Update app.py

app.py

@@ -1,6 +1,6 @@
 import gradio as gr
 import spaces
-import transformers
+from transformers import pipeline
 import torch
 
 DESCRIPTION="""
@@ -21,32 +21,86 @@ ner_example = [["Benim adım Turna."]]
 t2t_example = [["Paraphrase: Bu üründen çok memnun kaldım."]]
 nli_example = [["Bunu çok beğendim. Bunu çok sevdim."]]
 
-t2t_gen = gr.load("huggingface/boun-tabi-LMG/TURNA", examples =t2t_example,  title="Text-to-Text Generation", description="Please enter an instruction with a prefix to generate.")
-summarization = gr.load("huggingface/boun-tabi-LMG/turna_summarization_mlsum",examples =long_text, title="Summarization", description="TURNA fine-tuned on MLSUM. Enter a text to summarize below.")
-news_sum =  gr.load("huggingface/boun-tabi-LMG/turna_summarization_tr_news",examples =long_text, title="News Summarization", description="TURNA fine-tuned on News summarization. Enter a news to summarize.")
-paraphrase = gr.load("huggingface/boun-tabi-LMG/turna_paraphrasing_tatoeba", examples =long_text,title="Paraphrasing")
-paraphrasing_sub = gr.load("huggingface/boun-tabi-LMG/turna_paraphrasing_opensubtitles",examples =long_text, title="Paraphrasing on Subtitles")
 
-ttc = gr.load("huggingface/boun-tabi-LMG/turna_classification_ttc4900", examples =long_text, title="Text Categorization")
-product_reviews = gr.load("huggingface/boun-tabi-LMG/turna_classification_tr_product_reviews", examples=sentiment_example, title="Product Reviews Categorization")
-title_gen = gr.load("huggingface/boun-tabi-LMG/turna_title_generation_mlsum", examples =long_text, title="Title Generation", description="Enter a text to generate title to.")
-sentiment = gr.load("huggingface/boun-tabi-LMG/turna_classification_17bintweet_sentiment",examples=sentiment_example, title="Sentiment Analysis", description="Enter a text to generate title to.")
+t2t_gen_model = pipeline("boun-tabi-LMG/TURNA", device=0) examples =t2t_example,  title="Text-to-Text Generation", description="Please enter an instruction with a prefix to generate.")
+summarization_model = pipeline("boun-tabi-LMG/turna_summarization_mlsum", device=0) examples =long_text, title="Summarization", description="TURNA fine-tuned on MLSUM. Enter a text to summarize below.")
+news_sum = pipeline("boun-tabi-LMG/turna_summarization_tr_news", device=0) examples =long_text, title="News Summarization", description="TURNA fine-tuned on News summarization. Enter a news to summarize.")
+paraphrasing = pipeline("boun-tabi-LMG/turna_paraphrasing_tatoeba", device=0) examples =long_text,title="Paraphrasing")
+paraphrasing_sub = pipeline("boun-tabi-LMG/turna_paraphrasing_opensubtitles", device=0) examples =long_text, title="Paraphrasing on Subtitles")
 
