Spaces:
Build error
Build error
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,60 +1,390 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
import gradio as gr
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
|
| 3 |
-
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
from
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
{% for doc in documents %}
|
| 25 |
-
{{ doc.content }}
|
| 26 |
-
{% endfor %}
|
| 27 |
-
|
| 28 |
-
Answer the given question: {{question}}
|
| 29 |
-
Answer:
|
| 30 |
-
"""
|
| 31 |
-
prompt_builder = PromptBuilder(template=prompt_template)
|
| 32 |
-
|
| 33 |
-
pipeline = Pipeline()
|
| 34 |
-
pipeline.add_component("fetcher", fetcher)
|
| 35 |
-
pipeline.add_component("converter", converter)
|
| 36 |
-
pipeline.add_component("splitter", document_splitter)
|
| 37 |
-
pipeline.add_component("ranker", similarity_ranker)
|
| 38 |
-
pipeline.add_component("prompt_builder", prompt_builder)
|
| 39 |
-
pipeline.add_component("llm", generator)
|
| 40 |
-
|
| 41 |
-
pipeline.connect("fetcher.streams", "converter.sources")
|
| 42 |
-
pipeline.connect("converter.documents", "splitter.documents")
|
| 43 |
-
pipeline.connect("splitter.documents", "ranker.documents")
|
| 44 |
-
pipeline.connect("ranker.documents", "prompt_builder.documents")
|
| 45 |
-
pipeline.connect("prompt_builder.prompt", "llm.prompt")
|
| 46 |
-
|
| 47 |
-
def respond(prompt, use_rag):
|
| 48 |
-
if use_rag:
|
| 49 |
-
result = pipeline.run({"prompt_builder": {"question": prompt},
|
| 50 |
-
"ranker": {"query": prompt},
|
| 51 |
-
"fetcher": {"urls": ["https://haystack.deepset.ai/blog/introducing-haystack-2-beta-and-advent"]},
|
| 52 |
-
"llm":{"generation_kwargs": {"max_new_tokens": 350}}})
|
| 53 |
-
return result['llm']['replies'][0]
|
| 54 |
else:
|
| 55 |
-
|
| 56 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 57 |
|
| 58 |
-
|
| 59 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 60 |
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#from haystack.components.generators import HuggingFaceTGIGenerator
|
| 2 |
+
from llama_index.llms import HuggingFaceInferenceAPI
|
| 3 |
+
from llama_index.llms import ChatMessage, MessageRole
|
| 4 |
+
from llama_index.prompts import ChatPromptTemplate
|
| 5 |
+
from llama_index import VectorStoreIndex, SimpleDirectoryReader, ServiceContext #, LLMPredictor, StorageContext, load_index_from_storage
|
| 6 |
import gradio as gr
|
| 7 |
+
#import sys
|
| 8 |
+
#import logging
|
| 9 |
+
#import torch
|
| 10 |
+
#from huggingface_hub import InferenceClient
|
| 11 |
+
#import tqdm as notebook_tqdm
|
| 12 |
+
import requests
|
| 13 |
+
import os
|
| 14 |
+
import json
|
| 15 |
|
| 16 |
+
#generator = HuggingFaceTGIGenerator("mistralai/Mixtral-8x7B-Instruct-v0.1")
|
| 17 |
+
#generator.warm_up()
|
| 18 |
+
|
| 19 |
+
def download_file(url, filename):
|
| 20 |
+
"""
|
| 21 |
+
Download a file from the specified URL and save it locally under the given filename.
