Fix the bug in rag example (Neural Chat) (#226)

* code 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> * revision Signed-off-by: XuhuiRen <xuhui.ren@intel.com> * revise Signed-off-by: XuhuiRen <xuhui.ren@intel.com> * Update __init__.py * Update document_parser.py * fixed import error. Signed-off-by: Ye, Xinyu <xinyu.ye@intel.com> --------- Signed-off-by: XuhuiRen <xuhui.ren@intel.com> Signed-off-by: Ye, Xinyu <xinyu.ye@intel.com> Co-authored-by: Ye, Xinyu <xinyu.ye@intel.com>

Author: XuhuiRen

intel_extension_for_transformers/neural_chat/chatbot.py

@@ -30,9 +30,8 @@
 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.security import SafetyChecker
 from .pipeline.plugins.retrievals import QA_Client
-from .pipeline.plugins.security.safety_checker import SafetyChecker
-from .pipeline.plugins.intent_detector import IntentDetector
 from .models.llama_model import LlamaModel
 from .models.mpt_model import MptModel
 from .models.chatglm_model import ChatGlmModel

intel_extension_for_transformers/neural_chat/cli/cli_commands.py

@@ -20,7 +20,7 @@
 from typing import List
 from ..utils.command import NeuralChatCommandDict
 from .base_executor import BaseCommandExecutor
-from ..config import PipelineConfig, FinetuningConfig, GenerationConfig # pylint: disable=E0611
+from ..config import PipelineConfig, TextGenerationFinetuningConfig, GenerationConfig
 from ..config import ModelArguments, DataArguments, FinetuningArguments
 from ..plugins import plugins
 from transformers import TrainingArguments
@@ -311,7 +311,7 @@ def execute(self, argv: List[str]) -> bool:
  training_args = TrainingArguments(output_dir="./output")
  finetune_args= FinetuningArguments()
 
- self.finetuneCfg = FinetuningConfig(model_args, data_args, training_args, finetune_args)
+ self.finetuneCfg = TextGenerationFinetuningConfig(model_args, data_args, training_args, finetune_args)
  try:
  res = self()
  print(res)

intel_extension_for_transformers/neural_chat/examples/retrieval/retrieval.py renamed to intel_extension_for_transformers/neural_chat/examples/retrieval/retrieval_chat.py

@@ -18,35 +18,23 @@
 import os
 import sys
 from transformers import TrainingArguments, HfArgumentParser
-from intel_extension_for_transformers.neural_chat.config import (
- PipelineConfig,
- RetrieverConfig,
- SafetyConfig,
- GenerationConfig
-)
+from intel_extension_for_transformers.neural_chat.config import PipelineConfig
 from intel_extension_for_transformers.neural_chat.chatbot import build_chatbot
 
 
 def main():
  # See all possible arguments in config.py
  # or by passing the --help flag to this script.
  # We now keep distinct sets of args, for a cleaner separation of concerns.
- parser = HfArgumentParser(
- (PipelineConfig, RetrieverConfig, SafetyConfig, GenerationConfig)
- )
+ parser = HfArgumentParser(PipelineConfig)
 
  if len(sys.argv) == 2 and sys.argv[1].endswith(".json"):
  # If we pass only one argument to the script and it's the path to a json file,
  # let's parse it to get our arguments.
- pipeline_args, retriever_args, safety_args, generation_args = parser.parse_json_file(
- json_file = os.path.abspath(sys.argv[1])
- )
+ pipeline_args= parser.parse_json_file(json_file = os.path.abspath(sys.argv[1]))
  else:
- (pipeline_args, retriever_args, safety_args, generation_args) = parser.parse_args_into_dataclasses()
+ pipeline_args= parser.parse_args_into_dataclasses()
 
