Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,258 +1,72 @@
|
|
| 1 |
-
import
|
| 2 |
-
import
|
| 3 |
-
import sys
|
| 4 |
-
import requests
|
| 5 |
-
import pandas as pd
|
| 6 |
-
from peft import *
|
| 7 |
-
import bitsandbytes as bnb
|
| 8 |
-
import pandas as pd
|
| 9 |
import torch
|
| 10 |
-
import torch.nn as nn
|
| 11 |
-
import transformers
|
| 12 |
-
from datasets import load_dataset
|
| 13 |
-
from huggingface_hub import notebook_login
|
| 14 |
-
from peft import (
|
| 15 |
-
LoraConfig,
|
| 16 |
-
PeftConfig,
|
| 17 |
-
get_peft_model,
|
| 18 |
-
prepare_model_for_kbit_training,
|
| 19 |
-
)
|
| 20 |
-
from transformers import (
|
| 21 |
-
AutoConfig,
|
| 22 |
-
AutoModelForCausalLM,
|
| 23 |
-
AutoTokenizer,
|
| 24 |
-
BitsAndBytesConfig,
|
| 25 |
-
)
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
USER_ICON = "images/user-icon.png"
|
| 29 |
-
AI_ICON = "images/ai-icon.png"
|
| 30 |
-
MAX_HISTORY_LENGTH = 5
|
| 31 |
-
|
| 32 |
-
if 'user_id' in st.session_state:
|
| 33 |
-
user_id = st.session_state['user_id']
|
| 34 |
-
else:
|
| 35 |
-
user_id = str(uuid.uuid4())
|
| 36 |
-
st.session_state['user_id'] = user_id
|
| 37 |
-
|
| 38 |
-
if 'chat_history' not in st.session_state:
|
| 39 |
-
st.session_state['chat_history'] = []
|
| 40 |
-
|
| 41 |
-
if "chats" not in st.session_state:
|
| 42 |
-
st.session_state.chats = [
|
| 43 |
-
{
|
| 44 |
-
'id': 0,
|
| 45 |
-
'question': '',
|
| 46 |
-
'answer': ''
|
| 47 |
-
}
|
| 48 |
-
]
|
| 49 |
-
|
| 50 |
-
if "questions" not in st.session_state:
|
| 51 |
-
st.session_state.questions = []
|
| 52 |
-
|
| 53 |
-
if "answers" not in st.session_state:
|
| 54 |
-
st.session_state.answers = []
|
| 55 |
-
|
| 56 |
-
if "input" not in st.session_state:
|
| 57 |
-
st.session_state.input = ""
|
| 58 |
-
|
| 59 |
-
st.markdown("""
|
| 60 |
-
<style>
|
| 61 |
-
.block-container {
|
| 62 |
-
padding-top: 32px;
|
| 63 |
-
padding-bottom: 32px;
|
| 64 |
-
padding-left: 0;
|
| 65 |
-
padding-right: 0;
|
| 66 |
-
}
|
| 67 |
-
.element-container img {
|
| 68 |
-
background-color: #000000;
|
| 69 |
-
}
|
| 70 |
-
|
| 71 |
-
.main-header {
|
| 72 |
-
font-size: 24px;
|
| 73 |
-
}
|
| 74 |
-
</style>
|
| 75 |
-
""", unsafe_allow_html=True)
|
| 76 |
-
|
| 77 |
-
def write_top_bar():
|
| 78 |
-
col1, col2, col3 = st.columns([1,10,2])
|
| 79 |
-
with col1:
|
| 80 |
-
st.image(AI_ICON, use_column_width='always')
|
| 81 |
-
with col2:
|
| 82 |
-
header = "Cogwise Intelligent Assistant"
|
| 83 |
-
st.write(f"<h3 class='main-header'>{header}</h3>", unsafe_allow_html=True)
|
| 84 |
-
with col3:
|
| 85 |
-
clear = st.button("Clear Chat")
|
| 86 |
-
return clear
|
| 87 |
-
|
| 88 |
-
clear = write_top_bar()
|
| 89 |
-
|
| 90 |
-
if clear:
|
| 91 |
-
st.session_state.questions = []
|
| 92 |
-
st.session_state.answers = []
|
| 93 |
-
st.session_state.input = ""
|
| 94 |
-
st.session_state["chat_history"] = []
|
| 95 |
-
|
| 96 |
-
def handle_input():
|
| 97 |
-
input = st.session_state.input
|
| 98 |
-
question_with_id = {
|
| 99 |
-
'question': input,
|
| 100 |
-
'id': len(st.session_state.questions)
|
| 101 |
-
}
|
| 102 |
-
st.session_state.questions.append(question_with_id)
|
| 103 |
|
| 104 |
-
|
| 105 |
-
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
# answer = result['answer']
|
| 114 |
-
# !pip install -Uqqq pip --progress-bar off
|
| 115 |
-
# !pip install -qqq bitsandbytes == 0.39.0
|
| 116 |
-
# !pip install -qqqtorch --2.0.1 --progress-bar off
|
| 117 |
-
# !pip install -qqq -U git + https://github.com/huggingface/transformers.git@e03a9cc --progress-bar off
|
