Update app.py
Browse files
app.py
CHANGED
@@ -1,7 +1,9 @@
|
|
|
|
1 |
import streamlit as st
|
2 |
import uuid
|
3 |
import sys
|
4 |
import requests
|
|
|
5 |
import bitsandbytes as bnb
|
6 |
import pandas as pd
|
7 |
import torch
|
@@ -9,6 +11,12 @@ import torch.nn as nn
|
|
9 |
import transformers
|
10 |
from datasets import load_dataset
|
11 |
from huggingface_hub import notebook_login
|
|
|
|
|
|
|
|
|
|
|
|
|
12 |
from transformers import (
|
13 |
AutoConfig,
|
14 |
AutoModelForCausalLM,
|
@@ -66,13 +74,11 @@ st.markdown("""
|
|
66 |
</style>
|
67 |
""", unsafe_allow_html=True)
|
68 |
|
69 |
-
# Load the model
|
70 |
-
|
71 |
-
model =
|
72 |
-
|
73 |
-
|
74 |
-
# Load the dataset
|
75 |
-
dataset = load_dataset("nisaar/Lawyer_GPT_India")
|
76 |
|
77 |
def write_top_bar():
|
78 |
col1, col2, col3 = st.columns([1,10,2])
|
@@ -105,13 +111,8 @@ def handle_input():
|
|
105 |
if len(chat_history) == MAX_HISTORY_LENGTH:
|
106 |
chat_history = chat_history[:-1]
|
107 |
|
108 |
-
|
109 |
-
|
110 |
-
|
111 |
-
# Generate response using the model
|
112 |
-
inputs = tokenizer.encode(closest_example, return_tensors="pt")
|
113 |
-
outputs = model.generate(inputs)
|
114 |
-
answer = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
115 |
|
116 |
chat_history.append((input, answer))
|
117 |
|
@@ -121,12 +122,6 @@ def handle_input():
|
|
121 |
})
|
122 |
st.session_state.input = ""
|
123 |
|
124 |
-
def find_closest_example(input, dataset):
|
125 |
-
# Implement your own logic to find the closest example in the dataset based on the user input
|
126 |
-
# You can use techniques like cosine similarity, semantic similarity, or any other approach that fits your dataset and requirements
|
127 |
-
# Return the closest example as a string
|
128 |
-
pass
|
129 |
-
|
130 |
def write_user_message(md):
|
131 |
col1, col2 = st.columns([1,12])
|
132 |
|
@@ -153,4 +148,4 @@ with st.container():
|
|
153 |
write_chat_message(a, q)
|
154 |
|
155 |
st.markdown('---')
|
156 |
-
input = st.text_input("You are talking to an AI, ask any question.", key="input", on_change=handle_input)
|
|
|
1 |
+
|
2 |
import streamlit as st
|
3 |
import uuid
|
4 |
import sys
|
5 |
import requests
|
6 |
+
from peft import *
|
7 |
import bitsandbytes as bnb
|
8 |
import pandas as pd
|
9 |
import torch
|
|
|
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,
|
|
|
74 |
</style>
|
75 |
""", unsafe_allow_html=True)
|
76 |
|
77 |
+
# Load the model outside the handle_input() function
|
78 |
+
with open('model_saved.pkl', 'rb') as f:
|
79 |
+
model = pickle.load(f)
|
80 |
+
if not isinstance(model, str):
|
81 |
+
st.error("The loaded model is not valid.")
|
|
|
|
|
82 |
|
83 |
def write_top_bar():
|
84 |
col1, col2, col3 = st.columns([1,10,2])
|
|
|
111 |
if len(chat_history) == MAX_HISTORY_LENGTH:
|
112 |
chat_history = chat_history[:-1]
|
113 |
|
114 |
+
prompt = input
|
115 |
+
answer = model # Replace the predict() method with the model itself
|
|
|
|
|
|
|
|
|
|
|
116 |
|
117 |
chat_history.append((input, answer))
|
118 |
|
|
|
122 |
})
|
123 |
st.session_state.input = ""
|
124 |
|
|
|
|
|
|
|
|
|
|
|
|
|
125 |
def write_user_message(md):
|
126 |
col1, col2 = st.columns([1,12])
|
127 |
|
|
|
148 |
write_chat_message(a, q)
|
149 |
|
150 |
st.markdown('---')
|
151 |
+
input = st.text_input("You are talking to an AI, ask any question.", key="input", on_change=handle_input)
|