Fix the code structure and the plugin name in NeuralChat (#233)

* revision Signed-off-by: XuhuiRen <xuhui.ren@intel.com> * revision Signed-off-by: XuhuiRen <xuhui.ren@intel.com> * revision Signed-off-by: XuhuiRen <xuhui.ren@intel.com> * Update intent_detection.py * Rename context_util.py to context_utils.py * revision Signed-off-by: XuhuiRen <xuhui.ren@intel.com> * add Signed-off-by: XuhuiRen <xuhui.ren@intel.com> * fix parm value Signed-off-by: XuhuiRen <xuhui.ren@intel.com> * fix Signed-off-by: XuhuiRen <xuhui.ren@intel.com> * fix Signed-off-by: XuhuiRen <xuhui.ren@intel.com> --------- Signed-off-by: XuhuiRen <xuhui.ren@intel.com>
intel · Sep 5, 2023 · 486e9ec · 486e9ec
1 parent cd4c33d
commit 486e9ec
Show file tree

Hide file tree

Showing 16 changed files with 27 additions and 43 deletions.
diff --git a/intel_extension_for_transformers/neural_chat/chatbot.py b/intel_extension_for_transformers/neural_chat/chatbot.py
@@ -30,7 +30,7 @@
 from .pipeline.plugins.audio.tts import TextToSpeech
 from .pipeline.plugins.audio.tts_chinese import ChineseTextToSpeech
 from .pipeline.plugins.security import SafetyChecker
-from .pipeline.plugins.retrievals import QA_Client
+from .pipeline.plugins.retrieval import Agent_QA
 from .models.llama_model import LlamaModel
 from .models.mpt_model import MptModel
 from .models.chatglm_model import ChatGlmModel

diff --git a/intel_extension_for_transformers/neural_chat/config.py b/intel_extension_for_transformers/neural_chat/config.py
@@ -26,9 +26,9 @@
 from .pipeline.plugins.audio.asr_chinese import ChineseAudioSpeechRecognition
 from .pipeline.plugins.audio.tts import TextToSpeech
 from .pipeline.plugins.audio.tts_chinese import ChineseTextToSpeech
-from .pipeline.plugins.retrievals.indexing import DocumentIndexing
-from .pipeline.plugins.retrievals.retrieval import SparseBM25Retriever, ChromaRetriever
-from .pipeline.plugins.intent_detector import IntentDetector
+from .pipeline.plugins.retrieval.indexing import DocumentIndexing
+from .pipeline.plugins.retrieval import SparseBM25Retriever, ChromaRetriever
+from .pipeline.plugins.retrieval.detector import IntentDetector
 from .pipeline.plugins.security import SafetyChecker
 from .plugins import plugins
 

diff --git a/intel_extension_for_transformers/neural_chat/examples/retrieval/README.md b/intel_extension_for_transformers/neural_chat/examples/retrieval/README.md
@@ -41,7 +41,7 @@ The user can costomize the retrieval parameters to meet the personal demmads for
 ```
 >>>persist_dir [str]: The local path to save the processed database. Default to "./output".
 
->>>process [bool]: Select to process the too long document into small chucks. Default to "False".
+>>>process [bool]: Select to process the too long document into small chucks. Default to "True".
 
 >>>input_path [str]: The user local path to a file folder or a specific file path. The code itself will check the path is a folder or a file. If it is a folder, the code will process all the files in the given folder. If it is a file, the code will prcess this single file.
 
@@ -55,9 +55,9 @@ The user can costomize the retrieval parameters to meet the personal demmads for
     
 >>>top_k [int]: The number of the retrieved documents. Default to "1".
 
->>>search_type [str]: Select a ranking method for dense retrieval from "mmr" or "similarity". "similarity" will return the most similar docs to the input query. "mmr" will rank the docs using the maximal marginal relevance method. Deault to "mmr".
+>>>search_type [str]: Select a ranking method for dense retrieval from "mmr", "similarity" and "similarity_score_threshold". "similarity" will return the most similar docs to the input query. "mmr" will rank the docs using the maximal marginal relevance method. "similarity_score_threshold" will return the mosy similar docs that also meet the threshold. Deault to "mmr".
 
->>>search_kwargs [dict]: Used by dense retrieval. Should be in the same format with {"k":1, "fetch_k":5}. "fetch_k" determines the amount of documents to pass to the ranking algorithm. Default to {"k":1, "fetch_k":5}.
+>>>search_kwargs [dict]: Used by dense retrieval. Should be in the same format like {"k":1, "fetch_k":5}. "k" is the amount of documents to return. "score_threshold" is the minimal relevance threshold for "similarity_score_threshold" search. "lambda_mult" is the diversity of results returned by "mmr". "fetch_k" determines the amount of documents to pass to the "mmr" algorithm. Default to {"k":1, "fetch_k":5}.
 ```
 
