switch to a text generation model
Browse files- app.py +85 -13
- requirements.txt +2 -1
app.py
CHANGED
@@ -1,22 +1,94 @@
|
|
1 |
import gradio as gr
|
2 |
-
|
|
|
|
|
|
|
|
|
3 |
|
4 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
|
6 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
7 |
A resume question-answering interface where a recruter can ask the user about their achievements and skills without the need to interact with them directly or the need to read a really long resume
|
8 |
|
9 |
-
**LIMITATIONS:** the bot can only extract specific information and does not take into account multiple sentences at once.
|
10 |
"""
|
11 |
-
examples = [
|
12 |
-
|
13 |
-
|
14 |
-
|
15 |
-
|
|
|
|
|
|
|
16 |
|
17 |
-
def chat(question,history):
|
18 |
-
"""chat with the QA pipeline"""
|
19 |
-
return qa_model(question = question, context = context)["answer"].strip()
|
20 |
|
21 |
-
demo = gr.ChatInterface(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
demo.launch()
|
|
|
1 |
import gradio as gr
|
2 |
+
import torch
|
3 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
|
4 |
+
import os
|
5 |
+
from threading import Thread
|
6 |
+
import spaces
|
7 |
|
8 |
+
token = os.environ["HF_TOKEN"]
|
9 |
+
model = AutoModelForCausalLM.from_pretrained(
|
10 |
+
"google/gemma-7b-it", torch_dtype=torch.float16, token=token
|
11 |
+
)
|
12 |
+
tok = AutoTokenizer.from_pretrained("google/gemma-7b-it", token=token)
|
13 |
+
device = torch.device("cuda")
|
14 |
+
model = model.to(device)
|
15 |
|
16 |
+
with open("context.txt", "r") as f:
|
17 |
+
# read content from the resume
|
18 |
+
context = f.read()
|
19 |
+
|
20 |
+
|
21 |
+
def format_prompt(message):
|
22 |
+
prompt = f"""your name is hafedh hichri and given the following prompt:
|
23 |
+
{message}
|
24 |
+
you will reply directly without any extra info to the previous prompt given the following context:
|
25 |
+
{context}"""
|
26 |
+
return prompt
|
27 |
+
|
28 |
+
|
29 |
+
@spaces.GPU
|
30 |
+
def chat(message, history):
|
31 |
+
chat = []
|
32 |
+
for item in history:
|
33 |
+
chat.append({"role": "user", "content": item[0]})
|
34 |
+
if item[1] is not None:
|
35 |
+
chat.append({"role": "assistant", "content": item[1]})
|
36 |
+
chat.append({"role": "user", "content": format_prompt(message)})
|
37 |
+
messages = tok.apply_chat_template(chat, tokenize=False, add_generation_prompt=True)
|
38 |
+
# Tokenize the messages string
|
39 |
+
model_inputs = tok([messages], return_tensors="pt").to(device)
|
40 |
+
streamer = TextIteratorStreamer(
|
41 |
+
tok, timeout=10.0, skip_prompt=True, skip_special_tokens=True
|
42 |
+
)
|
43 |
+
generate_kwargs = dict(
|
44 |
+
model_inputs,
|
45 |
+
streamer=streamer,
|
46 |
+
max_new_tokens=1024,
|
47 |
+
do_sample=True,
|
48 |
+
top_p=0.95,
|
49 |
+
top_k=1000,
|
50 |
+
temperature=0.75,
|
51 |
+
num_beams=1,
|
52 |
+
)
|
53 |
+
t = Thread(target=model.generate, kwargs=generate_kwargs)
|
54 |
+
t.start()
|
55 |
+
|
56 |
+
# Initialize an empty string to store the generated text
|
57 |
+
partial_text = ""
|
58 |
+
for new_text in streamer:
|
59 |
+
# print(new_text)
|
60 |
+
partial_text += new_text
|
61 |
+
# Yield an empty string to cleanup the message textbox and the updated conversation history
|
62 |
+
yield partial_text
|
63 |
+
|
64 |
+
|
65 |
+
description = """
|
66 |
A resume question-answering interface where a recruter can ask the user about their achievements and skills without the need to interact with them directly or the need to read a really long resume
|
67 |
|
|
|
68 |
"""
|
69 |
+
examples = [
|
70 |
+
"what's your name?",
|
71 |
+
"what's your email adress ?",
|
72 |
+
"what did you study ?",
|
73 |
+
"are you open for work?",
|
74 |
+
"what are your skills ?",
|
75 |
+
"what's your most recent experience ?",
|
76 |
+
]
|
77 |
|
|
|
|
|
|
|
78 |
|
79 |
+
demo = gr.ChatInterface(
|
80 |
+
fn=chat,
|
81 |
+
chatbot=gr.Chatbot(
|
82 |
+
show_label=True,
|
83 |
+
show_share_button=True,
|
84 |
+
show_copy_button=True,
|
85 |
+
likeable=True,
|
86 |
+
layout="bubble",
|
87 |
+
bubble_full_width=False,
|
88 |
+
),
|
89 |
+
examples=examples,
|
90 |
+
title="Resume QA",
|
91 |
+
description=description,
|
92 |
+
autofocus=False,
|
93 |
+
)
|
94 |
demo.launch()
|
requirements.txt
CHANGED
@@ -1,2 +1,3 @@
|
|
1 |
transformers
|
2 |
-
torch
|
|
|
|
1 |
transformers
|
2 |
+
torch==2.2.0
|
3 |
+
spaces
|