Update app.py

app.py

@@ -1,28 +1,39 @@
 import streamlit as st
-import subprocess
 from transformers import AutoTokenizer, AutoModelForCausalLM
 import torch
-import re
 
-st.title('Eleanor Rigby')
+# Load the Phi 2 model and tokenizer
+tokenizer = AutoTokenizer.from_pretrained(
+    "microsoft/phi-2",
+    trust_remote_code=True
+)
 
-hf_token = st.secrets["hf_token"]
-inference_model = AutoModelForCausalLM.from_pretrained("arieridwans/phi_2-finetuned-lyrics", trust_remote_code=True, torch_dtype=torch.float32, token=hf_token)
-inference_tokenizer = AutoTokenizer.from_pretrained("microsoft/phi-2", use_fast=True)
-inference_tokenizer.pad_token=inference_tokenizer.eos_token
+model = AutoModelForCausalLM.from_pretrained(
+    "microsoft/phi-2",
+    device_map="auto",
+    trust_remote_code=True
+)
 
-user_prompt = st.text_area("Enter your prompt that can be song lyrics e.g. 'Yesterday, I saw you in my dream'")
+# Streamlit UI
+st.title("Eleanor Rigby")
 
-def run_inference():
+# User input prompt
+prompt = st.text_area("Enter your prompt:", """Write a story about Nasa""")
+
+# Generate output based on user input
+if st.button("Generate Output"):
     instruct_prompt = "Instruct:You are a song writer and your main reference is The Beatles. Write a song lyrics by completing these words:"
     output_prompt = "Output:"
-    input = inference_tokenizer(""" {0}{1}\n{2} """.format(instruct_prompt, user_prompt, output_prompt),
-                     return_tensors="pt",
-                     return_attention_mask=False,
-                     padding=True,
-                     truncation=True)
-    result = inference_model.generate(**input, repetition_penalty=1.2, max_length=1024)
-    output = inference_tokenizer.batch_decode(result, skip_special_tokens=True)[0]
-    st.text(output)
-    
-st.button('Generate Result', on_click=run_inference)
+    prompt = """ {0}{1}\n{2} """.format(instruct_prompt, user_prompt, output_prompt)
+    with torch.no_grad():
+        token_ids = tokenizer.encode(prompt, add_special_tokens=False, return_tensors="pt")
+        output_ids = model.generate(
+            token_ids.to(model.device),
+            max_new_tokens=512,
+            do_sample=True,
+            temperature=0.3
+        )
+
+    output = tokenizer.decode(output_ids[0][token_ids.size(1):])
+    st.text("Generated Output:")
+    st.write(output)