Create app.py

app.py

@@ -0,0 +1,71 @@
+# Run the script and open the link in the browser.
+
+import gradio as gr
+import os
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
+
+# finetuning with original tokenizer
+CHECKPOINT_PATH= '/work/pnrr_itserr/latin_model/cineca_checkpoints/finetuning-gpt2_2-nodes_tokenizer-original_lr-6e-4_fcocchi'
+# finetuning with our embedding and tokenizer
+##CHECKPOINT_PATH= '/work/pnrr_itserr/latin_model/cineca_checkpoints/finetuning-gpt2_2-nodes_tokenizer-our-embedding_lr-6e-4_fcocchi'
+# scratch with original tokenizer
+CHECKPOINT_PATH= '/work/pnrr_itserr/latin_model/cineca_checkpoints/scratch_2-nodes_tokenizer-original_packing_fcocchi'
+# scratch with latbert tokenizer
+CHECKPOINT_PATH= '/work/pnrr_itserr/latin_model/cineca_checkpoints/scratch_2-nodes_tokenizer_latbert-original_packing_fcocchi'
+print(f"Loading model from: {CHECKPOINT_PATH}")
+
+tokenizer = AutoTokenizer.from_pretrained(CHECKPOINT_PATH)
+model = AutoModelForCausalLM.from_pretrained(CHECKPOINT_PATH)
+
+description="""
+This is a Latin Language Model (LLM) based on GPT-2 and it was trained on a large corpus of Latin texts and can generate text in Latin. 
+Please enter a prompt in Latin to generate text.
+"""
+title= "(L<sup>3</sup>) - Latin Large Language Model"
+article= "hello world ..."
+examples= ['Accidere ex una scintilla', 'Audacter calumniare,', 'Consolatium misero comites']
+logo_image= '/homes/fcocchi/itserr/itserr_latin_llm/ITSERR_row_logo.png'
+
+def generate_text(prompt):
+    if torch.cuda.is_available(): device = torch.device("cuda")      
+    else: 
+        device = torch.device("cpu")
+        print("No GPU available")
+    
+    print("***** Generate *****")
+    text_generator = pipeline("text-generation", model=model, tokenizer=tokenizer, device=device)
+    #generated_text = text_generator(prompt, max_length=100)
+    generated_text = text_generator(prompt, max_length=50, do_sample=True, temperature=1.0, repetition_penalty=2.0, truncation=True)
+    return generated_text[0]['generated_text']
+
+custom_css = """
+#logo {
+    display: block;
+    margin-left: auto;
+    margin-right: auto;
+    width: 512px;
+    height: 256px;
+}
+"""
+
+with gr.Blocks(css=custom_css) as demo:
+    gr.Image(logo_image, elem_id="logo")
+    gr.Markdown(f"<h1 style='text-align: center;'>{title}</h1>")
+    gr.Markdown(description)
+    
+    with gr.Row():
+        with gr.Column():
+            input_text = gr.Textbox(lines=5, placeholder="Enter latin text here...", label="Input Text")
+        with gr.Column():
+            output_text = gr.Textbox(lines=5, placeholder="Output text will appear here...", label="Output Text")
+    
+    clean_button = gr.Button("Generate Text")
+    clean_button.click(fn=generate_text, inputs=input_text, outputs=output_text)
+    
+    gr.Examples(examples=examples, inputs=input_text)
+    gr.Markdown(article)
+
+demo.launch(share=True)
+# iface = gr.Interface(fn=generate_text, inputs="text", outputs="text", 
+#                      description=description, title=title, examples=examples, article=article)