| 118 |
-
# !pip install -qqq -U git + https://github.com/huggingface/peft.git@42a184f --progress-bar off
|
| 119 |
-
# !pip install -qqq -U git + https://github.com/huggingface/accelerate.git@c9fbb71 --progress-bar off
|
| 120 |
-
# !pip install -qqq datasets == 2.12.0 --progress-bar off
|
| 121 |
-
# !pip install -qqq loralib == 0.1.1 --progress-bar off
|
| 122 |
-
# !pip install einops
|
| 123 |
-
|
| 124 |
-
import os
|
| 125 |
-
# from pprint import pprint
|
| 126 |
-
# import json
|
| 127 |
-
|
| 128 |
-
|
| 129 |
-
|
| 130 |
-
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
|
| 131 |
-
|
| 132 |
-
# notebook_login()
|
| 133 |
-
# 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
|
| 134 |
-
|
| 135 |
-
# """### Load dataset"""
|
| 136 |
-
|
| 137 |
-
from datasets import load_dataset
|
| 138 |
-
|
| 139 |
-
dataset_name = "nisaar/Lawyer_GPT_India"
|
| 140 |
-
# dataset_name = "patrick11434/TEST_LLM_DATASET"
|
| 141 |
-
dataset = load_dataset(dataset_name, split="train")
|
| 142 |
-
|
| 143 |
-
# """## Load adapters from the Hub
|
| 144 |
-
|
| 145 |
-
# You can also directly load adapters from the Hub using the commands below:
|
| 146 |
-
# """
|
| 147 |
|
| 148 |
|
| 149 |
-
|
| 150 |
-
|
| 151 |
-
|
| 152 |
-
load_4bit_use_double_quant=True,
|
| 153 |
-
bnb_4bit_quant_type="nf4",
|
| 154 |
-
bnb_4bit_compute_dtype=torch.bfloat16,
|
| 155 |
-
)
|
| 156 |
|
| 157 |
-
|
| 158 |
-
|
| 159 |
-
|
| 160 |
-
|
| 161 |
-
|
| 162 |
-
|
| 163 |
-
|
| 164 |
-
|
| 165 |
)
|
| 166 |
-
tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)
|
| 167 |
-
tokenizer.pad_token = tokenizer.eos_token
|
| 168 |
-
|
| 169 |
-
model = PeftModel.from_pretrained(model, peft_model_id)
|
| 170 |
|
| 171 |
-
|
|
|
|
| 172 |
|
| 173 |
-
|
| 174 |
-
""
|
|
|
|
| 175 |
|
| 176 |
-
generation_config = model.generation_config
|
| 177 |
-
generation_config.max_new_tokens = 200
|
| 178 |
-
generation_config_temperature = 1
|
| 179 |
-
generation_config.top_p = 0.7
|
| 180 |
-
generation_config.num_return_sequences = 1
|
| 181 |
-
generation_config.pad_token_id = tokenizer.eos_token_id
|
| 182 |
-
generation_config_eod_token_id = tokenizer.eos_token_id
|
| 183 |
|
| 184 |
-
DEVICE = "cuda:0"
|
| 185 |
|
| 186 |
-
# Commented out IPython magic to ensure Python compatibility.
|
| 187 |
-
# %%time
|
| 188 |
-
# prompt = f"""
|
| 189 |
-
# <human>: Who appoints the Chief Justice of India?
|
| 190 |
-
# <assistant>:
|
| 191 |
-
# """.strip()
|
| 192 |
-
#
|
| 193 |
-
# encoding = tokenizer(prompt, return_tensors="pt").to(DEVICE)
|
| 194 |
-
# with torch.inference_mode():
|
| 195 |
-
# outputs = model.generate(
|
| 196 |
-
# input_ids=encoding.attention_mask,
|
| 197 |
-
# generation_config=generation_config,
|
| 198 |
-
# )
|
| 199 |
-
# print(tokenizer.decode(outputs[0],skip_special_tokens=True))
|
| 200 |
|
| 201 |
-
def generate_response(question: str) -> str:
|
| 202 |
-
prompt = f"""
|
| 203 |
-
<human>: {question}
|
| 204 |
-
<assistant>:
|
| 205 |
-
""".strip()
|
| 206 |
-
encoding = tokenizer(prompt, return_tensors="pt").to(DEVICE)
|
| 207 |
-
with torch.inference_mode():
|
| 208 |
-
outputs = model.generate(
|
| 209 |
-
input_ids=encoding.input_ids,
|
| 210 |
-
attention_mask=encoding.attention_mask,
|
| 211 |
-
generation_config=generation_config,
|
| 212 |
-
)
|
| 213 |
-
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
| 214 |
|
| 215 |
-
assistant_start = '<assistant>:'
|
| 216 |
-
response_start = response.find(assistant_start)
|
| 217 |
-
return response[response_start + len(assistant_start):].strip()
