Spaces:
Runtime error
Runtime error
JuliaUpton
commited on
Commit
•
8d63e99
1
Parent(s):
eec1d0a
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import torch
|
2 |
+
from peft import PeftModel, PeftConfig
|
3 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
4 |
+
|
5 |
+
peft_model_id = "JuliaUpton/Math_AI"
|
6 |
+
config = PeftConfig.from_pretrained(peft_model_id)
|
7 |
+
model = AutoModelForCausalLM.from_pretrained(config.base_model_name_or_path, return_dict=True, load_in_8bit=False)
|
8 |
+
tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)
|
9 |
+
"
|
10 |
+
# Load the Lora model
|
11 |
+
merged_model = PeftModel.from_pretrained(model, peft_model_id)
|
12 |
+
|
13 |
+
def input_from_text(instruction):
|
14 |
+
return f"<s>[INST]Below is a math inquiry, please answer it as a math expert showing your thought process.\n\n### Inquiry:\n{instruction}\n\n### Response:[/INST]"
|
15 |
+
|
16 |
+
def make_inference(instruction):
|
17 |
+
inputs = mixtral_tokenizer(input_from_text(instruction), return_tensors="pt")
|
18 |
+
|
19 |
+
outputs = merged_model.generate(
|
20 |
+
**inputs,
|
21 |
+
max_new_tokens=150,
|
22 |
+
generation_kwargs={"repetition_penalty" : 1.7}
|
23 |
+
)
|
24 |
+
# print(mixtral_tokenizer.decode(outputs[0], skip_special_tokens=True))
|
25 |
+
result = mixtral_tokenizer.decode(outputs[0], skip_special_tokens=True).split("[/INST]")[1]
|
26 |
+
return result
|
27 |
+
|
28 |
+
if __name__ == "__main__":
|
29 |
+
# make a gradio interface
|
30 |
+
import gradio as gr
|
31 |
+
|
32 |
+
gr.Interface(
|
33 |
+
make_inference,
|
34 |
+
[
|
35 |
+
gr.Textbox(lines=5, label="Instruction"),
|
36 |
+
],
|
37 |
+
gr.Textbox(label="Answer"),
|
38 |
+
title="Math-AI",
|
39 |
+
description="Math-AI is a generative model that answers math questions",
|
40 |
+
).launch()
|