Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,63 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
|
3 |
+
import torch
|
4 |
+
import transformers
|
5 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
6 |
+
from transformers import pipeline
|
7 |
+
from diffusers import StableDiffusionPipeline
|
8 |
+
|
9 |
+
summarizer = pipeline("summarization")
|
10 |
+
model_id = "runwayml/stable-diffusion-v1-5"
|
11 |
+
|
12 |
+
SAVED_CHECKPOINT = 'mikegarts/distilgpt2-lotr'
|
13 |
+
MIN_WORDS = 120
|
14 |
+
|
15 |
+
def get_image_pipe():
|
16 |
+
pipe = StableDiffusionPipeline.from_pretrained(model_id, torch_dtype=torch.float16, revision="fp16")
|
17 |
+
pipe.to(pipe.device)
|
18 |
+
return pipe
|
19 |
+
|
20 |
+
def get_model():
|
21 |
+
model = AutoModelForCausalLM.from_pretrained(SAVED_CHECKPOINT)
|
22 |
+
tokenizer = AutoTokenizer.from_pretrained(SAVED_CHECKPOINT)
|
23 |
+
return model, tokenizer
|
24 |
+
|
25 |
+
|
26 |
+
def generate(prompt):
|
27 |
+
model, tokenizer = get_model()
|
28 |
+
|
29 |
+
input_context = prompt
|
30 |
+
input_ids = tokenizer.encode(input_context, return_tensors="pt").to(model.device)
|
31 |
+
|
32 |
+
outputs = model.generate(
|
33 |
+
input_ids=input_ids,
|
34 |
+
max_length=100,
|
35 |
+
temperature=0.7,
|
36 |
+
num_return_sequences=3,
|
37 |
+
do_sample=True
|
38 |
+
)
|
39 |
+
|
40 |
+
return tokenizer.decode(outputs[0], skip_special_tokens=True).rsplit('.', 1)[0] + '.'
|
41 |
+
|
42 |
+
def make_image(prompt):
|
43 |
+
pipe = get_image_pipe()
|
44 |
+
image = pipe(prompt).images[0]
|
45 |
+
|
46 |
+
def predict(prompt):
|
47 |
+
story = generate(prompt=prompt)
|
48 |
+
summary = summarizer(story, min_length=5, max_length=20)[0]['summary_text']
|
49 |
+
image = make_image(summary)
|
50 |
+
return story, summarizer(story, min_length=5, max_length=20), image
|
51 |
+
|
52 |
+
|
53 |
+
title = "Lord of the rings app"
|
54 |
+
description = """A Lord of the rings insired app that combines text and image generation"""
|
55 |
+
|
56 |
+
gr.Interface(
|
57 |
+
fn=predict,
|
58 |
+
inputs="textbox",
|
59 |
+
outputs=["text", "text", "image"],
|
60 |
+
title=title,
|
61 |
+
description=description,
|
62 |
+
examples=[["My new adventure would be"], ["Then I a hobbit appeared"], ["Frodo told me"]]
|
63 |
+
).launch(share=True)
|