Spaces:
Running
Running
File size: 6,699 Bytes
e7fe8f1 0cd089d e7fe8f1 957d4a5 e7fe8f1 957d4a5 e7fe8f1 957d4a5 e7fe8f1 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 |
import gradio as gr
from langchain.chains import RetrievalQA
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.llms import HuggingFacePipeline
from langchain.memory import ConversationBufferWindowMemory
from langchain.vectorstores import Chroma
from langchain import PromptTemplate, LLMChain
from transformers import AutoTokenizer, pipeline
from typing import Dict, Any
import torch
# class AnswerConversationBufferMemory(ConversationBufferMemory):
class AnswerConversationBufferMemory(ConversationBufferWindowMemory):
def save_context(self, inputs: Dict[str, Any], outputs: Dict[str, str]) -> None:
return super(AnswerConversationBufferMemory, self).save_context(inputs,{'response': outputs['result']})
def clean_text(text):
# Remove excessive whitespace
cleaned_text = ' '.join(text.split())
# Keep max one newline character
cleaned_text = cleaned_text.replace('\n\n', '\n')
return cleaned_text
def chatbot_llm_response(llm_response):
text = clean_text(llm_response['result']) + '\nSources:\n'
for source in llm_response["source_documents"]:
text += source.metadata['source'] + '\n'
return text
model_name = "databricks/dolly-v2-3b"
tokenizer = AutoTokenizer.from_pretrained(model_name, padding_side="left")
generate_text = pipeline(model=model_name,
torch_dtype=torch.bfloat16,
trust_remote_code=True,
device_map="auto",
return_full_text=True,
max_new_tokens=256,
top_p=0.95,
top_k=50)
prompt = PromptTemplate(
input_variables=["instruction"],
template="{instruction}")
hf_pipeline = HuggingFacePipeline(pipeline=generate_text)
llm_chain = LLMChain(llm=hf_pipeline, prompt=prompt)
# top #2 when task = Retrieval June 2023 for under ~500 MB
model_name = "intfloat/e5-base-v2"
hf = HuggingFaceEmbeddings(model_name=model_name)
# Load up Vector Database
persist_directory = 'db'
vectordb = Chroma(persist_directory=persist_directory,
embedding_function=hf)
vectordb.get()
retriever = vectordb.as_retriever(search_kwargs={'k':3})
# Configure Conversation Chain
memory = AnswerConversationBufferMemory(k=3)
qa_chain_with_memory = RetrievalQA.from_chain_type(llm=hf_pipeline,
chain_type="stuff",
retriever=retriever,
return_source_documents=True,
memory=memory)
# try to set the tone
template = '''
You are the assistant to a tradesperson with knowledge of the Ontario Building Code. You provide specific details using the context given and the users question.
If you don't know the answer, you truthfully say you don't know and don't try to make up an answer.
----------------
{context}
Question: {question}
Helpful Answer:'''
qa_chain_with_memory.combine_documents_chain.llm_chain.prompt.template = template
examples = ["What's the minimum pipe size needed for sinks, toilets, and showers?",
"Are there any specific rules for installing backflow prevention devices?",
"Can you guide me on the approved materials and methods for installing underground sewer lines?",
"How much clearance is required for electrical panels and equipment like switchboards?",
"Are there any restrictions or guidelines for outdoor electrical wiring and fixtures?",
"Could you explain the proper bonding and grounding requirements for commercial buildings?",
"What's the load-bearing capacity for beams and columns?",
"Are there any specific rules for designing buildings to withstand earthquakes?",
"Can you provide information on the fire resistance ratings for walls, floors, and roofs?",
"What are the specific building code requirements for designing accessible entrances and pathways?",
"Can you explain the regulations for fire protection systems and how they should be integrated into architectural designs?",
"What are the foundation requirements in areas prone to earthquakes?",
"Are there any restrictions or guidelines for installing electrical wiring and fixtures in wet locations?"
]
def process_example(args):
for x in generate(args):
pass
return x
def generate(instruction):
response = qa_chain_with_memory(instruction)
processed_response = chatbot_llm_response(response)
result = ""
for word in processed_response.split(" "):
result += word + " "
yield result
with gr.Blocks(analytics_enabled=False) as demo:
with gr.Column():
gr.Markdown("""# 🐑 Dolly-Expert-Lite
Dolly-Expert-Lite is a bot for domain specific question
answering. Currently powered by the new Dolly-v2-3b open
source model. It's expert systems in the era of LLMs!
## 🏗️ Building Code Expert
In this example deployment, Dolly-Expert-Lite retrieves
information via a vector database made using the
[Ontario (Canada) Building Code](https://www.buildingcode.online)
sitemap LangChain loader. For details on the original Dolly
v2 model, please refer to the
[model card](https://huggingface.co/databricks/dolly-v2-12b)
### Type in the box below and click to ask the expert!
"""
)
with gr.Row():
with gr.Column(scale=3):
instruction = gr.Textbox(placeholder="Enter your question here", label="Question", elem_id="q-input")
with gr.Box():
gr.Markdown("**Answer**")
output = gr.Markdown(elem_id="q-output")
submit = gr.Button("Generate", variant="primary")
clear = gr.Button("Clear", variant="secondary")
gr.Examples(
examples=examples,
inputs=[instruction],
cache_examples=False,
fn=process_example,
outputs=[output],
)
submit.click(generate, inputs=[instruction], outputs=[output])
clear.click(lambda: None, [], [output])
instruction.submit(generate, inputs=[instruction], outputs=[output])
demo.queue(concurrency_count=16).launch()
demo.launch() |