Spaces:
Runtime error
Runtime error
Fecalisboa
commited on
Commit
•
04a7754
1
Parent(s):
510c455
Update app.py
Browse files
app.py
CHANGED
@@ -15,7 +15,6 @@ from langchain.chains import ConversationChain
|
|
15 |
from langchain.memory import ConversationBufferMemory
|
16 |
from langchain_community.llms import HuggingFaceEndpoint
|
17 |
import torch
|
18 |
-
|
19 |
api_token = os.getenv("HF_TOKEN")
|
20 |
|
21 |
list_llm = ["meta-llama/Meta-Llama-3-8B-Instruct", "mistralai/Mistral-7B-Instruct-v0.3"]
|
@@ -98,26 +97,40 @@ def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db, in
|
|
98 |
progress(0.9, desc="Done!")
|
99 |
return qa_chain
|
100 |
|
101 |
-
|
102 |
-
|
103 |
-
|
104 |
-
|
105 |
-
|
106 |
-
|
107 |
-
|
108 |
-
|
109 |
-
|
110 |
-
)
|
111 |
-
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
119 |
progress(0.9, desc="Done!")
|
120 |
-
return
|
|
|
|
|
|
|
|
|
|
|
|
|
121 |
|
122 |
def format_chat_history(message, chat_history):
|
123 |
formatted_chat_history = []
|
@@ -143,6 +156,30 @@ def conversation(qa_chain, message, history):
|
|
143 |
new_history = history + [(message, response_answer)]
|
144 |
return qa_chain, gr.update(value=""), new_history, response_source1, response_source1_page, response_source2, response_source2_page, response_source3, response_source3_page
|
145 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
146 |
def conversation_no_doc(llm, message, history):
|
147 |
formatted_chat_history = format_chat_history(message, history)
|
148 |
response = llm({"question": message, "chat_history": formatted_chat_history})
|
@@ -156,33 +193,6 @@ def upload_file(file_obj):
|
|
156 |
list_file_path.append(file.name)
|
157 |
return list_file_path
|
158 |
|
159 |
-
def initialize_database(list_file_obj, chunk_size, chunk_overlap, db_type, progress=gr.Progress()):
|
160 |
-
list_file_path = [x.name for x in list_file_obj if x is not None]
|
161 |
-
progress(0.1, desc="Creating collection name...")
|
162 |
-
collection_name = create_collection_name(list_file_path[0])
|
163 |
-
progress(0.25, desc="Loading document...")
|
164 |
-
doc_splits = load_doc(list_file_path, chunk_size, chunk_overlap)
|
165 |
-
progress(0.5, desc="Generating vector database...")
|
166 |
-
vector_db = create_db(doc_splits, collection_name, db_type)
|
167 |
-
progress(0.9, desc="Done!")
|
168 |
-
return vector_db, collection_name, "Complete!"
|
169 |
-
|
170 |
-
def create_collection_name(filepath):
|
171 |
-
collection_name = Path(filepath).stem
|
172 |
-
collection_name = collection_name.replace(" ", "-")
|
173 |
-
collection_name = unidecode(collection_name)
|
174 |
-
collection_name = re.sub('[^A-Za-z0-9]+', '-', collection_name)
|
175 |
-
collection_name = collection_name[:50]
|
176 |
-
if len(collection_name) < 3:
|
177 |
-
collection_name = collection_name + 'xyz'
|
178 |
-
if not collection_name[0].isalnum():
|
179 |
-
collection_name = 'A' + collection_name[1:]
|
180 |
-
if not collection_name[-1].isalnum():
|
181 |
-
collection_name = collection_name[:-1] + 'Z'
|
182 |
-
print('Filepath: ', filepath)
|
183 |
-
print('Collection name: ', collection_name)
|
184 |
-
return collection_name
|
185 |
-
|
186 |
def demo():
|
187 |
with gr.Blocks(theme="base") as demo:
