Update app.py

app.py

@@ -5,8 +5,8 @@ import torch
 import re
 
 # Initialize the model
-model = AutoModelForCausalLM.from_pretrained("Detsutut/Igea-1B-v0.0.1-Q4_K_M-GGUF", model_file="igea-1b-v0.0.1-q4_k_m.gguf", model_type="mistral", hf=True)
-tokenizer = AutoTokenizer.from_pretrained( "Detsutut/Igea-1B-v0.0.1")
+model = AutoModelForCausalLM.from_pretrained("Detsutut/Igea-1B-instruct-GGUF-Q4", model_file="unsloth.Q4_K_M.gguf", model_type="mistral", hf=True)
+tokenizer = AutoTokenizer.from_pretrained( "Detsutut/Igea-1B-instruct")
 
 
 gen_pipeline = pipeline(
@@ -15,11 +15,23 @@ gen_pipeline = pipeline(
     tokenizer=tokenizer
 )
 
+alpaca_instruct_prompt = """
+{}
+
+### Istruzione:
+{}
+
+### Risposta:
+{}"""
+
 # Define the function to generate text
-def generate_text(input_text, max_new_tokens=30, temperature=1, top_p=0.95, split_output=False):
-    if split_output:
-        max_new_tokens=30
-        top_p=0.95
+def generate_text(input_text, max_new_tokens=30, temperature=1, top_p=0.95):
+
+    prompt = alpaca_instruct_prompt.format("Di seguito è riportata un'istruzione che descrive un compito. Scrivi una risposta che completi in modo appropriato la richiesta.", 
+                                           input_text, 
+                                           ""
+    )
+    
     output = gen_pipeline(
         input_text,
         max_new_tokens=max_new_tokens,
@@ -28,34 +40,32 @@ def generate_text(input_text, max_new_tokens=30, temperature=1, top_p=0.95, spli
         return_full_text = False
     )
     generated_text = output[0]['generated_text']
-    if split_output:
-        sentences = re.split('(?<!\w\.\w.)(?<![A-Z][a-z]\.)(?<=\.|\?)\s', generated_text)
-        if sentences:
-            generated_text = sentences[0]
+    
+    if generated_text[-1] not in [".","!","?","\n"]:
+        generated_text = generated_text + " [...]"
+    
     return f"<span>{input_text}</span><b style='color: blue;'>{generated_text}</b>"
 
 # Create the Gradio interface
-input_text = gr.Textbox(lines=2, placeholder="Enter your text here...", label="Input Text")
+input_text = gr.Textbox(lines=2, placeholder="Enter your request here...", label="Input Text")
 
 max_new_tokens = gr.Slider(minimum=1, maximum=200, value=30, step=1, label="Max New Tokens")
 temperature = gr.Slider(minimum=0.1, maximum=2.0, value=1.0, step=0.1, label="Temperature")
 top_p = gr.Slider(minimum=0.0, maximum=1.0, value=0.95, step=0.01, label="Top-p")
-split_output = gr.Checkbox(label="Quick single-sentence output", value=True)
 
 with gr.Blocks(css="#outbox { border-radius: 8px !important; border: 1px solid #e5e7eb !important; padding: 8px !important; text-align:center !important;}") as iface:
-    gr.Markdown("# Igea Text Generation Interface ⚕️🩺")
-    gr.Markdown("🐢💬 To guarantee a reasonable througput (<1 min to answer with default settings), this space employs a **GGUF quantized version of [Igea 1B](https://huggingface.co/bmi-labmedinfo/Igea-1B-v0.0.1)**, optimized for **hardware-limited, CPU-only machines** like the free-tier HuggingFace space.")
+    gr.Markdown("# Igea Instruct Interface ⚕️🩺")
+    gr.Markdown("🐢💬 To guarantee a reasonable througput (<1 min to answer with default settings), this space employs a **GGUF quantized version of [Igea 1B](https://huggingface.co/bmi-labmedinfo/Igea-1B-v0.0.1)**, optimized for **hardware-limited, CPU-only machines** like the free-tier HuggingFace space. Quantized models may result in significant performance degradation and therefore are not representative of the original model capabilities.")
     gr.Markdown("⚠️ Read the **[bias, risks and limitations](https://huggingface.co/bmi-labmedinfo/Igea-1B-v0.0.1#%F0%9F%9A%A8%E2%9A%A0%EF%B8%8F%F0%9F%9A%A8-bias-risks-and-limitations-%F0%9F%9A%A8%E2%9A%A0%EF%B8%8F%F0%9F%9A%A8)** of Igea before use!")
     input_text.render()
     with gr.Accordion("Advanced Options", open=False):
         max_new_tokens.render()
         temperature.render()
         top_p.render()
-        split_output.render()
     output = gr.HTML(label="Generated Text",elem_id="outbox")
 
     btn = gr.Button("Generate")
-    btn.click(generate_text, [input_text, max_new_tokens, temperature, top_p, split_output], output)
+    btn.click(generate_text, [input_text, max_new_tokens, temperature, top_p], output)
 
 # Launch the interface
 if __name__ == "__main__":