BigSalmon commited on
Commit
0a6d2ab
·
1 Parent(s): 13a293f

Create app.py

Browse files
Files changed (1) hide show
  1. app.py +81 -0
app.py ADDED
@@ -0,0 +1,81 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import streamlit as st
2
+ from transformers import AutoTokenizer, AutoModelForCausalLM
3
+ import torch
4
+
5
+ first = """informal english: corn fields are all across illinois, visible once you leave chicago.\nTranslated into the Style of Abraham Lincoln: corn fields ( permeate illinois / span the state of illinois / ( occupy / persist in ) all corners of illinois / line the horizon of illinois / envelop the landscape of illinois ), manifesting themselves visibly as one ventures beyond chicago.\n\ninformal english:"""
6
+
7
+ @st.cache(allow_output_mutation=True)
8
+ def get_model():
9
+ print("hello")
10
+ model = AutoModelForCausalLM.from_pretrained("BigSalmon/InformalToFormalLincoln16")
11
+ print("hello")
12
+ tokenizer = AutoTokenizer.from_pretrained("BigSalmon/Points2")
13
+ print("hello")
14
+ return model, tokenizer
15
+
16
+ model, tokenizer = get_model()
17
+
18
+ st.text('''How To Make Prompt:
19
+ informal english: sega’s doing mobile gaming right with this approach to exploiting its back catalog, giving people a way to build up their library of classics on devices they’ll always have close to hand for cheap. it’s a far better experience than nintendo’s classic machine sales on its home gaming systems, which tend to be locked in and not passed between hardware generations.
20
+ Translated into the Style of Abraham Lincoln: pursuing the proper course, sega's foray into mobile gaming entails leveraging its back catalog of classics on portable devices that ( are certain to be in reach whenever / will be at the ready for use at all times ) ( without spending a fortune / at minimal cost ). such an endeavor comes as a welcome alternative to nintendo's propensity to ( sell / offer / market ) home gaming systems that are rigidly bound to a single hardware generation.
21
+ Translated into the Style of Abraham Lincoln: at a time when nintendo has become inflexible, ( stubbornly bent on / firmly set on / unyielding in its insistence on / steadfastly transfixed by / uncompromising in its commitment to / rigidly decided on ) consoles that are tethered to a fixed iteration, sega diligently curates its legacy of classic video games on handheld devices.
22
+ informal english: garage band has made people who know nothing about music good at creating music.
23
+ Translated into the Style of Abraham Lincoln: garage band ( offers the uninitiated in music the ability to produce professional-quality compositions / catapults those for whom music is an uncharted art the ability the realize masterpieces / stimulates music novice's competency to yield sublime arrangements / begets individuals of rudimentary musical talent the proficiency to fashion elaborate suites ).
24
+ informal english: chrome extensions can make doing regular tasks much easier to get done.
25
+ Translated into the Style of Abraham Lincoln: chrome extensions ( yield the boon of time-saving convenience / ( expedite the ability to / unlock the means to more readily ) accomplish everyday tasks / turbocharges the velocity with which one can conduct their obligations ).
26
+ informal english: broadband is finally expanding to rural areas, a great development that will thrust them into modern life.
27
+ Translated into the Style of Abraham Lincoln: broadband is ( ( finally / at last / after years of delay ) arriving in remote locations / springing to life in far-flung outposts / inching into even the most backwater corners of the nation ) that will ( hasten their transition into the modern age / leap-frog them into the twenty-first century / facilitate their integration into contemporary life ).
28
+ informal english: national parks are a big part of the us culture.
29
+ Translated into the Style of Abraham Lincoln: the culture of the united states is ( inextricably ( bound up with / molded by / enriched by / enlivened by ) its ( serene / picturesque / pristine / breathtaking ) national parks ).
30
+ informal english: corn fields are all across illinois, visible once you leave chicago.
31
+ Translated into the Style of Abraham Lincoln: corn fields ( permeate illinois / span the state of illinois / ( occupy / persist in ) all corners of illinois / line the horizon of illinois / envelop the landscape of illinois ), manifesting themselves visibly as one ventures beyond chicago.
32
+ informal english:''')
33
+
34
+ temp = st.sidebar.slider("Temperature", 0.7, 1.5)
35
+ number_of_outputs = st.sidebar.slider("Number of Outputs", 5, 50)
36
+ lengths = st.sidebar.slider("Length", 3, 10)
37
+ bad_words = st.text_input("Words You Do Not Want Generated", " core lemon height time ")
38
+
39
+ def run_generate(text, bad_words):
40
+ yo = []
41
+ input_ids = tokenizer.encode(text, return_tensors='pt')
42
+ res = len(tokenizer.encode(text))
43
+ bad_words = bad_words.split()
44
+ bad_word_ids = []
45
+ for bad_word in bad_words:
46
+ bad_word = " " + bad_word
47
+ ids = tokenizer(bad_word).input_ids
48
+ bad_word_ids.append(ids)
49
+ sample_outputs = model.generate(
50
+ input_ids,
51
+ do_sample=True,
52
+ max_length= res + lengths,
53
+ min_length = res + lengths,
54
+ top_k=50,
55
+ temperature=temp,
56
+ num_return_sequences=number_of_outputs,
57
+ bad_words_ids=bad_word_ids
58
+ )
59
+ for i in range(number_of_outputs):
60
+ e = tokenizer.decode(sample_outputs[i])
61
+ e = e.replace(text, "")
62
+ yo.append(e)
63
+ return yo
64
+ with st.form(key='my_form'):
65
+ text = st.text_area(label='Enter sentence', value=first)
66
+ submit_button = st.form_submit_button(label='Submit')
67
+ if submit_button:
68
+ translated_text = run_generate(text, bad_words)
69
+ st.write(translated_text if translated_text else "No translation found")
70
+ with torch.no_grad():
71
+ text2 = str(text)
72
+ print(text2)
73
+ text3 = tokenizer.encode(text2)
74
+ myinput, past_key_values = torch.tensor([text3]), None
75
+ myinput = myinput
76
+ logits, past_key_values = model(myinput, past_key_values = past_key_values, return_dict=False)
77
+ logits = logits[0,-1]
78
+ probabilities = torch.nn.functional.softmax(logits)
79
+ best_logits, best_indices = logits.topk(100)
80
+ best_words = [tokenizer.decode([idx.item()]) for idx in best_indices]
81
+ st.write(best_words)