from transformers import AutoModelForSeq2SeqLM, AutoTokenizer import gradio as gr # Load fine-tuned BanglaT5 models for different tasks translation_model_en_bn = AutoModelForSeq2SeqLM.from_pretrained("csebuetnlp/banglat5_nmt_en_bn") translation_tokenizer_en_bn = AutoTokenizer.from_pretrained("csebuetnlp/banglat5_nmt_en_bn") translation_model_bn_en = AutoModelForSeq2SeqLM.from_pretrained("csebuetnlp/banglat5_nmt_bn_en") translation_tokenizer_bn_en = AutoTokenizer.from_pretrained("csebuetnlp/banglat5_nmt_bn_en") summarization_model = AutoModelForSeq2SeqLM.from_pretrained("skl25/banglat5_xlsum_fine-tuned") summarization_tokenizer = AutoTokenizer.from_pretrained("skl25/banglat5_xlsum_fine-tuned") paraphrase_model = AutoModelForSeq2SeqLM.from_pretrained("csebuetnlp/banglat5_banglaparaphrase") paraphrase_tokenizer = AutoTokenizer.from_pretrained("csebuetnlp/banglat5_banglaparaphrase") # Define task functions def translate_text_en_bn(input_text): inputs = translation_tokenizer_en_bn(input_text, return_tensors="pt") outputs = translation_model_en_bn.generate(**inputs) return translation_tokenizer_en_bn.decode(outputs[0], skip_special_tokens=True) def translate_text_bn_en(input_text): inputs = translation_tokenizer_bn_en(input_text, return_tensors="pt") outputs = translation_model_bn_en.generate(**inputs) return translation_tokenizer_bn_en.decode(outputs[0], skip_special_tokens=True) def summarize_text(input_text): inputs = summarization_tokenizer(input_text, return_tensors="pt") outputs = summarization_model.generate(**inputs) return summarization_tokenizer.decode(outputs[0], skip_special_tokens=True) def paraphrase_text(input_text): inputs = paraphrase_tokenizer(input_text, return_tensors="pt") outputs = paraphrase_model.generate(**inputs) return paraphrase_tokenizer.decode(outputs[0], skip_special_tokens=True) # Process input based on task def process_text(text, task): task_funcs = { "Translate English to Bengali": translate_text_en_bn, "Translate Bengali to English": translate_text_bn_en, "Summarize": summarize_text, "Paraphrase": paraphrase_text } return task_funcs.get(task, lambda x: "Invalid Task")(text) # Task-specific examples examples_en_bn_translation = [ ["The sky is blue, and the weather is nice."], ["Artificial intelligence is shaping the future."], ["Bangladesh is known for its rich culture and heritage."] ] examples_bn_en_translation = [ ["বাংলাদেশ দক্ষিণ এশিয়ার একটি সার্বভৌম রাষ্ট্র।"], ["ঢাকা বাংলাদেশের রাজধানী।"], ["রবীন্দ্রনাথ ঠাকুরের গান বাংলা সংস্কৃতির একটি অবিচ্ছেদ্য অংশ।"] ] examples_summarization = [ ["The Department of Computer Science and Engineering, established in 1982, was the first of its kind in Bangladesh. " "Attracting top students from all over the country, it offers both undergraduate and postgraduate degrees."], ["Climate change is one of the biggest challenges we face today. With rising temperatures and unpredictable weather, " "the world needs to come together to find sustainable solutions."], ["Technology has advanced rapidly over the past decade, with innovations in fields like AI, robotics, and quantum computing."] ] examples_paraphrasing = [ ["The cat is sitting on the mat."], ["He was very happy to receive the award."], ["The weather today is sunny and warm."] ] # Define the Gradio interface with enhanced visuals iface = gr.Interface( fn=process_text, inputs=[ "text", gr.Dropdown( ["Translate English to Bengali", "Translate Bengali to English", "Summarize", "Paraphrase"], label="Choose Task", elem_id="dropdown-task" ) ], outputs="text", title="BanglaT5 Model Hub", description="A multi-functional tool for translation, summarization, and paraphrasing using BanglaT5 models.", theme="finlaydog/seafoam", examples=[ *examples_en_bn_translation, # Adding 3 examples for English to Bengali translation *examples_bn_en_translation, # Adding 3 examples for Bengali to English translation *examples_summarization, # Adding 3 examples for Summarization *examples_paraphrasing # Adding 3 examples for Paraphrasing ], allow_flagging="auto", ) # Launch the Gradio app iface.launch(inline=False)