import gradio as gr import numpy as np from transformers import Trainer, AutoTokenizer, XLMRobertaForSequenceClassification import torch import pickle import re import PyPDF2 import docx import xml.etree.ElementTree as ET def checker(content): #if isinstance(content, bytes):#falls es ein Byte-String ist # content = content.decode('utf-8') sentences = re.split(r"(?<=[!.;?:])\s+(?![a-zà-öø-ÿα-ωа-яא-תء-ي])", content) model_name = "abischoff/test-trainer" tokenizer = AutoTokenizer.from_pretrained(model_name) model = XLMRobertaForSequenceClassification.from_pretrained(model_name,num_labels=1) sum = 0.0 bias = -0.0666 for dic in sentences: inputs = tokenizer(dic.strip(), return_tensors="pt") with torch.no_grad(): logits = model(**inputs).logits.flatten()[0].numpy()+bias logits = np.clip(logits,0,1) sum += logits res = sum / len(sentences) return res # Funktion zur Verarbeitung der hochgeladenen Datei def process_file(file): if file: content = read_file(file) if content == "": return "Datei konnte nicht gelesen werden" else: res = checker(content) return res return "Fehler bei der Verarbeitung" # lese die Datei ein def read_file(file): content = "" if ".pdf" in file.name: # creating a pdf reader object reader = PyPDF2.PdfReader(file.name) counter = 0 while counter < len(reader.pages): content += reader.pages[counter].extract_text() counter += 1 elif ".docx" in file.name: doc = docx.Document(file.name) for para in doc.paragraphs: content += para.text elif ".xml" in file.name: tree = ET.parse(file.name) root = tree.getroot() for text in root.itertext(): if text != "": content += text else: with open(file.name, 'r', encoding='utf-8') as f: content = f.read() #with open(file.name, 'rb') as f: # content = pickle.load(f) print(content) return content def is_valid_file(file_path: str, file_types: list[str]) -> bool: file_extension = os.path.splitext(file_path)[1] return file_extension in file_types # Gradio Interface with gr.Blocks(css=".gr-box {border-radius: 8px;}") as demo: # App-Titel gr.Markdown( """ ## 📝 Checker Ziehe deine Datei hierhin und lasse den Quality Score errechnen! """ ) # Drag & Drop Datei-Upload with gr.Row(): with gr.Column(): file_input = gr.File(label="Datei hochladen", file_types=['file'], interactive=True) with gr.Column(): output_text = gr.Textbox(label="Dateiinhalt", lines=10) # Callback zum Verarbeiten der Datei file_input.change(fn=process_file, inputs=[file_input], outputs=[output_text]) # App starten demo.launch()