Spaces:
Runtime error
Runtime error
""" Modified from https://huggingface.co/spaces/flax-community/gpt2-indonesian/tree/main """ | |
import json | |
import requests | |
from mtranslate import translate | |
from prompts import PROMPT_LIST | |
import streamlit as st | |
import random | |
description = """ | |
## Overview | |
* **Overall Result:** So Fluent in Mongolian | |
* **Data:** [mC4-bn](https://huggingface.co/datasets/mc4) | |
* **Train Steps:** 250k steps | |
* **Contributors:** M Saiful Bari,Khalid Saifullah,Ibrahim Musa, Tasmiah Tahsin Mayeesha, Ritobrata Ghosh | |
* **link** [[🤗 huggingface](https://huggingface.co/flax-community/gpt2-bengali/)] | |
""" | |
headers = {} | |
MODELS = { | |
"GPT-2 Bengali": { | |
"url": "https://api-inference.huggingface.co/models/flax-community/gpt2-bengali" | |
}, | |
"GPT-2 Finetuned(On Bengali Songs)": { | |
"url": "https://api-inference.huggingface.co/models/khalidsaifullaah/bengali-lyricist-gpt2" | |
}, | |
} | |
def query(payload, model_name): | |
data = json.dumps(payload) | |
print("model url:", MODELS[model_name]["url"]) | |
response = requests.request("POST", MODELS[model_name]["url"], headers=headers, data=data) | |
return json.loads(response.content.decode("utf-8")) | |
def process(text: str, | |
model_name: str, | |
max_len: int, | |
temp: float, | |
top_k: int, | |
top_p: float): | |
payload = { | |
"inputs": text, | |
"parameters": { | |
"max_new_tokens": max_len, | |
"top_k": top_k, | |
"top_p": top_p, | |
"temperature": temp, | |
"repetition_penalty": 2.0, | |
}, | |
"options": { | |
"use_cache": True, | |
} | |
} | |
return query(payload, model_name) | |
st.set_page_config(page_title="Bengali GPT-2 Demo") | |
st.title("Bengali GPT-2") | |
st.sidebar.subheader("Configurable parameters") | |
max_len = st.sidebar.number_input( | |
"Maximum length", | |
value=30, | |
help="The maximum length of the sequence to be generated." | |
) | |
temp = st.sidebar.slider( | |
"Temperature", | |
value=1.0, | |
min_value=0.1, | |
max_value=100.0, | |
help="The value used to module the next token probabilities." | |
) | |
top_k = st.sidebar.number_input( | |
"Top k", | |
value=10, | |
help="The number of highest probability vocabulary tokens to keep for top-k-filtering." | |
) | |
top_p = st.sidebar.number_input( | |
"Top p", | |
value=0.95, | |
help=" If set to float < 1, only the most probable tokens with probabilities that add up to top_p or higher are kept for generation." | |
) | |
do_sample = st.sidebar.selectbox('Sampling?', (True, False), help="Whether or not to use sampling; use greedy decoding otherwise.") | |
st.markdown( | |
"""Bengali GPT-2 demo. Part of the [Huggingface JAX/Flax event](https://discuss.huggingface.co/t/open-to-the-community-community-week-using-jax-flax-for-nlp-cv/). Also features a finetuned version on bengali song lyrics.""" | |
) | |
st.write(description) | |
model_name = st.selectbox('Model',(['GPT-2 Bengali', 'GPT-2 Finetuned(On Bengali Songs)'])) | |
ALL_PROMPTS = list(PROMPT_LIST.keys())+["Custom"] | |
prompt = st.selectbox('Prompt', ALL_PROMPTS, index=len(ALL_PROMPTS)-1) | |
if prompt == "Custom": | |
prompt_box = "Enter your text here" | |
else: | |
prompt_box = random.choice(PROMPT_LIST[prompt]) | |
text = st.text_area("Enter text", prompt_box) | |
if st.button("Run"): | |
with st.spinner(text="Getting results..."): | |
st.subheader("Result") | |
print(f"maxlen:{max_len}, temp:{temp}, top_k:{top_k}, top_p:{top_p}") | |
result = process(text=text, | |
model_name=model_name, | |
max_len=int(max_len), | |
temp=temp, | |
top_k=int(top_k), | |
top_p=float(top_p)) | |
print("result:", result) | |
if "error" in result: | |
if type(result["error"]) is str: | |
st.write(f'{result["error"]}.', end=" ") | |
if "estimated_time" in result: | |
st.write(f'Please try it again in about {result["estimated_time"]:.0f} seconds') | |
else: | |
if type(result["error"]) is list: | |
for error in result["error"]: | |
st.write(f'{error}') | |
else: | |
result = result[0]["generated_text"] | |
st.write(result.replace("\n", " \n")) | |
st.text("English translation") | |
st.write(translate(result, "en", "bn").replace("\n", " \n")) | |