mirror of
https://github.com/BerriAI/litellm.git
synced 2026-11-18 09:45:23 +00:00
Merge e0aeab4edf into 3b8ad37296
This commit is contained in:
commit
f3a4e965bc
2 changed files with 48 additions and 7 deletions
|
|
@ -61,7 +61,7 @@ class DashScopeRerankConfig(BaseRerankConfig):
|
|||
Reference: https://help.aliyun.com/zh/model-studio/text-rerank-api
|
||||
|
||||
Targets DashScope's qwen3-rerank model. Request fields: model, query,
|
||||
documents, top_n, return_documents. Response: results[].index,
|
||||
documents, top_n, return_documents, instruct. Response: results[].index,
|
||||
results[].relevance_score, optionally results[].document.text (when
|
||||
return_documents=true), plus a top-level usage.total_tokens counter.
|
||||
"""
|
||||
|
|
@ -116,7 +116,7 @@ class DashScopeRerankConfig(BaseRerankConfig):
|
|||
}
|
||||
|
||||
def get_supported_cohere_rerank_params(self, model: str) -> list:
|
||||
return ["query", "documents", "top_n", "return_documents"]
|
||||
return ["query", "documents", "top_n", "return_documents", "instruction"]
|
||||
|
||||
def map_cohere_rerank_params(
|
||||
self,
|
||||
|
|
@ -133,8 +133,6 @@ class DashScopeRerankConfig(BaseRerankConfig):
|
|||
max_tokens_per_doc: int | None = None,
|
||||
instruction: str | None = None,
|
||||
) -> dict:
|
||||
# qwen3-rerank accepts query/documents/top_n/return_documents. The
|
||||
# rest (rank_fields, max_*_per_doc) are silently dropped.
|
||||
params: Final[OptionalRerankParams] = OptionalRerankParams(
|
||||
query=query,
|
||||
documents=documents,
|
||||
|
|
@ -143,12 +141,14 @@ class DashScopeRerankConfig(BaseRerankConfig):
|
|||
params["top_n"] = top_n
|
||||
if return_documents is not None:
|
||||
params["return_documents"] = return_documents
|
||||
if instruction is not None:
|
||||
params["instruction"] = instruction
|
||||
return dict(params)
|
||||
|
||||
def transform_rerank_request(
|
||||
self,
|
||||
model: str,
|
||||
optional_rerank_params: dict,
|
||||
optional_rerank_params: dict[str, object],
|
||||
headers: dict,
|
||||
litellm_params: dict | None = None,
|
||||
) -> dict:
|
||||
|
|
@ -166,7 +166,10 @@ class DashScopeRerankConfig(BaseRerankConfig):
|
|||
request["top_n"] = optional_rerank_params["top_n"]
|
||||
if optional_rerank_params.get("return_documents") is not None:
|
||||
request["return_documents"] = optional_rerank_params["return_documents"]
|
||||
return request
|
||||
instruction: Final = optional_rerank_params.get("instruction")
|
||||
if instruction is None:
|
||||
return request
|
||||
return dict(request, instruct=_STR.validate_python(instruction))
|
||||
|
||||
def transform_rerank_response(
|
||||
self,
|
||||
|
|
|
|||
|
|
@ -3,19 +3,22 @@ Unit tests for DashScope rerank transformation.
|
|||
"""
|
||||
|
||||
import json
|
||||
from typing import Final
|
||||
from unittest.mock import MagicMock
|
||||
|
||||
import httpx
|
||||
import pytest
|
||||
from pydantic import ValidationError
|
||||
|
||||
|
||||
from litellm import LlmProviders
|
||||
from litellm.llms.dashscope.common_utils import DashScopeError
|
||||
from litellm.llms.dashscope.rerank.transformation import (
|
||||
DEFAULT_RERANK_URL,
|
||||
DashScopeRerankConfig,
|
||||
)
|
||||
from litellm.rerank_api.rerank_utils import get_optional_rerank_params
|
||||
from litellm.types.rerank import RerankResponse
|
||||
from litellm.utils import ProviderConfigManager
|
||||
|
||||
|
||||
class TestDashScopeRerankURL:
|
||||
|
|
@ -112,8 +115,43 @@ class TestDashScopeRerankRequest:
|
|||
"documents",
|
||||
"top_n",
|
||||
"return_documents",
|
||||
"instruction",
|
||||
]
|
||||
|
||||
@pytest.mark.parametrize(
|
||||
"provider",
|
||||
[LlmProviders.DASHSCOPE, LlmProviders.QWENCLOUD, LlmProviders.QWEN_AI_PLATFORM],
|
||||
)
|
||||
@pytest.mark.parametrize("instruction", [None, "", "Retrieve semantically similar text."])
|
||||
def test_instruction_reaches_provider_request(
|
||||
self, provider: LlmProviders, instruction: str | None
|
||||
) -> None:
|
||||
config: Final = ProviderConfigManager.get_provider_rerank_config(
|
||||
model="qwen3-rerank", provider=provider, api_base=None, present_version_params=[]
|
||||
)
|
||||
assert config is not None
|
||||
params: Final = get_optional_rerank_params(
|
||||
rerank_provider_config=config,
|
||||
model="qwen3-rerank",
|
||||
drop_params=False,
|
||||
query="How do I reset my password?",
|
||||
documents=["How can I change my password?", "Open settings and choose Reset password."],
|
||||
top_n=1,
|
||||
return_documents=False,
|
||||
instruction=instruction,
|
||||
)
|
||||
body: Final = config.transform_rerank_request(
|
||||
model="qwen3-rerank", optional_rerank_params=params, headers={}
|
||||
)
|
||||
assert body == {
|
||||
"model": "qwen3-rerank",
|
||||
"query": "How do I reset my password?",
|
||||
"documents": ["How can I change my password?", "Open settings and choose Reset password."],
|
||||
"top_n": 1,
|
||||
"return_documents": False,
|
||||
**({"instruct": instruction} if instruction is not None else {}),
|
||||
}
|
||||
|
||||
def test_map_params_drops_unsupported(self):
|
||||
# qwen3-rerank accepts query/documents/top_n/return_documents.
|
||||
# rank_fields and max_*_per_doc are silently dropped.
|
||||
|
|
|
|||
Loading…
Add table
Reference in a new issue