-pos = gr.load("huggingface/boun-tabi-LMG/turna_pos_imst", title="Part of Speech Tagging", examples=ner_example,description="Enter a text to generate title to.")
-nli = gr.load("huggingface/boun-tabi-LMG/turna_nli_nli_tr", title="NLI",examples=nli_example, description="Enter two texts to infer entailment.")
-pos_boun = gr.load("huggingface/boun-tabi-LMG/turna_pos_boun", examples = ner_example, title="Part of Speech Tagging", description="Enter a text to tag parts of speech (POS).")
-stsb = gr.load("huggingface/boun-tabi-LMG/turna_semantic_similarity_stsb_tr", examples=nli_example, title="Semantic Similarity", description="Enter two texts in the input to assess semantic similarity.")
-ner = gr.load("huggingface/boun-tabi-LMG/turna_ner_milliyet", title="NER WikiANN", examples=ner_example, description="Enter a text for NER.")
-ner_wikiann = gr.load("huggingface/boun-tabi-LMG/turna_ner_wikiann", title="NER",examples=ner_example, description="Enter a text for NER.")
+ttc = pipeline("boun-tabi-LMG/turna_classification_ttc4900", device=0) examples =long_text, title="Text Categorization")
+product_reviews = pipeline("boun-tabi-LMG/turna_classification_tr_product_reviews", device=0) examples=sentiment_example, title="Product Reviews Categorization")
+title_gen = pipeline("boun-tabi-LMG/turna_title_generation_mlsum", device=0) examples =long_text, title="Title Generation", description="Enter a text to generate title to.")
+sentiment_model = pipeline("boun-tabi-LMG/turna_classification_17bintweet_sentiment", device=0) examples=sentiment_example, title="Sentiment Analysis", description="Enter a text to generate title to.")
+
+pos_imst = pipeline("boun-tabi-LMG/turna_pos_imst", device=0) title="Part of Speech Tagging", examples=ner_example,description="Enter a text to generate title to.")
+nli_model = pipeline("boun-tabi-LMG/turna_nli_nli_tr", device=0) title="NLI",examples=nli_example, description="Enter two texts to infer entailment.")
+pos_boun = pipeline("boun-tabi-LMG/turna_pos_boun", device=0) examples = ner_example, title="Part of Speech Tagging", description="Enter a text to tag parts of speech (POS).")
+stsb_model = pipeline("boun-tabi-LMG/turna_semantic_similarity_stsb_tr", device=0) examples=nli_example, title="Semantic Similarity", description="Enter two texts in the input to assess semantic similarity.")
+ner_model = pipeline("boun-tabi-LMG/turna_ner_milliyet", device=0) title="NER WikiANN", examples=ner_example, description="Enter a text for NER.")
+ner_wikiann = pipeline("boun-tabi-LMG/turna_ner_wikiann", device=0) title="NER",examples=ner_example, description="Enter a text for NER.")
+
+
+@spaces.GPU
+def sentiment_analysis(input, sentiment=True):
+    if sentiment==True:
+        return sentiment_model(input)
+    else:
+        return product_reviews(input)
+
+@spaces.GPU
+def nli_stsb(input, nli=True):
+    if nli==True:
+        return nli_model(input)
+    else:
+        return stsb_model(input)
+
+@spaces.GPU
+def t2t(input):
+    return t2t_gen_model(input)
+
+@spaces.GPU
+def pos(input, boun=True):
+    if boun==True:
+        return pos_boun(input)
+    else:
+        return pos_imst(input)
+
+@spaces.GPU
+def ner(input, wikiann=True):
+    if wikiann==True:
+        return ner_wikiann(input)
+    else:
+        return ner_model(input)
+
+
+@spaces.GPU
+def paraphrase(input, model_choice="turna_paraphrasing_tatoeba"):
+    if model_choice=="turna_paraphrasing_tatoeba":
+        return paraphrasing(input)
+    else:
+        return paraphrasing_sub(input)
+        
+@spaces.GPU    
+def summarize(input, model_choice="turna_summarization_tr_news"):
+    if model_choice=="turna_summarization_tr_news":
+        return news_sum(input)
+    else:
+        return summarization_model(input)
 
 
-interface_list = ["t2t_gen","summarization", "news_sum", "paraphrase", "paraphrasing_sub", "ttc",
-                 "product_reviews", "title_gen", "sentiment", "pos", "nli", "pos_boun",
-                 "stsb", "ner", "ner_wikiann"]
 
 with gr.Blocks(theme="shivi/calm_seafoam") as demo:
     gr.Markdown("# TURNA 🐦")
     gr.Markdown(DESCRIPTION)
-    gradio.TabbedInterface(interface_list)
-
+    with gr.Tab("Summarization"):
+        with gr.Markdown("TURNA fine-tuned on ummarization. Enter a news to summarize and pick the model.")
+        with gr.Column():
+            with gr.Row():
+                sum_choice = gr.Radio(choices = ["turna_summarization_mlsum", "turna_summarization_tr_news"])
+                sum_input = gr.Text()
+                sum_output = gr.Text()
+                sum_submit = gr.Button()
+                sum_submit.click(summarize, inputs=[sum_input, sum_choice], outputs=sum_output)
+                examples = gr.Examples(examples = long_text, inputs = [sum_input, sum_choice], outputs=sum_output, cache_examples=True, fn = summarize=True)
 demo.launch()