init

app.py

@@ -0,0 +1,145 @@
+import string
+from transformers import AutoModelForCausalLM,AutoTokenizer, pipeline
+import gradio as gr
+import os
+import time
+from wikipediaapi import Wikipedia
+
+import feedparser
+import html2text
+
+# url = "https://rss.hotnews.ro"
+# url = "https://www.thecipherbrief.com/feed"
+# url = "https://feeds.simplecast.com/54nAGcIl"
+# url = "https://www.reddit.com/r/news/.rss"
+
+# def init_model():
+#     system_prompt = "### System:\nYou are StableBeluga, an AI that follows instructions extremely well. Help as much as you can.\n\n"
+#     tokenizer = AutoTokenizer.from_pretrained("stabilityai/StableBeluga-7B", use_fast=True)
+#     # model = AutoModelForCausalLM.from_pretrained("stabilityai/StableBeluga-7B", load_in_8bit=True, low_cpu_mem_usage=True, device_map=0)
+#     model = AutoModelForCausalLM.from_pretrained("stabilityai/StableBeluga-7B", low_cpu_mem_usage=True, device_map=0)
+#     return system_prompt, tokenizer, model
+
+# system_prompt, tokenizer, model = init_model()
+
+# def ask_assistant(prompt, token=tokenizer, md=model):
+#     inputs = token(make_prompt(prompt), return_tensors="pt").to("cuda")
+#     # output = md.generate(**inputs, do_sample=True, top_p=0.95, top_k=0, max_new_tokens=256).to('cpu')
+#     output = md.generate(**inputs, max_new_tokens=150).to('cuda')
+#     # return tokenizer.decode(output[0], skip_special_tokens=True)
+#     response = tokenizer.decode(output[0])
+#     return response[0].split('### Assistant: \n')[1]
+
+# def make_prompt(user, syst=system_prompt): 
+#     return f"{syst}### User: {user}\n\n### Assistant:\n"
+
+# def ask_assistant_in_context(question, context):
+#     prompt = f"""Answer the question with the help of the provided context.
+
+#     ## Context
+
+#     {context}
+
+#     ## Question
+
+#     {question}"""
+
+#     return ask_assistant(prompt)
+
+def ask_assistant_in_context(question, context):
+    return "raspuns, bla, bla, bla"
+
+# def create_chat(id = 0, context=""):
+#     chat = st.container()
+
+#     # Initialize chat history
+#     if "messages" not in st.session_state:
+#         st.session_state.messages = []
+
+#     # Display chat messages from history on app rerun
+#     for message in st.session_state.messages:
+#         if message['id'] == id:
+#             chat.chat_message(message['role']).write(message['content'])
+#             # context += message['role'] + ": " + message['content'] + "\n"
+ 
+#     # Accept user input
+#     if prompt := st.chat_input(placeholder = "Don't waist time reading, ask me", key = id):
+
+#         chat.chat_message("user").write(prompt)
+#         with st.spinner('Wait for it...'):
+#             assistant_response = ask_assistant_in_context(prompt, context)
+#         chat.chat_message("assistant").write(f"{assistant_response}")
+
+#         st.session_state.messages.append({"id": id, "role": "user", "content": prompt})
+#         st.session_state.messages.append({"id": id, "role": "assistant", "content": assistant_response})
+
+# url = "https://www.europarl.europa.eu/rss/doc/last-news-committees/en.xml"
+# feed = feedparser.parse(url)
+
+with gr.Accordion("See Details"):
+    gr.Markdown("lorem ipsum")
+
+# st.subheader(feed.feed.title, divider='gray')
+# key = 0
+
+# for entry in feed.entries:
+#     title = entry.title.split('-')[1].strip()
+
+#     col1, col2 = st.columns([0.1,0.9])
+
+#     with col1.popover(""):
+#         content = html2text.html2text(entry.summary)
+#         st.write(content)
+#         st.divider()
+
+#         create_chat(key, content)
+#         key += 1
+    
+#     col2.write(f":gray[{entry.published}]")
+#     col2.write(f"**{title}**")
+
+#     st.divider()
+
+
+# def show_message(button_text):
+#     """Function to return a message based on which button was clicked."""
+#     return f"You clicked: {button_text}"
+
+# def main():
+#     with gr.Blocks() as app:
+#         # Create a container for buttons
+#         with gr.Column():
+#             button_texts = ["Item 1", "Item 2", "Item 3"]
+#             outputs = []
+            
+#             for text in button_texts:
+#                 # Create a button for each item
+#                 btn = gr.Button(value=text)
+#                 output_text = gr.Textbox()
+#                 btn.click(fn=show_message, inputs=[btn], outputs=output_text)
+#                 outputs.append(output_text)
+
+#     app.launch()
+
+# if __name__ == "__main__":
+#     main()
+
+def yes_man(message, history):
+    if message.endswith("?"):
+        return "Yes"
+    else:
+        return "Ask me anything!"
+
+gr.ChatInterface(
+    yes_man,
+    chatbot=gr.Chatbot(height=300),
+    textbox=gr.Textbox(placeholder="Ask me a yes or no question", container=False, scale=7),
+    title="Model",
+    description="Ask Model any question",
+    theme="soft",
+    examples=[""],
+    cache_examples=True,
+    retry_btn=None,
+    undo_btn=None,
+    clear_btn=None,
+).launch()

gradio_cached_examples/13/log.csv

@@ -0,0 +1,2 @@
+component 0,flag,username,timestamp
+"[["""", ""Ask me anything!""]]",,,2024-03-28 01:17:13.202639

gradio_cached_examples/15/log.csv

@@ -0,0 +1,4 @@
+component 0,flag,username,timestamp
+"[[""Hello"", ""Ask me anything!""]]",,,2024-03-28 01:10:33.526848
+"[[""Am I cool?"", ""Yes""]]",,,2024-03-28 01:10:33.530444
+"[[""Are tomatoes vegetables?"", ""Yes""]]",,,2024-03-28 01:10:33.539654

requirements.txt

@@ -0,0 +1,10 @@
+--extra-index-url https://download.pytorch.org/whl/cu113
+transformers
+datasets
+torch
+sentencepiece
+accelerate
+bitsandbytes
+wikipedia-api
+feedparser
+html2text