mirror of
https://github.com/infiniflow/ragflow.git
synced 2025-12-08 20:42:30 +08:00
build python version rag-flow (#21)
* clean rust version project * clean rust version project * build python version rag-flow
This commit is contained in:
KevinHuSh
32
rag/llm/__init__.py
Normal file
32
rag/llm/__init__.py
Normal file
@ -0,0 +1,32 @@
|
||||
#
|
||||
# Copyright 2019 The FATE Authors. All Rights Reserved.
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
#
|
||||
from .embedding_model import *
|
||||
from .chat_model import *
|
||||
from .cv_model import *
|
||||
|
||||
|
||||
EmbeddingModel = {
|
||||
"local": HuEmbedding,
|
||||
"OpenAI": OpenAIEmbed,
|
||||
"通义千问": QWenEmbed,
|
||||
}
|
||||
|
||||
|
||||
CvModel = {
|
||||
"OpenAI": GptV4,
|
||||
"通义千问": QWenCV,
|
||||
}
|
||||
|
||||
52
rag/llm/chat_model.py
Normal file
52
rag/llm/chat_model.py
Normal file
@ -0,0 +1,52 @@
|
||||
#
|
||||
# Copyright 2019 The FATE Authors. All Rights Reserved.
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
#
|
||||
from abc import ABC
|
||||
from openai import OpenAI
|
||||
import os
|
||||
|
||||
|
||||
class Base(ABC):
|
||||
def chat(self, system, history, gen_conf):
|
||||
raise NotImplementedError("Please implement encode method!")
|
||||
|
||||
|
||||
class GptTurbo(Base):
|
||||
def __init__(self):
|
||||
self.client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])
|
||||
|
||||
def chat(self, system, history, gen_conf):
|
||||
history.insert(0, {"role": "system", "content": system})
|
||||
res = self.client.chat.completions.create(
|
||||
model="gpt-3.5-turbo",
|
||||
messages=history,
|
||||
**gen_conf)
|
||||
return res.choices[0].message.content.strip()
|
||||
|
||||
|
||||
class QWenChat(Base):
|
||||
def chat(self, system, history, gen_conf):
|
||||
from http import HTTPStatus
|
||||
from dashscope import Generation
|
||||
# export DASHSCOPE_API_KEY=YOUR_DASHSCOPE_API_KEY
|
||||
history.insert(0, {"role": "system", "content": system})
|
||||
response = Generation.call(
|
||||
Generation.Models.qwen_turbo,
|
||||
messages=history,
|
||||
result_format='message'
|
||||
)
|
||||
if response.status_code == HTTPStatus.OK:
|
||||
return response.output.choices[0]['message']['content']
|
||||
return response.message
|
||||
89
rag/llm/cv_model.py
Normal file
89
rag/llm/cv_model.py
Normal file
@ -0,0 +1,89 @@
|
||||
#
|
||||
# Copyright 2019 The FATE Authors. All Rights Reserved.
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
#
|
||||
from abc import ABC
|
||||
from openai import OpenAI
|
||||
import os
|
||||
import base64
|
||||
from io import BytesIO
|
||||
|
||||
|
||||
class Base(ABC):
|
||||
def __init__(self, key, model_name):
|
||||
pass
|
||||
|
||||
def describe(self, image, max_tokens=300):
|
||||
raise NotImplementedError("Please implement encode method!")
