kadirnar's picture
Update app.py
5ee913f
raw
history blame
2.35 kB
from transformers import pipeline, set_seed
from transformers import BioGptTokenizer, BioGptForCausalLM
from multilingual_translation import translate
from utils import lang_ids
import gradio as gr
model_list = [
"microsoft/biogpt",
"microsoft/BioGPT-Large-PubMedQA"
]
lang_list = list(lang_ids.keys())
def translate_to_english(text, base_lang):
if base_lang == "English":
return text
else:
base_lang = lang_ids[base_lang]
new_text = translate("facebook/m2m100_418M", text, base_lang, "en")
return new_text
def biogpt(
prompt: str,
model_id: str,
max_length: int = 25,
num_return_sequences: int = 5,
base_lang: str = "en"
):
en_prompt = translate_to_english(prompt, base_lang)
model = BioGptForCausalLM.from_pretrained(model_id)
tokenizer = BioGptTokenizer.from_pretrained(model_id)
generator = pipeline('text-generation', model=model, tokenizer=tokenizer)
set_seed(42)
output = generator(en_prompt, max_length=max_length, num_return_sequences=num_return_sequences, do_sample=True)
output_dict = {
"1": output[0]['generated_text'],
"2": output[1]['generated_text'],
"3": output[2]['generated_text'],
"4": output[3]['generated_text'],
"5": output[4]['generated_text']
}
output_text = f'{output_dict["1"]}\n\n{output_dict["2"]}\n\n{output_dict["3"]}\n\n{output_dict["4"]}\n\n{output_dict["5"]}'
return en_prompt, output_text
inputs = [
gr.inputs.Textbox(label="Prompt", lines=5, default="COVID-19 is"),
gr.Dropdown(model_list, value="microsoft/biogpt", label="Model ID"),
gr.inputs.Slider(5, 100, 25, default=25, label="Max Length"),
gr.inputs.Slider(1, 10, 5, default=5, label="Num Return Sequences"),
gr.Dropdown(lang_list, value="en", label="Base Language")
]
outputs = [
gr.outputs.Textbox(label="Prompt"),
gr.outputs.Textbox(label="Output")
]
examples = [
["COVID-19 is", "microsoft/biogpt", 25, 5, "en"],
["Kanser", "microsoft/biogpt", 25, 5, "tr"]
]
title = " BioGPT: Generative Pre-trained Transformer for Biomedical Text Generation and Mining"
demo_app = gr.Interface(
biogpt,
inputs,
outputs,
title=title,
examples=examples,
cache_examples=True,
)
demo_app.launch(debug=True, enable_queue=True)