|
188 |
vector_db = gr.State()
|
@@ -257,20 +267,6 @@ def demo():
|
|
257 |
clear_btn = gr.ClearButton([msg, chatbot], value="Clear conversation")
|
258 |
|
259 |
with gr.Tab("Step 6 - Chatbot without document"):
|
260 |
-
with gr.Row():
|
261 |
-
llm_no_doc_btn = gr.Radio(list_llm_simple,
|
262 |
-
label="LLM models", value=list_llm_simple[0], type="index", info="Choose your LLM model for chatbot without document")
|
263 |
-
with gr.Accordion("Advanced options - LLM model", open=False):
|
264 |
-
with gr.Row():
|
265 |
-
slider_temperature_no_doc = gr.Slider(minimum=0.01, maximum=1.0, value=0.7, step=0.1, label="Temperature", info="Model temperature", interactive=True)
|
266 |
-
with gr.Row():
|
267 |
-
slider_maxtokens_no_doc = gr.Slider(minimum=224, maximum=4096, value=1024, step=32, label="Max Tokens", info="Model max tokens", interactive=True)
|
268 |
-
with gr.Row():
|
269 |
-
slider_topk_no_doc = gr.Slider(minimum=1, maximum=10, value=3, step=1, label="top-k samples", info="Model top-k samples", interactive=True)
|
270 |
-
with gr.Row():
|
271 |
-
llm_no_doc_progress = gr.Textbox(value="None", label="LLM initialization for chatbot without document")
|
272 |
-
with gr.Row():
|
273 |
-
llm_no_doc_init_btn = gr.Button("Initialize LLM for Chatbot without document")
|
274 |
chatbot_no_doc = gr.Chatbot(height=300)
|
275 |
with gr.Row():
|
276 |
msg_no_doc = gr.Textbox(placeholder="Type message to chat with lucIAna", container=True)
|
@@ -282,10 +278,10 @@ def demo():
|
|
282 |
db_btn.click(initialize_database,
|
283 |
inputs=[document, slider_chunk_size, slider_chunk_overlap, db_type_radio],
|
284 |
outputs=[vector_db, collection_name, db_progress])
|
285 |
-
set_prompt_btn.click(lambda prompt:
|
286 |
inputs=prompt_input,
|
287 |
outputs=initial_prompt)
|
288 |
-
qachain_btn.click(
|
289 |
inputs=[llm_btn, slider_temperature, slider_maxtokens, slider_topk, vector_db, initial_prompt],
|
290 |
outputs=[qa_chain, llm_progress]).then(lambda:[None,"",0,"",0,"",0],
|
291 |
inputs=None,
|
@@ -306,11 +302,7 @@ def demo():
|
|
306 |
outputs=[chatbot, doc_source1, source1_page, doc_source2, source2_page, doc_source3, source3_page],
|
307 |
queue=False)
|
308 |
|
309 |
-
#
|
310 |
-
llm_no_doc_init_btn.click(initialize_llm_no_doc,
|
311 |
-
inputs=[llm_no_doc_btn, slider_temperature_no_doc, slider_maxtokens_no_doc, slider_topk_no_doc, initial_prompt],
|
312 |
-
outputs=[llm_no_doc, llm_no_doc_progress])
|
313 |
-
|
314 |
submit_btn_no_doc.click(conversation_no_doc,
|
315 |
inputs=[llm_no_doc, msg_no_doc, chatbot_no_doc],
|
316 |
outputs=[llm_no_doc, msg_no_doc, chatbot_no_doc],
|
@@ -320,6 +312,27 @@ def demo():
|
|
320 |
outputs=[chatbot_no_doc, msg_no_doc],
|
321 |
queue=False)
|
322 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
323 |
demo.queue().launch(debug=True, share=True)
|
324 |
|
325 |
if __name__ == "__main__":
|
|
|
15 |
from langchain.memory import ConversationBufferMemory
|
16 |
from langchain_community.llms import HuggingFaceEndpoint
|
17 |
import torch
|
|
|
18 |
api_token = os.getenv("HF_TOKEN")
|
19 |
|
20 |
list_llm = ["meta-llama/Meta-Llama-3-8B-Instruct", "mistralai/Mistral-7B-Instruct-v0.3"]
|
|
|
97 |
progress(0.9, desc="Done!")
