github/DB-GPT
1
0
Fork 0
mirror of https://github.com/csunny/DB-GPT.git synced 2025-09-04 10:34:30 +00:00
Code Issues Packages Projects Releases Wiki Activity

chore(model): Modify siliconflow name (#2194)

Browse Source
This commit is contained in:
Fangyin Cheng
2024-12-12 16:47:14 +08:00
committed by GitHub
parent 4da1809b31
commit dab55493f5
10 changed files with 48 additions and 28 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
dbgpt/model/adapter/embeddings_loader.py
View File

@@ -109,7 +109,7 @@ class EmbeddingLoader:
if proxy_param.proxy_backend:
openapi_param["model_name"] = proxy_param.proxy_backend
return OpenAPIRerankEmbeddings(**openapi_param)
elif model_name in ["rerank_proxy_silicon_flow"]:
elif model_name in ["rerank_proxy_siliconflow"]:
from dbgpt.rag.embedding.rerank import SiliconFlowRerankEmbeddings
proxy_param = cast(ProxyEmbeddingParameters, param)

6
dbgpt/model/adapter/proxy_adapter.py
View File

@@ -349,7 +349,7 @@ class SiliconFlowProxyLLMModelAdapter(ProxyLLMModelAdapter):
return True
def do_match(self, lower_model_name_or_path: Optional[str] = None):
return lower_model_name_or_path == "silicon_flow_proxyllm"
return lower_model_name_or_path == "siliconflow_proxyllm"
def get_llm_client_class(
self, params: ProxyModelParameters
@@ -359,9 +359,9 @@ class SiliconFlowProxyLLMModelAdapter(ProxyLLMModelAdapter):
return SiliconFlowLLMClient
def get_async_generate_stream_function(self, model, model_path: str):
from dbgpt.model.proxy.llms.siliconflow import silicon_flow_generate_stream
from dbgpt.model.proxy.llms.siliconflow import siliconflow_generate_stream
return silicon_flow_generate_stream
return siliconflow_generate_stream
register_model_adapter(OpenAIProxyLLMModelAdapter)

2
dbgpt/model/parameter.py
View File

@@ -621,7 +621,7 @@ _EMBEDDING_PARAMETER_CLASS_TO_NAME_CONFIG = {
"proxy_tongyi",
"proxy_qianfan",
"rerank_proxy_http_openapi",
"rerank_proxy_silicon_flow",
"rerank_proxy_siliconflow",
]
}

16
dbgpt/model/proxy/llms/siliconflow.py
View File

@@ -12,10 +12,10 @@ if TYPE_CHECKING:
ClientType = Union[AsyncAzureOpenAI, AsyncOpenAI]
_SILICON_FLOW_DEFAULT_MODEL = "Qwen/Qwen2.5-Coder-32B-Instruct"
_SILICONFLOW_DEFAULT_MODEL = "Qwen/Qwen2.5-Coder-32B-Instruct"
async def silicon_flow_generate_stream(
async def siliconflow_generate_stream(
model: ProxyModel, tokenizer, params, device, context_len=2048
):
client: SiliconFlowLLMClient = model.proxy_llm_client
@@ -39,7 +39,7 @@ class SiliconFlowLLMClient(OpenAILLMClient):
model: Optional[str] = None,
proxies: Optional["ProxiesTypes"] = None,
timeout: Optional[int] = 240,
model_alias: Optional[str] = "silicon_flow_proxyllm",
model_alias: Optional[str] = "siliconflow_proxyllm",
context_length: Optional[int] = None,
openai_client: Optional["ClientType"] = None,
openai_kwargs: Optional[Dict[str, Any]] = None,
@@ -47,11 +47,11 @@ class SiliconFlowLLMClient(OpenAILLMClient):
):
api_base = (
api_base
or os.getenv("SILICON_FLOW_API_BASE")
or os.getenv("SILICONFLOW_API_BASE")
or "https://api.siliconflow.cn/v1"
)
api_key = api_key or os.getenv("SILICON_FLOW_API_KEY")
model = model or _SILICON_FLOW_DEFAULT_MODEL
api_key = api_key or os.getenv("SILICONFLOW_API_KEY")
model = model or _SILICONFLOW_DEFAULT_MODEL
if not context_length:
if "200k" in model:
context_length = 200 * 1024
@@ -60,7 +60,7 @@ class SiliconFlowLLMClient(OpenAILLMClient):
if not api_key:
raise ValueError(
"SiliconFlow API key is required, please set 'SILICON_FLOW_API_KEY' in environment "
"SiliconFlow API key is required, please set 'SILICONFLOW_API_KEY' in environment "
"or pass it as an argument."
)
@@ -83,5 +83,5 @@ class SiliconFlowLLMClient(OpenAILLMClient):
def default_model(self) -> str:
model = self._model
if not model:
model = _SILICON_FLOW_DEFAULT_MODEL
model = _SILICONFLOW_DEFAULT_MODEL
return model