|
| 218 |
|
| 219 |
-
# prompt = "Debate the merits and demerits of introducing simultaneous elections in India?"
|
| 220 |
-
prompt=input
|
| 221 |
-
answer=print(generate_response(prompt))
|
| 222 |
|
| 223 |
-
# answer='Yes'
|
| 224 |
-
chat_history.append((input, answer))
|
| 225 |
|
| 226 |
-
st.session_state.answers.append({
|
| 227 |
-
'answer': answer,
|
| 228 |
-
'id': len(st.session_state.questions)
|
| 229 |
-
})
|
| 230 |
-
st.session_state.input = ""
|
| 231 |
|
| 232 |
-
def write_user_message(md):
|
| 233 |
-
col1, col2 = st.columns([1,12])
|
| 234 |
|
| 235 |
-
|
| 236 |
-
st.image(USER_ICON, use_column_width='always')
|
| 237 |
-
with col2:
|
| 238 |
-
st.warning(md['question'])
|
| 239 |
|
| 240 |
-
def render_answer(answer):
|
| 241 |
-
col1, col2 = st.columns([1,12])
|
| 242 |
-
with col1:
|
| 243 |
-
st.image(AI_ICON, use_column_width='always')
|
| 244 |
-
with col2:
|
| 245 |
-
st.info(answer)
|
| 246 |
|
| 247 |
-
|
| 248 |
-
|
| 249 |
-
|
| 250 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 251 |
|
| 252 |
-
|
| 253 |
-
for (q, a) in zip(st.session_state.questions, st.session_state.answers):
|
| 254 |
-
write_user_message(q)
|
| 255 |
-
write_chat_message(a, q)
|
| 256 |
|
| 257 |
-
|
| 258 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
import torch
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 4 |
|
| 5 |
+
model = AutoModelForCausalLM.from_pretrained(
|
| 6 |
+
"CogwiseAI/testchatexample",
|
| 7 |
+
torch_dtype=torch.bfloat16,
|
| 8 |
+
trust_remote_code=True,
|
| 9 |
+
device_map="auto",
|
| 10 |
+
low_cpu_mem_usage=True,
|
| 11 |
+
)
|
| 12 |
+
tokenizer = AutoTokenizer.from_pretrained("CogwiseAI/testchatexample")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
|
| 14 |
|
| 15 |
+
def generate_text(input_text):
|
| 16 |
+
input_ids = tokenizer.encode(input_text, return_tensors="pt")
|
| 17 |
+
attention_mask = torch.ones(input_ids.shape)
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
|
| 19 |
+
output = model.generate(
|
| 20 |
+
input_ids,
|
| 21 |
+
attention_mask=attention_mask,
|
| 22 |
+
max_length=200,
|
| 23 |
+
do_sample=True,
|
| 24 |
+
top_k=10,
|
| 25 |
+
num_return_sequences=1,
|
| 26 |
+
eos_token_id=tokenizer.eos_token_id,
|
| 27 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
| 28 |
|
| 29 |
+
output_text = tokenizer.decode(output[0], skip_special_tokens=True)
|
| 30 |
+
print(output_text)
|
| 31 |
|
| 32 |
+
# Remove Prompt Echo from Generated Text
|
| 33 |
+
cleaned_output_text = output_text.replace(input_text, "")
|
| 34 |
+
return cleaned_output_text
|
| 35 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
|
|
|
|
| 37 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 38 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
|
|
|
|
|
|
|
|
|
|
| 40 |
|
|
|
|
|
|
|
|
|
|
| 41 |
|
|
|
|
|
|
|
| 42 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 43 |
|
|
|
|
|
|
|
| 44 |
|
| 45 |
+
block = gr.Blocks()
|
|
|
|
|
|
|
|
|
|
| 46 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 47 |
|
| 48 |
+
with block:
|
| 49 |
+
gr.Markdown("""<h1><center>Cogwise AI Falcon-7B Instruct</center></h1>
|
| 50 |
+
""")
|
| 51 |
+
chatbot = gr.Chatbot()
|
| 52 |
+
message = gr.Textbox(placeholder=prompt)
|
| 53 |
+
state = gr.State()
|
| 54 |
+
submit = gr.Button("SEND")
|
| 55 |
+
submit.click(generate_text, inputs=[message, state], outputs=[chatbot, state])
|
| 56 |
|
| 57 |
+
block.launch(debug = True)
|
|
|
|
|
|
|
|
|
|
| 58 |
|
| 59 |
+
# logo = (
|
| 60 |
+
# "<div >"
|
| 61 |
+
# "<img src='ai-icon.png'alt='image One'>"
|
| 62 |
+
# + "</div>"
|
| 63 |
+
# )
|
| 64 |
+
# text_generation_interface = gr.Interface(
|
| 65 |
+
# fn=generate_text,
|
| 66 |
+
# inputs=[
|
| 67 |
+
# gr.inputs.Textbox(label="Input Text"),
|
| 68 |
+
# ],
|
| 69 |
+
# outputs=gr.inputs.Textbox(label="Generated Text"),
|
| 70 |
+
# title="Falcon-7B Instruct",
|
| 71 |
+
# image=logo
|
| 72 |
+
# ).launch()
|