|
| 22 |
+
"""
|
| 23 |
+
|
| 24 |
+
response = requests.get(url, stream=True)
|
| 25 |
+
|
| 26 |
+
# Check if the request was successful
|
| 27 |
+
|
| 28 |
+
if filename in os.listdir('content/'): return
|
| 29 |
+
if filename == '': return
|
| 30 |
+
|
| 31 |
+
if response.status_code == 200:
|
| 32 |
+
with open('content/' + filename, 'wb') as file:
|
| 33 |
+
for chunk in response.iter_content(chunk_size=1024):
|
| 34 |
+
if chunk: # filter out keep-alive new chunks
|
| 35 |
+
file.write(chunk)
|
| 36 |
+
print(f"Download complete: {filename}")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 37 |
else:
|
| 38 |
+
print(f"Error: Unable to download file. HTTP status code: {response.status_code}")
|
| 39 |
+
|
| 40 |
+
#def save_answer(prompt, rag_answer, norag_answer):
|
| 41 |
+
# json_dict = dict()
|
| 42 |
+
# json_dict['prompt'] = prompt
|
| 43 |
+
# json_dict['rag_answer'] = rag_answer
|
| 44 |
+
# json_dict['norag_answer'] = norag_answer
|
| 45 |
+
#
|
| 46 |
+
# file_path = 'saved_answers.json'
|
| 47 |
+
#
|
| 48 |
+
# # Check if the file exists
|
| 49 |
+
# if not os.path.isfile(file_path):
|
| 50 |
+
# with open(file_path, 'w') as f:
|
| 51 |
+
# # Create an empty list in the file to store dictionaries
|
| 52 |
+
# json.dump([], f)
|
| 53 |
+
# f.write('\n') # Add a newline to separate the list and future entries
|
| 54 |
+
#
|
| 55 |
+
# # Open the file in append mode
|
| 56 |
+
# with open(file_path, 'a+') as f:
|
| 57 |
+
# # Read the existing data
|
| 58 |
+
# f.seek(0)
|
| 59 |
+
# data = json.load(f)
|
| 60 |
+
#
|
| 61 |
+
# # Append the new dictionary to the list
|
| 62 |
+
# data.append(json_dict)
|
| 63 |
+
#
|
| 64 |
+
# # Move the cursor to the beginning of the file
|
| 65 |
+
# f.seek(0)
|
| 66 |
+
#
|
| 67 |
+
# # Write the updated list of dictionaries
|
| 68 |
+
# json.dump(data, f)
|
| 69 |
+
# f.write('\n') # Add a newline to separate the list and future entries
|
| 70 |
+
#
|
| 71 |
+
#
|
| 72 |
+
#def check_answer(prompt):
|
| 73 |
+
# file_path = 'saved_answers.json'
|
| 74 |
+
#
|
| 75 |
+
# if not os.path.isfile(file_path):
|
| 76 |
+
# with open(file_path, 'w') as f:
|
| 77 |
+
# # Create an empty list in the file to store dictionaries
|
| 78 |
+
# json.dump([], f)
|
| 79 |
+
# f.write('\n') # Add a newline to separate the list and future entries
|
| 80 |
+
# with open('saved_answers.json', 'r') as f:
|
| 81 |
+
# data = json.load(f)
|
| 82 |
+
# for entry in data:
|
| 83 |
+
# if entry['prompt'] == prompt:
|
| 84 |
+
# return entry['rag_answer'], entry['norag_answer']
|
| 85 |
+
# return None, None # Return None if the prompt is not found
|
| 86 |
+
|
| 87 |
+
|
| 88 |
+
def save_answer(prompt, rag_answer, norag_answer):
|
| 89 |
+
file_path = 'saved_answers.jsonl'
|
| 90 |
+
|
| 91 |
+
# Create a dictionary for the current answer
|
| 92 |
+
json_dict = {
|
| 93 |
+
'prompt': prompt,
|
| 94 |
+
'rag_answer': rag_answer,
|
| 95 |
+
'norag_answer': norag_answer
|
| 96 |
+
}
|
| 97 |
+
|
| 98 |
+
# Check if the file exists, and create it if not
|
| 99 |
+
#if not os.path.isfile(file_path):
|
| 100 |
+
# with open(file_path, 'w') as f:
|
| 101 |
+
# # Create an empty list in the file to store dictionaries
|
| 102 |
+
# json.dump([], f)
|
| 103 |
+
# f.write('\n') # Add a newline to separate the list and future entries
|
| 104 |
+
|
| 105 |
+
# Load existing data from the file
|
| 106 |
+
existing_data = load_jsonl(file_path)
|
| 107 |
+
|
| 108 |
+
# Append the new answer to the existing data
|
| 109 |
+
existing_data.append(json_dict)
|
| 110 |
+
|
| 111 |
+
# Save the updated data back to the file
|
| 112 |
+
write_to_jsonl(file_path, existing_data)
|
| 113 |
+
|
| 114 |
+