|
||||
|
||||
def image2base64(self, image):
|
||||
if isinstance(image, BytesIO):
|
||||
return base64.b64encode(image.getvalue()).decode("utf-8")
|
||||
buffered = BytesIO()
|
||||
try:
|
||||
image.save(buffered, format="JPEG")
|
||||
except Exception as e:
|
||||
image.save(buffered, format="PNG")
|
||||
return base64.b64encode(buffered.getvalue()).decode("utf-8")
|
||||
|
||||
def prompt(self, b64):
|
||||
return [
|
||||
{
|
||||
"role": "user",
|
||||
"content": [
|
||||
{
|
||||
"type": "text",
|
||||
"text": "请用中文详细描述一下图中的内容，比如时间，地点，人物，事情，人物心情等。",
|
||||
},
|
||||
{
|
||||
"type": "image_url",
|
||||
"image_url": {
|
||||
"url": f"data:image/jpeg;base64,{b64}"
|
||||
},
|
||||
},
|
||||
],
|
||||
}
|
||||
]
|
||||
|
||||
|
||||
class GptV4(Base):
|
||||
def __init__(self, key, model_name="gpt-4-vision-preview"):
|
||||
self.client = OpenAI(key)
|
||||
self.model_name = model_name
|
||||
|
||||
def describe(self, image, max_tokens=300):
|
||||
b64 = self.image2base64(image)
|
||||
|
||||
res = self.client.chat.completions.create(
|
||||
model=self.model_name,
|
||||
messages=self.prompt(b64),
|
||||
max_tokens=max_tokens,
|
||||
)
|
||||
return res.choices[0].message.content.strip()
|
||||
|
||||
|
||||
class QWenCV(Base):
|
||||
def __init__(self, key, model_name="qwen-vl-chat-v1"):
|
||||
import dashscope
|
||||
dashscope.api_key = key
|
||||
self.model_name = model_name
|
||||
|
||||
def describe(self, image, max_tokens=300):
|
||||
from http import HTTPStatus
|
||||
from dashscope import MultiModalConversation
|
||||
response = MultiModalConversation.call(model=self.model_name,
|
||||
messages=self.prompt(self.image2base64(image)))
|
||||
if response.status_code == HTTPStatus.OK:
|
||||
return response.output.choices[0]['message']['content']
|
||||
return response.message
|
||||
94
rag/llm/embedding_model.py
Normal file
94
rag/llm/embedding_model.py
Normal file
@ -0,0 +1,94 @@
|
||||
#
|
||||
# Copyright 2019 The FATE Authors. All Rights Reserved.
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
#
|
||||
from abc import ABC
|
||||
|
||||
import dashscope
|
||||
from openai import OpenAI
|
||||
from FlagEmbedding import FlagModel
|
||||
import torch
|
||||
import os
|
||||
import numpy as np
|
||||
|
||||
from rag.utils import num_tokens_from_string
|
||||
|
||||
|
||||
class Base(ABC):
|
||||
def __init__(self, key, model_name):
|
||||
pass
|
||||
|
||||
|
||||
def encode(self, texts: list, batch_size=32):
|
||||
raise NotImplementedError("Please implement encode method!")
|
||||
|
||||
|
||||
class HuEmbedding(Base):
|
||||
def __init__(self):
|
||||
"""
|
||||
If you have trouble downloading HuggingFace models, -_^ this might help!!
|
||||
|
||||
For Linux:
|
||||
export HF_ENDPOINT=https://hf-mirror.com
|
||||
|
||||
For Windows:
|
||||
Good luck
|
||||
^_-
|
||||
|
||||
"""
|
||||
self.model = FlagModel("BAAI/bge-large-zh-v1.5",
|
||||
query_instruction_for_retrieval="为这个句子生成表示以用于检索相关文章：",
|
||||
use_fp16=torch.cuda.is_available())
|
||||
|
||||
|
||||
def encode(self, texts: list, batch_size=32):
|
||||
token_count = 0
|
||||
for t in texts: token_count += num_tokens_from_string(t)
|
||||
res = []
|
||||
for i in range(0, len(texts), batch_size):
|
||||
res.extend(self.model.encode(texts[i:i + batch_size]).tolist())
|
||||
return np.array(res), token_count
|
||||
|
||||
|
||||
class OpenAIEmbed(Base):
|
||||
def __init__(self, key, model_name="text-embedding-ada-002"):
|
||||
self.client = OpenAI(key)
|
||||
self.model_name = model_name
|
||||
|
||||
def encode(self, texts: list, batch_size=32):
|
||||
token_count = 0
|
||||
for t in texts: token_count += num_tokens_from_string(t)
|
||||
res = self.client.embeddings.create(input=texts,
|
||||
model=self.model_name)
|
||||
return [d["embedding"] for d in res["data"]], token_count
|
||||
|
||||
|
||||
class QWenEmbed(Base):
|
||||
def __init__(self, key, model_name="text_embedding_v2"):
|
||||
dashscope.api_key = key
|
||||
self.model_name = model_name
|
||||
|
||||
def encode(self, texts: list, batch_size=32, text_type="document"):
|
||||
import dashscope
|
||||
res = []
|
||||
token_count = 0
|
||||
for txt in texts:
|
||||
resp = dashscope.TextEmbedding.call(
|
||||
model=self.model_name,
|
||||
input=txt[:2048],
|
||||
text_type=text_type
|
||||
)
|
||||
res.append(resp["output"]["embeddings"][0]["embedding"])
|
||||
token_count += resp["usage"]["total_tokens"]
|
||||
return res, token_count
|
||||
Reference in New Issue
Block a user