- pipeline_args.saftey_config = safety_args
- pipeline_args.retrieval_config = retriever_args
- pipeline_args.generation_config = generation_args
  chatbot = build_chatbot(pipeline_args)
 
  response = chatbot.predict(query="What is IDM 2.0?", config=pipeline_args)

intel_extension_for_transformers/neural_chat/pipeline/plugins/retrievals/indexing/document_parser.py

@@ -97,14 +97,17 @@ def KB_construct(self, input):
 
  documents = []
  for data, meta in data_collection:
+ if len(data) < 5:
+ continue
  metadata = {"source": meta}
  new_doc = Document(page_content=data, metadata=metadata)
  documents.append(new_doc)
+ assert documents!= [], "The given file/files cannot be loaded." 
  embedding = HuggingFaceInstructEmbeddings(model_name=self.embedding_model)
  vectordb = Chroma.from_documents(documents=documents, embedding=embedding,
  persist_directory=self.persist_dir)
  vectordb.persist()
- print("success")
+ print("The local knowledge base has been successfully built!")
  return vectordb
  else:
  print("There might be some errors, please wait and try again!")
@@ -125,11 +128,13 @@ def KB_construct(self, input):
  documents = []
  for data, meta in data_collection:
  metadata = {"source": meta}
- # pylint: disable=E1123
- new_doc = SDocument(content=data, metadata=metadata)
+ if len(data) < 5:
+ continue
+ new_doc = SDocument(content=data, meta=metadata)
  documents.append(new_doc)
+ assert documents != [], "The given file/files cannot be loaded."
  document_store.write_documents(documents)
- print("success")
+ print("The local knowledge base has been successfully built!")
  return document_store
  else:
  print("There might be some errors, please wait and try again!")

intel_extension_for_transformers/neural_chat/pipeline/plugins/retrievals/indexing/utils.py

@@ -165,19 +165,21 @@ def laod_structured_data(input, process, max_length):
 
 def get_chuck_data(content, max_length, input):
  """Process the context to make it maintain a suitable length for the generation."""
- sentences = re.split('(?<=[;!.?])', content)
+ sentences = re.split('(?<=[!.?])', content)
 
  paragraphs = []
  current_length = 0
  count = 0
  current_paragraph = ""
  for sub_sen in sentences:
+ if sub_sen == "":
+ continue
  count +=1
  sentence_length = len(sub_sen)
  if current_length + sentence_length <= max_length:
  current_paragraph += sub_sen
  current_length += sentence_length
- if count == len(sentences):
+ if count == len(sentences) and len(current_paragraph.strip())>5:
  paragraphs.append([current_paragraph.strip() ,input])
  else:
  paragraphs.append([current_paragraph.strip() ,input])

intel_extension_for_transformers/neural_chat/pipeline/plugins/retrievals/retrieval/bm25_retrieval.py

@@ -22,7 +22,7 @@ class SparseBM25Retriever():
 
  def __int__(self, document_store = None, top_k = 1):
  assert document_store is not None, "Please give a document database for retrieving."
- self.retriever = BM25Retriever(document_store=document_store)
+ self.retriever = BM25Retriever(document_store=document_store, top_k=top_k)
 
  def query_the_database(self, query):
  documents = self.retriever.retrieve(query)

intel_extension_for_transformers/neural_chat/server/restful/finetune_api.py

@@ -26,7 +26,7 @@
  ModelArguments,
  DataArguments,
  FinetuningArguments,
- FinetuningConfig,
+ TextGenerationFinetuningConfig,
 )
 from intel_extension_for_transformers.neural_chat.server.restful.request import FinetuneRequest
 
@@ -66,7 +66,7 @@ def handle_finetune_request(self, request: FinetuneRequest) -> str:
  overwrite_output_dir=request.overwrite_output_dir
  )
  finetune_args = FinetuningArguments(peft=request.peft)
- finetune_cfg = FinetuningConfig(
+ finetune_cfg = TextGenerationFinetuningConfig(
  model_args=model_args,
  data_args=data_args,
  training_args=training_args,