def check_answer(prompt):
|
| 115 |
+
file_path = 'saved_answers.jsonl'
|
| 116 |
+
|
| 117 |
+
## Check if the file exists, and create it if not
|
| 118 |
+
#if not os.path.isfile(file_path):
|
| 119 |
+
# with open(file_path, 'w') as f:
|
| 120 |
+
# # Create an empty list in the file to store dictionaries
|
| 121 |
+
# json.dump([], f)
|
| 122 |
+
# f.write('\n') # Add a newline to separate the list and future entries
|
| 123 |
+
|
| 124 |
+
# Load existing data from the file
|
| 125 |
+
|
| 126 |
+
try:
|
| 127 |
+
existing_data = load_jsonl(file_path)
|
| 128 |
+
|
| 129 |
+
except:
|
| 130 |
+
return None, None
|
| 131 |
+
|
| 132 |
+
if len(existing_data) == 0:
|
| 133 |
+
return None, None
|
| 134 |
+
|
| 135 |
+
# Find the answer for the given prompt, if it exists
|
| 136 |
+
for entry in existing_data:
|
| 137 |
+
if entry['prompt'] == prompt:
|
| 138 |
+
return entry['rag_answer'], entry['norag_answer']
|
| 139 |
+
|
| 140 |
+
# Return None if the prompt is not found
|
| 141 |
+
return None, None
|
| 142 |
+
|
| 143 |
+
# Helper functions
|
| 144 |
+
def load_jsonl(file_path):
|
| 145 |
+
data = []
|
| 146 |
+
with open(file_path, 'r') as file:
|
| 147 |
+
for line in file:
|
| 148 |
+
# Each line is a JSON object
|
| 149 |
+
item = json.loads(line)
|
| 150 |
+
data.append(item)
|
| 151 |
+
return data
|
| 152 |
+
|
| 153 |
+
def write_to_jsonl(file_path, data):
|
| 154 |
+
with open(file_path, 'a+') as file:
|
| 155 |
+
for item in data:
|
| 156 |
+
# Convert Python object to JSON string and write it to the file
|
| 157 |
+
json_line = json.dumps(item)
|
| 158 |
+
file.write(json_line + '\n')
|
| 159 |
+
|
| 160 |
+
|
| 161 |
+
|
| 162 |
+
def generate(prompt, history, rag_only, file_link, temperature=0.9, max_new_tokens=256, top_p=0.95, repetition_penalty=1.0,):
|
| 163 |
+
|
| 164 |
+
rag_answer, norag_answer = check_answer(prompt)
|
| 165 |
+
|
| 166 |
+
if rag_answer != None:
|
| 167 |
+
if rag_only:
|
| 168 |
+
return f'* Mixtral + RAG Output:\n{rag_answer}'
|
| 169 |
+
else:
|
| 170 |
+
return f'* Mixtral Output:\n{norag_answer}\n\n* Mixtral + RAG Output:\n{rag_answer}'
|
| 171 |
+
|
| 172 |
+
mixtral = HuggingFaceInferenceAPI(
|
| 173 |
+
model_name="mistralai/Mixtral-8x7B-Instruct-v0.1"
|
| 174 |
+
#Mistral-7B-Instruct-v0.2
|
| 175 |
+
)
|
| 176 |
+
|
| 177 |
+
service_context = ServiceContext.from_defaults(
|
| 178 |
+
llm=mixtral, embed_model="local:BAAI/bge-small-en-v1.5"
|
| 179 |
+
)
|
| 180 |
+
|
| 181 |
+
download = download_file(file_link,file_link.split("/")[-1])
|
| 182 |
+
|
| 183 |
+
documents = SimpleDirectoryReader("content/").load_data()
|
| 184 |
+
|
| 185 |
+
index = VectorStoreIndex.from_documents(documents,service_context=service_context)
|
| 186 |
+
|
| 187 |
+
# Text QA Prompt
|
| 188 |
+
chat_text_qa_msgs = [
|
| 189 |
+
ChatMessage(
|
| 190 |
+
role=MessageRole.SYSTEM,
|
| 191 |
+
content=(
|
| 192 |
+
"Always answer the question, even if the context isn't helpful."
|
| 193 |
+
),
|
| 194 |
+
),
|
| 195 |
+
ChatMessage(
|
| 196 |
+
role=MessageRole.USER,
|
| 197 |
+
content=(
|
| 198 |
+
"Context information is below.\n"
|
| 199 |
+
"---------------------\n"
|
| 200 |
+
"{context_str}\n"
|
| 201 |
+
"---------------------\n"
|
| 202 |
+
"Given the context information and not prior knowledge, "
|
| 203 |
+
"answer the question: {query_str}\n"
|
| 204 |
+
),
|
| 205 |
+
),
|
| 206 |
+
]
|
| 207 |
+
text_qa_template = ChatPromptTemplate(chat_text_qa_msgs)
|
| 208 |
+
|
| 209 |
+
# Refine Prompt
|
| 210 |
+
chat_refine_msgs = [
|
| 211 |
+
ChatMessage(
|
| 212 |
+
role=MessageRole.SYSTEM,
|
| 213 |
+
content=(
|
| 214 |
+
"Always answer the question, even if the context isn't helpful."
