Update app.py
Browse files
app.py
CHANGED
|
@@ -119,24 +119,38 @@ model_kwargs = {"max_length": max_length, "pad_token_id": 50256, "pad_token": "[
|
|
| 119 |
# "pad_token_id": 50256, "pad_token": "[PAD]",
|
| 120 |
# "padding": 'longest' #for open end generation
|
| 121 |
# })
|
| 122 |
-
llm = ChatOpenAI(streaming=True,
|
| 123 |
prompt_template = """You are going to
|
| 124 |
receive a context with information about related works, papers and tools for the PhD of Martín Dévora Pajares
|
| 125 |
(exoplanets field), also appearing as M. Dévora-Pajares.
|
| 126 |
-
You can only answer things based on the question
|
| 127 |
-
with the user.
|
| 128 |
Context: {context}
|
| 129 |
-
Question: {question}
|
| 130 |
-
Memory: {chat_history}"""
|
| 131 |
|
| 132 |
PROMPT = PromptTemplate(
|
| 133 |
-
template=prompt_template, input_variables=["context", "question"
|
| 134 |
)
|
| 135 |
chain_type_kwargs = {"prompt": PROMPT}
|
| 136 |
-
memory = ConversationBufferWindowMemory(k=
|
| 137 |
-
|
| 138 |
-
|
| 139 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 140 |
# while True:
|
| 141 |
# query = input("Question: ")
|
| 142 |
# #docs = vector_store.similarity_search(query)
|
|
@@ -144,10 +158,18 @@ qa = RetrievalQA.from_chain_type(llm=llm, chain_type="stuff", retriever=vector_s
|
|
| 144 |
# response = qa({"query": query})
|
| 145 |
# print(f"Answer: {response['result']}")
|
| 146 |
# print(f"Sources: {response['source_documents']}")
|
| 147 |
-
|
| 148 |
def chat(message, history):
|
| 149 |
result = qa({"query": message})
|
| 150 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 151 |
# print(f"Answer: {response['result']}")
|
| 152 |
# print(f"Sources: {response['source_documents']}")
|
| 153 |
|
|
@@ -156,8 +178,10 @@ gr.ChatInterface(chat, chatbot=gr.Chatbot(),
|
|
| 156 |
title="PlanetHunters informative chatbot",
|
| 157 |
description="Ask anything related to SHERLOCK, MATRIX, WATSON and their related papers!",
|
| 158 |
theme="soft",
|
| 159 |
-
examples=["What is SHERLOCK?", "What is MATRIX?", "What are the foundational papers of SHERLOCK?"
|
| 160 |
-
|
|
|
|
| 161 |
retry_btn=None,
|
| 162 |
undo_btn="Delete Previous",
|
| 163 |
clear_btn="Clear").queue().launch()
|
|
|
|
|
|
| 119 |
# "pad_token_id": 50256, "pad_token": "[PAD]",
|
| 120 |
# "padding": 'longest' #for open end generation
|
| 121 |
# })
|
| 122 |
+
llm = ChatOpenAI(streaming=True, model_name=model_name, temperature=0.9)
|
| 123 |
prompt_template = """You are going to
|
| 124 |
receive a context with information about related works, papers and tools for the PhD of Martín Dévora Pajares
|
| 125 |
(exoplanets field), also appearing as M. Dévora-Pajares.
|
| 126 |
+
You can only answer things based on the question and a context to answer it.
|
|
|
|
| 127 |
Context: {context}
|
| 128 |
+
Question: {question}"""
|
| 129 |
+
#Memory: {chat_history}"""
|
| 130 |
|
| 131 |
PROMPT = PromptTemplate(
|
| 132 |
+
template=prompt_template, input_variables=["context", "question"]
|
| 133 |
)
|
| 134 |
chain_type_kwargs = {"prompt": PROMPT}
|
| 135 |
+
memory = ConversationBufferWindowMemory(k=4, memory_key="chat_history", input_key="question", output_key='answer',
|
| 136 |
+
return_messages=True)
|
| 137 |
+
|
| 138 |
+
retriever = vector_store.as_retriever()
|
| 139 |
+
# memory = VectorStoreRetrieverMemory(
|
| 140 |
+
# memory_key="chat_history",
|
| 141 |
+
# return_docs=False,
|
| 142 |
+
# retriever=retriever,
|
| 143 |
+
# return_messages=True,
|
| 144 |
+
# )
|
| 145 |
+
qa = RetrievalQA.from_chain_type(llm=llm, chain_type="stuff", retriever=retriever,
|
| 146 |
+
chain_type_kwargs=chain_type_kwargs, return_source_documents=True)
|
| 147 |
+
#memory=memory)
|
| 148 |
+
# qa = ConversationalRetrievalChain.from_llm(
|
| 149 |
+
# llm=llm, retriever=retriever, memory=memory,
|
| 150 |
+
# return_source_documents=True,
|
| 151 |
+
# combine_docs_chain_kwargs=chain_type_kwargs)
|
| 152 |
+
|
| 153 |
+
|
| 154 |
# while True:
|
| 155 |
# query = input("Question: ")
|
| 156 |
# #docs = vector_store.similarity_search(query)
|
|
|
|
| 158 |
# response = qa({"query": query})
|
| 159 |
# print(f"Answer: {response['result']}")
|
| 160 |
# print(f"Sources: {response['source_documents']}")
|
| 161 |
+
|
| 162 |
def chat(message, history):
|
| 163 |
result = qa({"query": message})
|
| 164 |
+
sources_list = []
|
| 165 |
+
for item in result['source_documents']:
|
| 166 |
+
if 'http' in item.metadata['source']:
|
| 167 |
+
sources_list = sources_list + [item.metadata['source']]
|
| 168 |
+
sources_list = list(set(sources_list))
|
| 169 |
+
response = result['result']
|
| 170 |
+
if len(sources_list) > 0:
|
| 171 |
+
response = response + '\n\nSources:' + '\n\n'.join(sources_list)
|
| 172 |
+
yield response
|
| 173 |
# print(f"Answer: {response['result']}")
|
| 174 |
# print(f"Sources: {response['source_documents']}")
|
| 175 |
|
|
|
|
| 178 |
title="PlanetHunters informative chatbot",
|
| 179 |
description="Ask anything related to SHERLOCK, MATRIX, WATSON and their related papers!",
|
| 180 |
theme="soft",
|
| 181 |
+
examples=["What is SHERLOCK?", "What is MATRIX?", "What are the foundational papers of SHERLOCK?",
|
| 182 |
+
"How should I cite SHERLOCK?", "How should I cite MATRIX?"],
|
| 183 |
+
cache_examples=False,
|
| 184 |
retry_btn=None,
|
| 185 |
undo_btn="Delete Previous",
|
| 186 |
clear_btn="Clear").queue().launch()
|
| 187 |
+
|