 
diff --git a/...chat/pipeline/plugins/prompts/__init__.py → ..._chat/pipeline/plugins/prompt/__init__.py b/...chat/pipeline/plugins/prompts/__init__.py → ..._chat/pipeline/plugins/prompt/__init__.py
@@ -15,4 +15,4 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from .prompt import generate_intent_prompt, generate_qa_prompt, generate_prompt
+from .prompt_template import generate_intent_prompt, generate_qa_prompt, generate_prompt
diff --git a/...l_chat/pipeline/plugins/prompts/prompt.py → ...ipeline/plugins/prompt/prompt_template.py b/...l_chat/pipeline/plugins/prompts/prompt.py → ...ipeline/plugins/prompt/prompt_template.py
diff --git a/...e/plugins/retrievals/indexing/__init__.py → ...at/pipeline/plugins/retrieval/__init__.py b/...e/plugins/retrievals/indexing/__init__.py → ...at/pipeline/plugins/retrieval/__init__.py
@@ -15,4 +15,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from .document_parser import DocumentIndexing
+from .retrieval_agent import Agent_QA
+from .retrieval_base import Retriever
+from .retrieval_bm25 import SparseBM25Retriever
+from .retrieval_chroma import ChromaRetriever
diff --git a/...eline/plugins/intent_detector/__init__.py → ...ne/plugins/retrieval/detector/__init__.py b/...eline/plugins/intent_detector/__init__.py → ...ne/plugins/retrieval/detector/__init__.py
diff --git a/...ugins/intent_detector/intent_detection.py → ...ns/retrieval/detector/intent_detection.py b/...ugins/intent_detector/intent_detection.py → ...ns/retrieval/detector/intent_detection.py
@@ -18,7 +18,7 @@
 
 import transformers
 import torch
-from ..prompts import generate_intent_prompt
+from intel_extension_for_transformers.neural_chat.pipeline.plugins.prompt import generate_intent_prompt
 from intel_extension_for_transformers.llm.inference import predict
 
 class IntentDetector:

diff --git a/...t/pipeline/plugins/retrievals/__init__.py → ...ne/plugins/retrieval/indexing/__init__.py b/...t/pipeline/plugins/retrievals/__init__.py → ...ne/plugins/retrieval/indexing/__init__.py
@@ -15,4 +15,4 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from .retrievals import QA_Client
+from .indexing import DocumentIndexing
diff --git a/...line/plugins/retrievals/indexing/utils.py → ...ugins/retrieval/indexing/context_utils.py b/...line/plugins/retrievals/indexing/utils.py → ...ugins/retrieval/indexing/context_utils.py
@@ -73,7 +73,7 @@ def read_docx(doc_path):
 
 def read_md(md_path):
     """Read docx file."""
-    loader = UnstructuredMarkdownLoader("instruction_data.md")
+    loader = UnstructuredMarkdownLoader(md_path)
     text = loader.load()[0].page_content
     return text
 

diff --git a/...ns/retrievals/indexing/document_parser.py → ...ne/plugins/retrieval/indexing/indexing.py b/...ns/retrievals/indexing/document_parser.py → ...ne/plugins/retrieval/indexing/indexing.py
@@ -24,12 +24,12 @@
 from langchain.embeddings import HuggingFaceEmbeddings, HuggingFaceInstructEmbeddings
 from langchain.vectorstores import Chroma
 from haystack.schema import Document as SDocument
-from .utils import load_unstructured_data, laod_structured_data, get_chuck_data
+from .context_utils import load_unstructured_data, laod_structured_data, get_chuck_data
 