|
| 215 |
+
),
|
| 216 |
+
),
|
| 217 |
+
ChatMessage(
|
| 218 |
+
role=MessageRole.USER,
|
| 219 |
+
content=(
|
| 220 |
+
"We have the opportunity to refine the original answer "
|
| 221 |
+
"(only if needed) with some more context below.\n"
|
| 222 |
+
"------------\n"
|
| 223 |
+
"{context_msg}\n"
|
| 224 |
+
"------------\n"
|
| 225 |
+
"Given the new context, refine the original answer to better "
|
| 226 |
+
"answer the question: {query_str}. "
|
| 227 |
+
"If the context isn't useful, output the original answer again.\n"
|
| 228 |
+
"Original Answer: {existing_answer}"
|
| 229 |
+
),
|
| 230 |
+
),
|
| 231 |
+
]
|
| 232 |
+
refine_template = ChatPromptTemplate(chat_refine_msgs)
|
| 233 |
+
|
| 234 |
+
temperature = float(temperature)
|
| 235 |
+
if temperature < 1e-2:
|
| 236 |
+
temperature = 1e-2
|
| 237 |
+
top_p = float(top_p)
|
| 238 |
+
|
| 239 |
+
stream= index.as_query_engine(
|
| 240 |
+
text_qa_template=text_qa_template, refine_template=refine_template, similarity_top_k=6, temperature = temperature,
|
| 241 |
+
max_new_tokens=max_new_tokens, top_p=top_p, repetition_penalty = repetition_penalty
|
| 242 |
+
).query(prompt)
|
| 243 |
+
print(str(stream))
|
| 244 |
+
|
| 245 |
+
output_rag= str(stream) #""
|
| 246 |
+
|
| 247 |
+
#output_norag = mixtral.complete(prompt, details=True, similarity_top_k=6, temperature = temperature,
|
| 248 |
+
# max_new_tokens=max_new_tokens, top_p=top_p, repetition_penalty = repetition_penalty)
|
| 249 |
+
|
| 250 |
+
#for response in str(stream):
|
| 251 |
+
# output += response
|
| 252 |
+
# yield output
|
| 253 |
+
|
| 254 |
+
#print(output_norag)
|
| 255 |
+
|
| 256 |
+
|
| 257 |
+
#result = generator.run(prompt, generation_kwargs={"max_new_tokens": 350})
|
| 258 |
+
#output_norag = result["replies"][0]
|
| 259 |
+
|
| 260 |
+
|
| 261 |
+
### NORAG
|
| 262 |
+
|
| 263 |
+
if rag_only == False:
|
| 264 |
+
chat_text_qa_msgs_nr = [
|
| 265 |
+
ChatMessage(
|
| 266 |
+
role=MessageRole.SYSTEM,
|
| 267 |
+
content=(
|
| 268 |
+
"Always answer the question"
|
| 269 |
+
),
|
| 270 |
+
),
|
| 271 |
+
ChatMessage(
|
| 272 |
+
role=MessageRole.USER,
|
| 273 |
+
content=(
|
| 274 |
+
"answer the question: {query_str}\n"
|
| 275 |
+
),
|
| 276 |
+
),
|
| 277 |
+
]
|
| 278 |
+
text_qa_template_nr = ChatPromptTemplate(chat_text_qa_msgs_nr)
|
| 279 |
+
|
| 280 |
+
# Refine Prompt
|
| 281 |
+
chat_refine_msgs_nr = [
|
| 282 |
+
ChatMessage(
|
| 283 |
+
role=MessageRole.SYSTEM,
|
| 284 |
+
content=(
|
| 285 |
+
"Always answer the question"
|
| 286 |
+
),
|
| 287 |
+
),
|
| 288 |
+
ChatMessage(
|
| 289 |
+
role=MessageRole.USER,
|
| 290 |
+
content=(
|
| 291 |
+
"answer the question: {query_str}. "
|
| 292 |
+
"If the context isn't useful, output the original answer again.\n"
|
| 293 |
+
"Original Answer: {existing_answer}"
|
| 294 |
+
),
|
| 295 |
+
),
|
| 296 |
+
]
|
| 297 |
+
refine_template_nr = ChatPromptTemplate(chat_refine_msgs_nr)
|
| 298 |
+
|
| 299 |
+
stream_nr= index.as_query_engine(
|
| 300 |
+
text_qa_template=text_qa_template_nr, refine_template=refine_template_nr, similarity_top_k=6
|
| 301 |
+
).query(prompt)
|
| 302 |
+
|
| 303 |
+
###
|
| 304 |
+
|
| 305 |
+
output_norag = str(stream_nr)
|
| 306 |
+
save_answer(prompt, output_rag, output_norag)
|
| 307 |
+
|
| 308 |
+
