Spaces:
No application file
No application file
gradio bigtranslate
Browse files- gradio_bigtranslate.py +64 -0
- requirements.txt +4 -0
gradio_bigtranslate.py
ADDED
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# -*- coding: utf-8 -*-
|
2 |
+
"""gradio-bigtranslate.ipynb
|
3 |
+
|
4 |
+
Automatically generated by Colab.
|
5 |
+
|
6 |
+
Original file is located at
|
7 |
+
https://colab.research.google.com/drive/1Rtw0lupjDrxW3bRiuFmxFlxKO40X6AuU
|
8 |
+
"""
|
9 |
+
|
10 |
+
# ! pip install gradio
|
11 |
+
|
12 |
+
# ! pip install transformers
|
13 |
+
|
14 |
+
from huggingface_hub import notebook_login
|
15 |
+
|
16 |
+
notebook_login()
|
17 |
+
|
18 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer
|
19 |
+
|
20 |
+
|
21 |
+
# Load the model and tokenizer
|
22 |
+
# ! pip install optimum auto-gptq
|
23 |
+
|
24 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer, GPTQConfig # Import necessary modules
|
25 |
+
|
26 |
+
# Load the model and tokenizer
|
27 |
+
model_name = "TheBloke/BigTranslate-13B-GPTQ"
|
28 |
+
# Configure GPTQ to disable Exllama and use the CUDA backend
|
29 |
+
quantization_config = GPTQConfig(bits=4, disable_exllama=True)
|
30 |
+
model = AutoModelForCausalLM.from_pretrained(model_name, quantization_config=quantization_config)
|
31 |
+
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
32 |
+
|
33 |
+
import gradio as gr
|
34 |
+
|
35 |
+
supported_languages = {
|
36 |
+
"English": "en",
|
37 |
+
"French": "fr",
|
38 |
+
"Spanish": "es",
|
39 |
+
"German": "de",
|
40 |
+
# Add more languages and their codes as needed
|
41 |
+
}
|
42 |
+
|
43 |
+
def translate_text(input_text, output_language):
|
44 |
+
# Prefix the input text with the target language code
|
45 |
+
prefixed_input_text = f">>{output_language}<< {input_text}"
|
46 |
+
# Tokenize the input text
|
47 |
+
inputs = tokenizer(prefixed_input_text, return_tensors="pt")
|
48 |
+
# Generate translation
|
49 |
+
outputs = model.generate(inputs['input_ids'], max_length=40, num_beams=4, early_stopping=True)
|
50 |
+
# Decode the output
|
51 |
+
translated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
52 |
+
return translated_text
|
53 |
+
|
54 |
+
# Create the Gradio interface
|
55 |
+
iface = gr.Interface(
|
56 |
+
fn=translate_text,
|
57 |
+
inputs=[
|
58 |
+
gr.Textbox(lines=2, placeholder="Enter text here..."),
|
59 |
+
gr.Dropdown(choices=list(supported_languages.keys()), label="Select output language")
|
60 |
+
],
|
61 |
+
outputs="text"
|
62 |
+
)
|
63 |
+
# Launch the interface
|
64 |
+
iface.launch()
|
requirements.txt
ADDED
@@ -0,0 +1,4 @@
|
|
|
|
|
|
|
|
|
|
|
1 |
+
gradio
|
2 |
+
transformers
|
3 |
+
optimum
|
4 |
+
auto-gptq
|