Merge pull request #11 from csunny/dev

架构方案添加

README.md

@@ -6,6 +6,9 @@ A Open Database-GPT Experiment
 [DB-GPT](https://github.com/csunny/DB-GPT) 是一个实验性的开源应用程序，它基于[FastChat](https://github.com/lm-sys/FastChat)，并使用[vicuna-13b](https://huggingface.co/Tribbiani/vicuna-13b)作为基础模型。此外，此程序结合了[langchain](https://github.com/hwchase17/langchain)和[llama-index](https://github.com/jerryjliu/llama_index)基于现有知识库进行[In-Context Learning](https://arxiv.org/abs/2301.00234)来对其进行数据库相关知识的增强。它可以进行SQL生成、SQL诊断、数据库知识问答等一系列的工作。 
 
 
+## 项目方案
+![](ttps://github.com/csunny/DB-GPT/blob/dev/asserts/pilot.png)
+
 [DB-GPT](https://github.com/csunny/DB-GPT) is an experimental open-source application that builds upon the [FastChat](https://github.com/lm-sys/FastChat) model and uses vicuna as its base model. Additionally, it looks like this application incorporates langchain and llama-index embedding knowledge to improve Database-QA capabilities. 
 
 Overall, it appears to be a sophisticated and innovative tool for working with databases. If you have any specific questions about how to use or implement DB-GPT in your work, please let me know and I'll do my best to assist you.
@@ -15,6 +18,7 @@ Run on an RTX 4090 GPU (The origin mov not sped up!, [YouTube地址](https://www
 
 ![](https://github.com/csunny/DB-GPT/blob/dev/asserts/演示.gif)
 
+
 - SQL生成示例
 首先选择对应的数据库, 然后模型即可根据对应的数据库Schema信息生成SQL
 

pilot/configs/model_config.py

@@ -13,9 +13,11 @@ LOGDIR = os.path.join(root_path, "logs")
 DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
 llm_model_config = {
     "flan-t5-base": os.path.join(model_path, "flan-t5-base"),
-    "vicuna-13b": os.path.join(model_path, "vicuna-13b")
+    "vicuna-13b": os.path.join(model_path, "vicuna-13b"),
+    "sentence-transforms": os.path.join(model_path, "all-MiniLM-L6-v2")
 }
 
+
 LLM_MODEL = "vicuna-13b"
 LIMIT_MODEL_CONCURRENCY = 5
 MAX_POSITION_EMBEDDINGS = 2048

pilot/vector_store/extract_tovec.py

@@ -5,6 +5,8 @@
 from langchain.text_splitter import CharacterTextSplitter
 from langchain.vectorstores import Chroma
 from pilot.model.vicuna_llm import VicunaEmbeddingLLM
+# from langchain.embeddings import SentenceTransformerEmbeddings
+
 
 embeddings = VicunaEmbeddingLLM()
 
@@ -20,4 +22,17 @@ def knownledge_tovec(filename):
     return docsearch
 
 
+# def knownledge_tovec_st(filename):
+#     """ Use sentence transformers to embedding the document.
+#         https://github.com/UKPLab/sentence-transformers
+#     """
+#     from pilot.configs.model_config import llm_model_config
+#     embeddings = SentenceTransformerEmbeddings(model=llm_model_config["sentence-transforms"])
 
+#     with open(filename, "r") as f:
+#         knownledge = f.read()
+    
+#     text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=0)
+#     texts = text_splitter(knownledge)
+#     docsearch = Chroma.from_texts(texts, embeddings, metadatas=[{"source": str(i)} for i in range(len(texts))])
+#     return docsearch