 
 class DocumentIndexing:
     def __init__(self, retrieval_type="dense", document_store=None, persist_dir="./output",
-                 process=False, embedding_model="hkunlp/instructor-large", max_length=512):
+                 process=True, embedding_model="hkunlp/instructor-large", max_length=512):
         """
         Wrapper for document indexing. Support dense and sparse indexing method.
         """

diff --git a/...pipeline/plugins/retrievals/retrievals.py → ...line/plugins/retrieval/retrieval_agent.py b/...pipeline/plugins/retrievals/retrievals.py → ...line/plugins/retrieval/retrieval_agent.py
@@ -18,15 +18,16 @@
 import os
 import torch
 import transformers
-from intel_extension_for_transformers.neural_chat.pipeline.plugins.intent_detector import IntentDetector
-from intel_extension_for_transformers.neural_chat.pipeline.plugins.retrievals.indexing import DocumentIndexing
-from intel_extension_for_transformers.neural_chat.pipeline.plugins.retrievals.retrieval import Retriever
+from intel_extension_for_transformers.neural_chat.pipeline.plugins.retrieval import Retriever
+from intel_extension_for_transformers.neural_chat.pipeline.plugins.retrieval.detector import IntentDetector
+from intel_extension_for_transformers.neural_chat.pipeline.plugins.retrieval.indexing import DocumentIndexing
+from intel_extension_for_transformers.neural_chat.pipeline.plugins.prompt import generate_qa_prompt, generate_prompt
 from intel_extension_for_transformers.neural_chat.plugins import register_plugin
-from intel_extension_for_transformers.neural_chat.pipeline.plugins.prompts import generate_qa_prompt, generate_prompt
+
 
 @register_plugin("retrieval")
-class QA_Client():
-    def __init__(self, persist_dir="./output", process=False, input_path=None,
+class Agent_QA():
+    def __init__(self, persist_dir="./output", process=True, input_path=None,
                  embedding_model="hkunlp/instructor-large", max_length=512, retrieval_type="dense",
                  document_store=None, top_k=1, search_type="mmr", search_kwargs={"k": 1, "fetch_k": 5}):
         self.model = None

diff --git a/...plugins/retrievals/retrieval/retrieval.py → ...eline/plugins/retrieval/retrieval_base.py b/...plugins/retrievals/retrieval/retrieval.py → ...eline/plugins/retrieval/retrieval_base.py
@@ -17,11 +17,11 @@
 
 """The class defination for the retriever. Supporting langchain-based and haystack-based retriever."""
 
-from .bm25_retrieval import SparseBM25Retriever
-from .chroma_retrieval import ChromaRetriever
+from .retrieval_bm25 import SparseBM25Retriever
+from .retrieval_chroma import ChromaRetriever
 
 class Retriever():
-    """Retrieve the document database with BM25 sparse algorithm."""
+    """The wrapper for sparse retriever and dense retriever."""
 
     def __init__(self, retrieval_type="dense", document_store=None,
                  top_k=1, search_type="mmr", search_kwargs={"k": 1, "fetch_k": 5}):

diff --git a/...ns/retrievals/retrieval/bm25_retrieval.py → ...eline/plugins/retrieval/retrieval_bm25.py b/...ns/retrievals/retrieval/bm25_retrieval.py → ...eline/plugins/retrieval/retrieval_bm25.py
diff --git a/.../retrievals/retrieval/chroma_retrieval.py → ...ine/plugins/retrieval/retrieval_chroma.py b/.../retrievals/retrieval/chroma_retrieval.py → ...ine/plugins/retrieval/retrieval_chroma.py
diff --git a/..._extension_for_transformers/neural_chat/pipeline/plugins/retrievals/retrieval/__init__.py b/..._extension_for_transformers/neural_chat/pipeline/plugins/retrievals/retrieval/__init__.py