update
Browse files- app.py +8 -63
- complete_poem.py +65 -0
- generate_poem.py +14 -0
app.py
CHANGED
|
@@ -1,66 +1,11 @@
|
|
| 1 |
import gradio as gr
|
| 2 |
-
from
|
|
|
|
| 3 |
|
| 4 |
-
|
| 5 |
-
|
| 6 |
-
|
| 7 |
-
"
|
| 8 |
-
"Libosa2707/vietnamese-poem-luc-bat-gpt2"
|
| 9 |
-
),
|
| 10 |
-
"Bay Chu": AutoModelForCausalLM.from_pretrained(
|
| 11 |
-
"Libosa2707/vietnamese-poem-bay-chu-gpt2"
|
| 12 |
-
),
|
| 13 |
-
"Tam Chu": AutoModelForCausalLM.from_pretrained(
|
| 14 |
-
"Libosa2707/vietnamese-poem-tam-chu-gpt2"
|
| 15 |
-
),
|
| 16 |
-
"Nam Chu": AutoModelForCausalLM.from_pretrained(
|
| 17 |
-
"Libosa2707/vietnamese-poem-nam-chu-gpt2"
|
| 18 |
-
),
|
| 19 |
-
}
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
def generate_poem(text, style):
|
| 23 |
-
# Preprocess the input text
|
| 24 |
-
text = text.strip()
|
| 25 |
-
text = text.lower()
|
| 26 |
-
|
| 27 |
-
# Choose the model based on the selected style
|
| 28 |
-
model = models[style]
|
| 29 |
-
|
| 30 |
-
# Tokenize the input line
|
| 31 |
-
input_ids = tokenizer.encode(text, return_tensors="pt")[:, :-1]
|
| 32 |
-
|
| 33 |
-
# Generate text
|
| 34 |
-
output = model.generate(input_ids, max_length=100, do_sample=True, temperature=0.7)
|
| 35 |
-
|
| 36 |
-
# Decode the output
|
| 37 |
-
generated_text = tokenizer.decode(
|
| 38 |
-
output[:, input_ids.shape[-1] :][0], skip_special_tokens=True
|
| 39 |
-
)
|
| 40 |
-
|
| 41 |
-
text = text + " " + generated_text
|
| 42 |
-
|
| 43 |
-
# Post-process the output
|
| 44 |
-
text = text.replace("<unk>", "\n")
|
| 45 |
-
pretty_text = ""
|
| 46 |
-
for idx, line in enumerate(text.split("\n")):
|
| 47 |
-
line = line.strip()
|
| 48 |
-
if not line:
|
| 49 |
-
continue
|
| 50 |
-
line = line[0].upper() + line[1:]
|
| 51 |
-
pretty_text += line + "\n"
|
| 52 |
-
|
| 53 |
-
return pretty_text
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
gradio_interface = gr.Interface(
|
| 57 |
-
fn=generate_poem,
|
| 58 |
-
inputs=[
|
| 59 |
-
gr.components.Textbox(lines=1, placeholder="First words of the poem"),
|
| 60 |
-
gr.components.Dropdown(
|
| 61 |
-
choices=["Luc Bat", "Bay Chu", "Tam Chu", "Nam Chu"], label="Style"
|
| 62 |
-
),
|
| 63 |
-
],
|
| 64 |
-
outputs="text",
|
| 65 |
)
|
| 66 |
-
|
|
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
+
from complete_poem import complete_poem_interface
|
| 3 |
+
from generate_poem import generate_poem_interface
|
| 4 |
|
| 5 |
+
gr_interface = gr.TabbedInterface(
|
| 6 |
+
[complete_poem_interface, generate_poem_interface],
|
| 7 |
+
title="Thơ AI",
|
| 8 |
+
description="Ứng dụng mô hình ngôn ngữ cực lớn để sinh ra các bài thơ Việt Nam",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
