app.py

from cProfile import label
from fastapi import File
import gradio as gr
from gib_detect_module import detect
import csv
import torch
import tensorflow as tf


from transformers import AutoModelForSequenceClassification, AutoTokenizer
DLmodel = AutoModelForSequenceClassification.from_pretrained("madhurjindal/autonlp-Gibberish-Detector-492513457", use_auth_token=True)
tokenizer = AutoTokenizer.from_pretrained("madhurjindal/autonlp-Gibberish-Detector-492513457", use_auth_token=True)



def greet(name):
    return "Hello " + name + "!!"

def detect_gibberish(line,f):
    if line:
        if detect(line):
            return "Valid!!!!", None,None
        else:
            return "Bollocks Giberrish",None,None
    elif f:
        return None, annotate_csv(f), None

 


def annotate_csv(f):
    with open(f.name) as csvfile:
        creader = csv.reader(csvfile, delimiter=',', quotechar='"')
    
        with open('out.csv', 'w', newline='') as csvout:
            cwriter = csv.writer(csvout, delimiter=',',
                                quotechar='"', quoting=csv.QUOTE_MINIMAL)
            for row in creader:
                row.append(str(detect(row[0])))
                cwriter.writerow(row) 
            return "out.csv" 


def annotate_csv_deep(f):
    labels = DLmodel.config.id2label
    with open(f.name) as csvfile:
        creader = csv.reader(csvfile, delimiter=',', quotechar='"')
        with open('out.csv', 'w', newline='') as csvout:
            cwriter = csv.writer(csvout, delimiter=',',
                                quotechar='"', quoting=csv.QUOTE_MINIMAL)
            for row in creader:
                inputs = tokenizer(row, return_tensors="pt")
                outputs = DLmodel(**inputs)
                probs = outputs.logits.softmax(dim=-1).detach().cpu().flatten().numpy().tolist()
                idx = probs.index(max(probs))
                
                row.append(labels[idx])
                row.append("{:.0%}".format(probs[idx]) )
                cwriter.writerow(row) 
            return "out.csv" 


def detect_gibberish_deep(line,f):
    if line:
        inputs = tokenizer(line, return_tensors="pt")
        labels = DLmodel.config.id2label
        outputs = DLmodel(**inputs)
        probs = outputs.logits.softmax(dim=-1).detach().cpu().flatten().numpy().tolist()
        output=dict(zip(labels.values(), probs))
        readable_output=""
        for k,v in output.items():
            readable_output+=k+" : "+ "{:.0%}".format(v) + "\n"
        return readable_output, None, output
    if f:
        return None, annotate_csv_deep(f),None



def detect_gibberish_abstract(model, line,f):
    if model == "Deep Learning Model":
        return detect_gibberish_deep(line,f)
    else:
        return detect_gibberish(line, f)


inputLine=gr.inputs.Textbox(lines=1, placeholder="Input text here, if both text and file have values, only the text input will be processed.", default="", label="Text", optional=False)
inputFile=gr.inputs.File(file_count="single", type="file", label="File to Annotate", optional=True)


choices = ["Deep Learning Model", "Markov Chain"]
inputModel=gr.inputs.Dropdown(choices)


outputLine=gr.outputs.Textbox(type="auto", label=None)
outputFile=gr.outputs.File( label="Annotated CSV")
label = gr.outputs.Label(num_top_classes=4)


examples=[
    ["Deep Learning Model","quetzalcoatl","demo_blank.csv"],
    ["Deep Learning Model","aasdf","demo_blank.csv"],
    ["Deep Learning Model","Covfefe","demo_blank.csv"],
    ["Markov Chain","quetzalcoatl","demo_blank.csv"],
    ["Markov Chain","aasdf","demo_blank.csv"],
    ["Markov Chain","Covfefe","demo_blank.csv"],
    ["Deep Learning Model","","demo_bad.txt"],
    ["Deep Learning Model","","demo_mixed.txt"],
    ["Markov Chain","","demo_bad.txt"],
    ["Markov Chain","","demo_mixed.txt"],
]
#iface = gr.Interface(fn=[detect_gibberish], inputs=["text",inputFile], outputs=["text",outputFile],examples=examples, allow_flagging='never')

#iface.launch()


iface = gr.Interface(fn=[detect_gibberish_abstract], inputs=[inputModel,inputLine,inputFile], outputs=["text",outputFile,label],examples=examples, allow_flagging='never')
iface.launch()