return f'* Mixtral Output:\n{output_norag}\n\n* Mixtral + RAG Output:\n{output_rag}'
|
| 309 |
+
|
| 310 |
+
return f'* Mixtral + RAG Output:\n{output_rag}'
|
| 311 |
+
|
| 312 |
+
#for response in formatted_output:
|
| 313 |
+
# output += response
|
| 314 |
+
# yield output
|
| 315 |
+
#return formatted_output
|
| 316 |
+
|
| 317 |
+
def upload_file(files):
|
| 318 |
+
file_paths = [file.name for file in files]
|
| 319 |
+
return file_paths
|
| 320 |
+
|
| 321 |
+
additional_inputs=[
|
| 322 |
+
gr.Checkbox(
|
| 323 |
+
label="RAG Only",
|
| 324 |
+
interactive=True,
|
| 325 |
+
value= False
|
| 326 |
+
),
|
| 327 |
+
gr.Textbox(
|
| 328 |
+
label="File Link",
|
| 329 |
+
max_lines=1,
|
| 330 |
+
interactive=True,
|
| 331 |
+
value= "https://arxiv.org/pdf/2401.10020.pdf"
|
| 332 |
+
),
|
| 333 |
+
gr.Slider(
|
| 334 |
+
label="Temperature",
|
| 335 |
+
value=0.9,
|
| 336 |
+
minimum=0.0,
|
| 337 |
+
maximum=1.0,
|
| 338 |
+
step=0.05,
|
| 339 |
+
interactive=True,
|
| 340 |
+
info="Higher values produce more diverse outputs",
|
| 341 |
+
),
|
| 342 |
+
gr.Slider(
|
| 343 |
+
label="Max new tokens",
|
| 344 |
+
value=1024,
|
| 345 |
+
minimum=0,
|
| 346 |
+
maximum=2048,
|
| 347 |
+
step=64,
|
| 348 |
+
interactive=True,
|
| 349 |
+
info="The maximum numbers of new tokens",
|
| 350 |
+
),
|
| 351 |
+
gr.Slider(
|
| 352 |
+
label="Top-p (nucleus sampling)",
|
| 353 |
+
value=0.90,
|
| 354 |
+
minimum=0.0,
|
| 355 |
+
maximum=1,
|
| 356 |
+
step=0.05,
|
| 357 |
+
interactive=True,
|
| 358 |
+
info="Higher values sample more low-probability tokens",
|
| 359 |
+
),
|
| 360 |
+
gr.Slider(
|
| 361 |
+
label="Repetition penalty",
|
| 362 |
+
value=1.2,
|
| 363 |
+
minimum=1.0,
|
| 364 |
+
maximum=2.0,
|
| 365 |
+
step=0.05,
|
| 366 |
+
interactive=True,
|
| 367 |
+
info="Penalize repeated tokens",
|
| 368 |
+
)
|
| 369 |
+
]
|
| 370 |
+
|
| 371 |
+
examples=[["What is a trustworthy digital repository, where can you find this information?", None, None, None, None, None, None, ],
|
| 372 |
+
["What are things a repository must have?", None, None, None, None, None, None,],
|
| 373 |
+
["What principles should record creators follow?", None, None, None, None, None, None,],
|
| 374 |
+
["Write a very short summary of Data Sanitation Techniques by Edgar Dale, and write a citation in APA style.", None, None, None, None, None, None,],
|
| 375 |
+
["Can you explain how the QuickSort algorithm works and provide a Python implementation?", None, None, None, None, None, None,],
|
| 376 |
+
["What are some unique features of Rust that make it stand out compared to other systems programming languages like C++?", None, None, None, None, None, None,],
|
| 377 |
+
]
|
| 378 |
|
| 379 |
+
gr.ChatInterface(
|
| 380 |
+
fn=generate,
|
| 381 |
+
chatbot=gr.Chatbot(show_label=False, show_share_button=False, show_copy_button=True, likeable=True, layout="panel"),
|
| 382 |
+
additional_inputs=additional_inputs,
|
| 383 |
+
title="RAG Demo",
|
| 384 |
+
examples=examples,
|
| 385 |
+
#concurrency_limit=20,
|
| 386 |
+
).queue().launch(show_api=False,debug=True,share=True)
|
| 387 |
|
| 388 |
+
#iface = gr.Interface(fn=generate, inputs=["text"], outputs=["text", "text"],
|
| 389 |
+
# additional_inputs=additional_inputs, title="RAG Demo", examples=examples)
|
| 390 |
+
#iface.launch()
|