pengjunjie/ragflow
1
0
Fork 1
mirror of https://github.com/infiniflow/ragflow.git synced 2025-12-08 20:42:30 +08:00
Code Issues Packages Projects Releases Wiki Activity

refactor some llm api using openai api format (#1692)

Browse Source 
### What problem does this PR solve?

refactor some llm api using openai api format

### Type of change

- [x] Refactoring

---------

Co-authored-by: Zhedong Cen <cenzhedong2@126.com>
This commit is contained in:
黄腾
2024-07-25 10:23:35 +08:00
committed by GitHub
parent d5f87a5498
commit e67bfca552
3 changed files with 58 additions and 240 deletions
Download Patch File Download Diff File Expand all files Collapse all files

38
rag/llm/embedding_model.py
View File

@ -113,21 +113,24 @@ class OpenAIEmbed(Base):
class LocalAIEmbed(Base):
def __init__(self, key, model_name, base_url):
self.base_url = base_url + "/embeddings"
self.headers = {
"Content-Type": "application/json",
}
if not base_url:
raise ValueError("Local embedding model url cannot be None")
if base_url.split("/")[-1] != "v1":
base_url = os.path.join(base_url, "v1")
self.client = OpenAI(api_key="empty", base_url=base_url)
self.model_name = model_name.split("___")[0]
def encode(self, texts: list, batch_size=None):
data = {"model": self.model_name, "input": texts, "encoding_type": "float"}
res = requests.post(self.base_url, headers=self.headers, json=data).json()
return np.array([d["embedding"] for d in res["data"]]), 1024
def encode(self, texts: list, batch_size=32):
res = self.client.embeddings.create(input=texts, model=self.model_name)
return (
np.array([d.embedding for d in res.data]),
1024,
) # local embedding for LmStudio donot count tokens
def encode_queries(self, text):
embds, cnt = self.encode([text])
return np.array(embds[0]), cnt
res = self.client.embeddings.create(text, model=self.model_name)
return np.array(res.data[0].embedding), 1024
class AzureEmbed(OpenAIEmbed):
def __init__(self, key, model_name, **kwargs):
@ -502,7 +505,7 @@ class NvidiaEmbed(Base):
return np.array(embds[0]), cnt
class LmStudioEmbed(Base):
class LmStudioEmbed(LocalAIEmbed):
def __init__(self, key, model_name, base_url):
if not base_url:
raise ValueError("Local llm url cannot be None")
@ -510,14 +513,3 @@ class LmStudioEmbed(Base):
self.base_url = os.path.join(base_url, "v1")
self.client = OpenAI(api_key="lm-studio", base_url=self.base_url)
self.model_name = model_name
def encode(self, texts: list, batch_size=32):
res = self.client.embeddings.create(input=texts, model=self.model_name)
return (
np.array([d.embedding for d in res.data]),
1024,
) # local embedding for LmStudio donot count tokens
def encode_queries(self, text):
res = self.client.embeddings.create(text, model=self.model_name)
return np.array(res.data[0].embedding), 1024