Upload folder using huggingface_hub

app.py

@@ -0,0 +1,137 @@
+import streamlit as st
+import torch
+import os
+import time
+import pandas as pd
+from transformers import AutoTokenizer, AutoModelForCausalLM
+from peft import PeftModel
+from huggingface_hub import HfApi, Table
+
+# CONFIGURATION
+BASE_MODEL = "microsoft/Phi-3-mini-4k-instruct"
+ADAPTER_REPO = "aniketp2009gmail/phi3-bilora-code-review"
+FEEDBACK_DATASET = "aniketp2009gmail/code-review-feedback"
+
+st.set_page_config(page_title="BiLoRA Code Assistant", page_icon="🧠")
+
+# Load the base model and adapters from Hub
+@st.cache_resource(show_spinner="Loading model (this takes ~2 mins on CPU)...")
+def load_model():
+    tokenizer = AutoTokenizer.from_pretrained(ADAPTER_REPO)
+    if tokenizer.pad_token is None:
+        tokenizer.pad_token = tokenizer.eos_token
+    tokenizer.padding_side = "right"
+
+    model = AutoModelForCausalLM.from_pretrained(
+        BASE_MODEL,
+        torch_dtype=torch.float16,
+        low_cpu_mem_usage=True,
+        trust_remote_code=False,
+        attn_implementation="eager"
+    )
+    
+    model = PeftModel.from_pretrained(model, ADAPTER_REPO, subfolder="task_1", adapter_name="task_1")
+    model.load_adapter(ADAPTER_REPO, subfolder="task_2", adapter_name="task_2")
+    model.set_adapter("task_1")
+    return model, tokenizer
+
+import datetime
+
+def log_feedback(prompt, response, task, rating):
+    """Save feedback to a Hugging Face Dataset CSV"""
+    try:
+        hf_token = st.secrets.get("HF_TOKEN") or os.environ.get("HF_TOKEN")
+        if not hf_token:
+            st.error("HF_TOKEN not found. Feedback not saved.")
+            return
+        
+        # Create a dictionary for the new entry
+        new_data = {
+            "timestamp": [str(datetime.datetime.now())],
+            "task": [task],
+            "prompt": [prompt],
+            "response": [response],
+            "rating": [rating]
+        }
+        df = pd.DataFrame(new_data)
+        
+        # In a real Space, we append to a CSV in the Dataset repo
+        # Using HfApi to upload/append
+        api = HfApi(token=hf_token)
+        csv_data = df.to_csv(index=False, header=False)
+        
+        # Path to feedback file in your dataset repo
+        repo_id = "aniketp2009gmail/bilora-user-feedback"
+        
+        # Create repo if it doesn't exist
+        api.create_repo(repo_id=repo_id, repo_type="dataset", exist_ok=True)
+        
+        # Upload as a new file named by day to prevent massive file conflicts
+        filename = f"feedback_{datetime.date.today()}.csv"
+        
+        # For simplicity in Streamlit, we'll use upload_file. 
+        # In high-traffic apps, you'd use a more robust queue.
+        api.upload_file(
+            path_or_fileobj=csv_data.encode(),
+            path_in_repo=filename,
+            repo_id=repo_id,
+            repo_type="dataset",
+            run_as_future=True
+        )
+        st.session_state.feedback_sent = True
+    except Exception as e:
+        st.error(f"Feedback error: {e}")
+
+# UI
+st.title("🧠 BiLoRA: Dual-Adapter Code Assistant")
+
+try:
+    model, tokenizer = load_model()
+
+    task_option = st.sidebar.radio(
+        "Select Task:",
+        options=["task_1", "task_2"],
+        format_func=lambda x: "Code Generation" if x == "task_1" else "Docstring Generation"
+    )
+
+    user_input = st.text_area("Input:", height=150)
+
+    if st.button("Generate"):
+        if user_input.strip():
+            with st.spinner("Generating..."):
+                model.set_adapter(task_option)
+                prefix = "Generate code: " if task_option == "task_1" else "Generate docstring: "
+                suffix = "\nCode: " if task_option == "task_1" else "\nDocstring: "
+                
+                full_prompt = f"{prefix}{user_input}{suffix}"
+                inputs = tokenizer(full_prompt, return_tensors="pt")
+                
+                with torch.no_grad():
+                    outputs = model.generate(**inputs, max_new_tokens=200, do_sample=True, temperature=0.7)
+                
+                input_length = inputs['input_ids'].shape[1]
+                output_text = tokenizer.decode(outputs[0][input_length:], skip_special_tokens=True)
+                
+                st.session_state.last_prompt = user_input
+                st.session_state.last_response = output_text
+                st.session_state.last_task = task_option
+                st.session_state.generated = True
+
+    if st.session_state.get("generated"):
+        st.subheader("Result:")
+        st.code(st.session_state.last_response, language="python")
+        
+        st.write("---")
+        st.write("Help improve BiLoRA! Was this helpful?")
+        col1, col2 = st.columns(5)
+        with col1:
+            if st.button("👍 Yes"):
+                log_feedback(st.session_state.last_prompt, st.session_state.last_response, st.session_state.last_task, 1)
+                st.success("Thanks!")
+        with col2:
+            if st.button("👎 No"):
+                log_feedback(st.session_state.last_prompt, st.session_state.last_response, st.session_state.last_task, 0)
+                st.error("Feedback received.")
+
+except Exception as e:
+    st.error(f"Error: {e}")

requirements.txt

@@ -1,3 +1,9 @@
-altair
+streamlit
+torch
+transformers
+peft
+accelerate
+sentencepiece
+protobuf
+huggingface_hub
 pandas
-streamlit