|
98 |
return qa_chain
|
99 |
|
100 |
+
# Generate collection name for vector database
|
101 |
+
def create_collection_name(filepath):
|
102 |
+
collection_name = Path(filepath).stem
|
103 |
+
collection_name = collection_name.replace(" ", "-")
|
104 |
+
collection_name = unidecode(collection_name)
|
105 |
+
collection_name = re.sub('[^A-Za-z0-9]+', '-', collection_name)
|
106 |
+
collection_name = collection_name[:50]
|
107 |
+
if len(collection_name) < 3:
|
108 |
+
collection_name = collection_name + 'xyz'
|
109 |
+
if not collection_name[0].isalnum():
|
110 |
+
collection_name = 'A' + collection_name[1:]
|
111 |
+
if not collection_name[-1].isalnum():
|
112 |
+
collection_name = collection_name[:-1] + 'Z'
|
113 |
+
print('Filepath: ', filepath)
|
114 |
+
print('Collection name: ', collection_name)
|
115 |
+
return collection_name
|
116 |
+
|
117 |
+
# Initialize database
|
118 |
+
def initialize_database(list_file_obj, chunk_size, chunk_overlap, db_type, progress=gr.Progress()):
|
119 |
+
list_file_path = [x.name for x in list_file_obj if x is not None]
|
120 |
+
progress(0.1, desc="Creating collection name...")
|
121 |
+
collection_name = create_collection_name(list_file_path[0])
|
122 |
+
progress(0.25, desc="Loading document...")
|
123 |
+
doc_splits = load_doc(list_file_path, chunk_size, chunk_overlap)
|
124 |
+
progress(0.5, desc="Generating vector database...")
|
125 |
+
vector_db = create_db(doc_splits, collection_name, db_type)
|
126 |
progress(0.9, desc="Done!")
|
127 |
+
return vector_db, collection_name, "Complete!"
|
128 |
+
|
129 |
+
def initialize_LLM(llm_option, llm_temperature, max_tokens, top_k, vector_db, initial_prompt, progress=gr.Progress()):
|
130 |
+
llm_name = list_llm[llm_option]
|
131 |
+
print("llm_name: ", llm_name)
|
132 |
+
qa_chain = initialize_llmchain(llm_name, llm_temperature, max_tokens, top_k, vector_db, initial_prompt, progress)
|
133 |
+
return qa_chain, "Complete!"
|
134 |
|
135 |
def format_chat_history(message, chat_history):
|
136 |
formatted_chat_history = []
|
|
|
156 |
new_history = history + [(message, response_answer)]
|
157 |
return qa_chain, gr.update(value=""), new_history, response_source1, response_source1_page, response_source2, response_source2_page, response_source3, response_source3_page
|
158 |
|
159 |
+
def initialize_llm_no_doc(llm_model, temperature, max_tokens, top_k, initial_prompt, progress=gr.Progress()):
|
160 |
+
progress(0.1, desc="Initializing HF tokenizer...")
|
161 |
+
|
162 |
+
progress(0.5, desc="Initializing HF Hub...")
|
163 |
+
|
164 |
+
llm = HuggingFaceEndpoint(
|
165 |
+
repo_id=llm_model,
|
166 |
+
huggingfacehub_api_token=api_token,
|
167 |
+
temperature=temperature,
|
168 |
+
max_new_tokens=max_tokens,
|
169 |
+
top_k=top_k,
|
170 |
+
)
|
171 |
+
|
172 |
+
progress(0.75, desc="Defining buffer memory...")
|
173 |
+
memory = ConversationBufferMemory(
|
174 |
+
memory_key="chat_history",
|
175 |
+
output_key='answer',
|
176 |
+
return_messages=True
|
177 |
+
)
|
178 |
+
conversation_chain = ConversationChain(llm=llm, memory=memory, verbose=False)
|
179 |
+
conversation_chain({"question": initial_prompt})
|
180 |
+
progress(0.9, desc="Done!")
