker_cannot_overwrite_new_run(monkeypatch: Any) -> N
assert store.tasks[task_id]["status"] == "running"
def test_reasoning_output_requires_generation_model() -> None:
store = FakeDataProcessStore()
task = store.create_task(
{
"name": "思维链模型校验",
"process_type": "structured",
"config": {"output_type": "reasoning"},
}
)
task_id = task["id"]
store.replace_preview_items(
task_id,
[
{
"source_file_id": None,
"original_content": "需要推理的来源内容",
"edited_content": "需要推理的来源内容",
"status": "manual",
}
],
)
started = store.start_generation(task_id, replace_existing=True)
data_process_endpoint._run_generation(
store,
task_id,
started["generation_run_id"],
)
failed = store.get_task(task_id)
assert failed["status"] == "failed"
assert failed["failure_reason"] == "思维链输出必须配置可用的数据生成模型"
def test_result_status_cannot_be_forged_by_client(tmp_path: Path) -> None:
client, _, _ = make_client(tmp_path)
task_id = client.post(

125
backend/tests/test_data_process_generation.py
View File

@@ -86,6 +86,119 @@ def test_generate_model_records_uses_prompt_auth_and_stable_split() -> None:
assert progress_updates == [(1, 1)]
def test_generate_model_records_builds_reasoning_output_with_think_tags() -> None:
def handler(request: httpx.Request) -> httpx.Response:
payload = json.loads(request.content)
system_prompt = payload["messages"][0]["content"]
assert '"reasoning":"...","answer":"..."' in system_prompt
assert "系统会在保存时统一组装" in system_prompt
content = "<think>模型接口自己的分析</think>" + json.dumps(
{
"items": [
{
"instruction": "计算两项费用合计",
"input": "交通费 30 元，餐费 20 元",
"reasoning": "先识别两项费用，再计算 30 + 20。",
"answer": "合计 50 元。",
}
]
},
ensure_ascii=False,
)
return httpx.Response(
200,
json={"choices": [{"message": {"content": content}}]},
)
records = generate_model_records(
[{"id": "preview-reasoning", "edited_content": "交通费 30 元，餐费 20 元"}],
model={"name": "model", "api_url": "https://model.example/v1"},
config={"output_type": "reasoning"},
task_id="task-reasoning",
split={"train": 100, "validation": 0, "test": 0},
qa_pairs_per_item=1,
client=httpx.Client(transport=httpx.MockTransport(handler)),
)
assert records[0]["status"] == "valid"
assert records[0]["output"] == (
"<think>\n先识别两项费用,再计算 30 + 20。\n</think>\n合计 50 元。"
)
def test_generate_model_records_marks_reasoning_without_reasoning_field_invalid() -> None:
response = {
"choices": [
{
"message": {
"content": json.dumps(
{
"items": [
{
"instruction": "问题",
"answer": "只有最终答案",
}
]
},
ensure_ascii=False,
)
}
}
]
}
client = httpx.Client(
transport=httpx.MockTransport(lambda _: httpx.Response(200, json=response))
)
records = generate_model_records(
[{"id": "preview-missing-reasoning", "edited_content": "来源正文"}],
model={"name": "model", "api_url": "https://model.example/v1"},
config={"output_type": "reasoning"},
task_id="task-missing-reasoning",
split={"train": 100, "validation": 0, "test": 0},
qa_pairs_per_item=1,
client=client,
)
assert records[0]["status"] == "invalid"
assert records[0]["output"] == "只有最终答案"
assert "reasoning" in records[0]["error"]
def test_standard_output_removes_model_think_block() -> None:
content = json.dumps(
{
"items": [
{
"instruction": "问题",
"output": "<think>不应保存的分析</think>最终答案",
}
]
},
ensure_ascii=False,
)
client = httpx.Client(
transport=httpx.MockTransport(
lambda _: httpx.Response(
200,
json={"choices": [{"message": {"content": content}}]},
)
)
)
records = generate_model_records(
[{"id": "preview-standard", "edited_content": "来源正文"}],
model={"name": "model", "api_url": "https://model.example/v1"},
config={"output_type": "standard"},
task_id="task-standard",
split={"train": 100, "validation": 0, "test": 0},
qa_pairs_per_item=1,
client=client,
)
assert records[0]["output"] == "最终答案"
def test_generate_model_records_keeps_partial_failure_for_manual_repair() -> None:
client = httpx.Client(
transport=httpx.MockTransport(
@@ -267,3 +380,15 @@ def test_generate_model_records_rejects_out_of_range_count(
split={"train": 100, "validation": 0, "test": 0},
qa_pairs_per_item=qa_pairs_per_item,
)
def test_generate_model_records_rejects_unknown_output_type() -> None:
with pytest.raises(ModelGenerationError, match="output_type"):
generate_model_records(
[],
model={"name": "model", "api_url": "https://model.example/v1"},
config={"output_type": "unknown"},
task_id="task-invalid-output",
split={"train": 100, "validation": 0, "test": 0},
qa_pairs_per_item=1,
)

