File size: 3,070 Bytes
f97bf68
 
69855af
f97bf68
 
b0413b4
 
 
 
 
 
 
 
 
 
 
 
f97bf68
1aa631a
244f082
 
 
 
 
 
 
f97bf68
244f082
1aa631a
f97bf68
244f082
e1957fa
69855af
a81c6ef
ead2968
 
 
244f082
 
 
 
 
 
 
 
 
 
f97bf68
09e9faf
f97bf68
 
1aa631a
 
 
 
 
 
 
 
 
 
 
 
 
 
2d82305
f97bf68
69855af
244f082
 
1aa631a
 
ead2968
 
 
 
 
6753cc6
244f082
 
f97bf68
1aa631a
f97bf68
e1957fa
6c7765b
f97bf68
12e69c8
f97bf68
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
import gradio as gr
import torch
import random
from transformers import T5Tokenizer, T5ForConditionalGeneration

tokenizer = T5Tokenizer.from_pretrained("google/flan-t5-small")
model = T5ForConditionalGeneration.from_pretrained("roborovski/superprompt-v1", torch_dtype=torch.float16)

if torch.cuda.is_available():
    device = "cuda"
    print("Using GPU")
else:
    device = "cpu"
    print("Using CPU")

model.to(device)

def generate(
    system_prompt,
    prompt,
    max_new_tokens,
    repetition_penalty,
    temperature,
    top_p,
    top_k,
    seed,
):

    input_text = f"{system_prompt}, {prompt}"
    input_ids = tokenizer(input_text, return_tensors="pt").input_ids.to(device)

    if seed == 0:
        seed = random.randint(1, 100000)
        torch.manual_seed(seed)
    else:
        torch.manual_seed(seed)
        
    outputs = model.generate(
        input_ids,
        max_new_tokens=max_new_tokens,
        repetition_penalty=repetition_penalty,
        do_sample=True,
        temperature=temperature,
        top_p=top_p,
        top_k=top_k,
    )

    better_prompt = tokenizer.decode(outputs[0])
    better_prompt = better_prompt.replace("<pad»", "").replace("</s>", "")
    return better_prompt

prompt = gr.Textbox(label="Prompt", interactive=True)

system_prompt = gr.Textbox(label="System Prompt", interactive=True)

max_new_tokens = gr.Slider(value=512, minimum=250, maximum=512, step=1, interactive=True, label="Max New Tokens", info="The maximum numbers of new tokens, controls how long is the output")
    
repetition_penalty = gr.Slider(value=1.2, minimum=0, maximum=2, step=0.05, interactive=True, label="Repetition Penalty", info="Penalize repeated tokens, making the AI repeat less itself")
    
temperature = gr.Slider(value=0.5, minimum=0, maximum=1, step=0.05, interactive=True, label="Temperature", info="Higher values produce more diverse outputs")

top_p = gr.Slider(value=1, minimum=0, maximum=2, step=0.05, interactive=True, label="Top P", info="Higher values sample more low-probability tokens")

top_k = gr.Slider(value=1, minimum=1, maximum=100, step=1, interactive=True, label="Top K", info="Higher k means more diverse outputs by considering a range of tokens")

seed = gr.Number(value=42, interactive=True, label="Seed", info="A starting point to initiate the generation process, put 0 for a random one")


examples = [
    [
        "A storefront with 'Text to Image' written on it.",
        "Expand the following prompt to add more detail:",
        512,
        1.2,
        0.5,
        1,
        50,
        42,
    ]
]

gr.Interface(
    fn=generate,
    inputs=[prompt, system_prompt, max_new_tokens, repetition_penalty, temperature, top_p, top_k, seed],
    outputs=gr.Textbox(label="Better Prompt", interactive=True),
    title="SuperPrompt-v1",
    description="Make your prompts more detailed!<br>Model used: https://huggingface.co/roborovski/superprompt-v1<br>Hugging Face Space made by [Nick088](https://linktr.ee/Nick088)",
    examples=examples,
    concurrency_limit=20,
).launch(show_api=False)