|
181 |
+
return conversation_chain
|
182 |
+
|
183 |
def conversation_no_doc(llm, message, history):
|
184 |
formatted_chat_history = format_chat_history(message, history)
|
185 |
response = llm({"question": message, "chat_history": formatted_chat_history})
|
|
|
193 |
list_file_path.append(file.name)
|
194 |
return list_file_path
|
195 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
196 |
def demo():
|
197 |
with gr.Blocks(theme="base") as demo:
|
198 |
vector_db = gr.State()
|
|
|
267 |
clear_btn = gr.ClearButton([msg, chatbot], value="Clear conversation")
|
268 |
|
269 |
with gr.Tab("Step 6 - Chatbot without document"):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
270 |
chatbot_no_doc = gr.Chatbot(height=300)
|
271 |
with gr.Row():
|
272 |
msg_no_doc = gr.Textbox(placeholder="Type message to chat with lucIAna", container=True)
|
|
|
278 |
db_btn.click(initialize_database,
|
279 |
inputs=[document, slider_chunk_size, slider_chunk_overlap, db_type_radio],
|
280 |
outputs=[vector_db, collection_name, db_progress])
|
281 |
+
set_prompt_btn.click(lambda prompt: prompt,
|
282 |
inputs=prompt_input,
|
283 |
outputs=initial_prompt)
|
284 |
+
qachain_btn.click(initialize_LLM,
|
285 |
inputs=[llm_btn, slider_temperature, slider_maxtokens, slider_topk, vector_db, initial_prompt],
|
286 |
outputs=[qa_chain, llm_progress]).then(lambda:[None,"",0,"",0,"",0],
|
287 |
inputs=None,
|
|
|
302 |
outputs=[chatbot, doc_source1, source1_page, doc_source2, source2_page, doc_source3, source3_page],
|
303 |
queue=False)
|
304 |
|
305 |
+
# Chatbot events without document
|
|
|
|
|
|
|
|
|
306 |
submit_btn_no_doc.click(conversation_no_doc,
|
307 |
inputs=[llm_no_doc, msg_no_doc, chatbot_no_doc],
|
308 |
outputs=[llm_no_doc, msg_no_doc, chatbot_no_doc],
|
|
|
312 |
outputs=[chatbot_no_doc, msg_no_doc],
|
313 |
queue=False)
|
314 |
|
315 |
+
# Initialize LLM without document for conversation
|
316 |
+
with gr.Tab("Initialize LLM for Chatbot without document"):
|
317 |
+
with gr.Row():
|
318 |
+
llm_no_doc_btn = gr.Radio(list_llm_simple,
|
319 |
+
label="LLM models", value=list_llm_simple[0], type="index", info="Choose your LLM model for chatbot without document")
|
320 |
+
with gr.Accordion("Advanced options - LLM model", open=False):
|
321 |
+
with gr.Row():
|
322 |
+
slider_temperature_no_doc = gr.Slider(minimum=0.01, maximum=1.0, value=0.7, step=0.1, label="Temperature", info="Model temperature", interactive=True)
|
323 |
+
with gr.Row():
|
324 |
+
slider_maxtokens_no_doc = gr.Slider(minimum=224, maximum=4096, value=1024, step=32, label="Max Tokens", info="Model max tokens", interactive=True)
|
325 |
+
with gr.Row():
|
326 |
+
slider_topk_no_doc = gr.Slider(minimum=1, maximum=10, value=3, step=1, label="top-k samples", info="Model top-k samples", interactive=True)
|
327 |
+
with gr.Row():
|
328 |
+
llm_no_doc_progress = gr.Textbox(value="None", label="LLM initialization for chatbot without document")
|
329 |
+
with gr.Row():
|
330 |
+
llm_no_doc_init_btn = gr.Button("Initialize LLM for Chatbot without document")
|
331 |
+
|
332 |
+
llm_no_doc_init_btn.click(initialize_llm_no_doc,
|
333 |
+
inputs=[llm_no_doc_btn, slider_temperature_no_doc, slider_maxtokens_no_doc, slider_topk_no_doc, initial_prompt],
|
334 |
+
outputs=[llm_no_doc, llm_no_doc_progress])
|
335 |
+
|
336 |
demo.queue().launch(debug=True, share=True)
|
337 |
|
338 |
if __name__ == "__main__":
|