71
backend/tests/test_data_process_store.py
View File

@@ -15,6 +15,7 @@ from app.modules.data_process.store import (
InvalidStateError,
_decode_row,
_preview_config_changed,
_reasoning_output_is_valid,
_source_storage_descriptor,
)
@@ -114,14 +115,25 @@ class _PublishConnection:
)
if normalized.startswith("INSERT INTO dataset_records"):
self.records.append(
{"dataset_id": params[1], "line_no": params[4], "split": params[5]}
{
"dataset_id": params[1],
"line_no": params[4],
"split": params[5],
"output": params[8],
"raw": json.loads(params[9]),
}
)
return _Result()
class _PublishStore(DataProcessStore):
def __init__(self, conn: _PublishConnection):
def __init__(
self,
conn: _PublishConnection,
task_config: dict[str, Any] | None = None,
):
self._conn = conn
self._task_config = task_config or {}
@contextmanager
def connect(self) -> Iterator[_PublishConnection]:
@@ -135,7 +147,7 @@ class _PublishStore(DataProcessStore):
"id": task_id,
"status": "completed",
"description": "",
"config": {},
"config": self._task_config,
"output_dataset_id": train_dataset and train_dataset["id"],
}
@@ -967,7 +979,11 @@ def test_publish_creates_three_independent_datasets_with_exact_counts() -> None:
"status": "valid",
"instruction": f"问题 {index}",
"input": "",
"output": f"答案 {index}",
"output": (
"<think>\n先读取制度条款。\n</think>\n答案 0"
if index == 0
else f"答案 {index}"
),
"preview_item_id": f"preview-{index}",
}
for index in range(28)
@@ -993,6 +1009,10 @@ def test_publish_creates_three_independent_datasets_with_exact_counts() -> None:
item["id"] for item in conn.datasets
}
assert len(conn.records) == 28
reasoning_record = next(
item for item in conn.records if item["output"].startswith("<think>")
)
assert reasoning_record["raw"]["output"] == reasoning_record["output"]
assert published["dataset"]["type"] == "train"
assert len(published["datasets"]) == 3
assert published["split_counts"] == {"train": 22, "validation": 3, "test": 3}
@@ -1078,6 +1098,49 @@ def test_publish_keeps_all_three_datasets_when_a_small_split_is_empty() -> None:
assert len(conn.files) == 3
@pytest.mark.parametrize(
("output", "expected"),
[
("<think>\n推理步骤\n</think>\n最终答案", True),
("<think></think>\n最终答案", False),
("<think>只有推理</think>", False),
("没有标签的最终答案", False),
("<think>外层<think>嵌套</think></think>答案", False),
],
)
def test_reasoning_output_validator_requires_one_complete_pair(
output: str,
expected: bool,
) -> None:
assert _reasoning_output_is_valid(output) is expected
def test_publish_rejects_invalid_reasoning_output_format() -> None:
conn = _PublishConnection(
[
{
"id": "result-reasoning-invalid",
"status": "valid",
"instruction": "需要推理的问题",
"input": "",
"output": "只有最终答案",
"preview_item_id": "preview-reasoning-invalid",
}
]
)
with pytest.raises(InvalidStateError, match="1 invalid results"):
_PublishStore(conn, {"output_type": "reasoning"}).publish(
"task-reasoning-invalid",
{
"dataset_name": "无效思维链",
"storage_type": "local",
"format": "alpaca_jsonl",
"split": {"train": 80, "validation": 10, "test": 10},
},
)
def test_source_storage_descriptor_accepts_owned_local_and_legacy_db_references() -> None:
task_id = "dpt_task"
source_file_id = "dpsf_source"