)
|
| 10 |
+
|
| 11 |
+
gr_interface.launch()
|
complete_poem.py
ADDED
|
@@ -0,0 +1,65 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
| 3 |
+
|
| 4 |
+
tokenizer = AutoTokenizer.from_pretrained("vinai/phobert-base")
|
| 5 |
+
# Define your models
|
| 6 |
+
models = {
|
| 7 |
+
"Luc Bat": AutoModelForCausalLM.from_pretrained(
|
| 8 |
+
"Libosa2707/vietnamese-poem-luc-bat-gpt2"
|
| 9 |
+
),
|
| 10 |
+
"Bay Chu": AutoModelForCausalLM.from_pretrained(
|
| 11 |
+
"Libosa2707/vietnamese-poem-bay-chu-gpt2"
|
| 12 |
+
),
|
| 13 |
+
"Tam Chu": AutoModelForCausalLM.from_pretrained(
|
| 14 |
+
"Libosa2707/vietnamese-poem-tam-chu-gpt2"
|
| 15 |
+
),
|
| 16 |
+
"Nam Chu": AutoModelForCausalLM.from_pretrained(
|
| 17 |
+
"Libosa2707/vietnamese-poem-nam-chu-gpt2"
|
| 18 |
+
),
|
| 19 |
+
}
|
| 20 |
+
|
| 21 |
+
|
| 22 |
+
def complete_poem(text, style):
|
| 23 |
+
# Preprocess the input text
|
| 24 |
+
text = text.strip()
|
| 25 |
+
text = text.lower()
|
| 26 |
+
|
| 27 |
+
# Choose the model based on the selected style
|
| 28 |
+
model = models[style]
|
| 29 |
+
|
| 30 |
+
# Tokenize the input line
|
| 31 |
+
input_ids = tokenizer.encode(text, return_tensors="pt")[:, :-1]
|
| 32 |
+
|
| 33 |
+
# Generate text
|
| 34 |
+
output = model.generate(input_ids, max_length=100, do_sample=True, temperature=0.7)
|
| 35 |
+
|
| 36 |
+
# Decode the output
|
| 37 |
+
generated_text = tokenizer.decode(
|
| 38 |
+
output[:, input_ids.shape[-1] :][0], skip_special_tokens=True
|
| 39 |
+
)
|
| 40 |
+
|
| 41 |
+
text = text + " " + generated_text
|
| 42 |
+
|
| 43 |
+
# Post-process the output
|
| 44 |
+
text = text.replace("<unk>", "\n")
|
| 45 |
+
pretty_text = ""
|
| 46 |
+
for idx, line in enumerate(text.split("\n")):
|
| 47 |
+
line = line.strip()
|
| 48 |
+
if not line:
|
| 49 |
+
continue
|
| 50 |
+
line = line[0].upper() + line[1:]
|
| 51 |
+
pretty_text += line + "\n"
|
| 52 |
+
|
| 53 |
+
return pretty_text
|
| 54 |
+
|
| 55 |
+
|
| 56 |
+
complete_poem_interface = gr.Interface(
|
| 57 |
+
fn=complete_poem,
|
| 58 |
+
inputs=[
|
| 59 |
+
gr.components.Textbox(lines=1, placeholder="First words of the poem"),
|
| 60 |
+
gr.components.Dropdown(
|
| 61 |
+
choices=["Luc Bat", "Bay Chu", "Tam Chu", "Nam Chu"], label="Style"
|
| 62 |
+
),
|
| 63 |
+
],
|
| 64 |
+
outputs="text",
|
| 65 |
+
)
|
generate_poem.py
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
|
| 3 |
+
|
| 4 |
+
def generate_poem(text):
|
| 5 |
+
pass
|
| 6 |
+
|
| 7 |
+
|
| 8 |
+
generate_poem_interface = gr.Interface(
|
| 9 |
+
fn=generate_poem,
|
| 10 |
+
inputs=[
|
| 11 |
+
gr.components.Textbox(lines=1, placeholder="First words of the poem"),
|
| 12 |
+
],
|
| 13 |
+
outputs="text",
|
| 14 |
+
)
|