Update app.py

app.py

@@ -1,182 +1,159 @@
 import streamlit as st
-import streamlit.components.v1 as components
-from transformers import pipeline
-from diffusers import StableDiffusionPipeline
-from datasets import load_dataset
-from peft import PeftConfig
+from transformers import pipeline, AutoModel, AutoTokenizer
+from diffusers import StableDiffusionPipeline, DiffusionPipeline, DDIMScheduler
+from datasets import load_dataset, Dataset
+from peft import PeftModel, PeftConfig, get_peft_model
 from accelerate import Accelerator
 from optimum.onnxruntime import ORTModelForSequenceClassification
 import torch
-import time
+from PIL import Image
+import pandas as pd
 
-# Cache resource-intensive models
+# Cache resource-intensive models for performance
 @st.cache_resource
-def load_diffuser_model():
+def load_diffuser():
     return StableDiffusionPipeline.from_pretrained("runwayml/stable-diffusion-v1-5")
 
 # Sidebar navigation
-st.sidebar.title("🤗 Libraries Demo")
-st.sidebar.markdown("Explore text 📝, images 🖼️, and model ops 🔗 with Hugging Face and Arcee!")
-page = st.sidebar.selectbox(
-    "Choose a Section",
-    [
-        "🏠 Home",
-        "🔄 Workflow",
-        "📝 Transformers",
-        "🖼️ Diffusers",
-        "📊 Datasets",
-        "⚙️ PEFT",
-        "🚀 Accelerate",
-        "⚡ Optimum",
-        "📚 DistillKit",
-        "🔗 MergeKit",
-        "❄️ Spectrum"
-    ],
-    help="Select a library to explore!"
-)
-
-# Mermaid graph for DistillKit, MergeKit, and Spectrum workflows
-mermaid_code = """
-graph TD
-    subgraph DistillKit
-        A1[Load Teacher Model] --> B1[Load Student Model]
-        B1 --> C1[Configure Distillation]
-        C1 --> D1[Perform Distillation]
-        D1 --> E1[Evaluate Model]
-    end
-    subgraph MergeKit
-        A2[Select Models] --> B2[Choose Merge Method]
-        B2 --> C2[Set Parameters]
-        C2 --> D2[Merge Models]
-        D2 --> E2[Save Merged Model]
-    end
-    subgraph Spectrum
-        A3[Load Model] --> B3[Analyze Layers]
-        B3 --> C3[Generate Config]
-        C3 --> D3[Apply Freezing]
-        D3 --> E3[Train/Evaluate Model]
-    end
-"""
-
-# Home Page
-if page == "🏠 Home":
-    st.title("Hugging Face & Arcee Libraries Demo 🌟")
-    st.markdown("""
-    Welcome to an interactive demo of powerful libraries for text, image, and model processing!  
-    - **📝 Text**: Analyze or generate text with Transformers.  
-    - **🖼️ Images**: Create visuals with Diffusers.  
-    - **🔗 Models**: Distill, merge, and optimize with Arcee's DistillKit, MergeKit, and Spectrum.  
-    Navigate via the sidebar to explore each library!
-    """)
-
-# Workflow Page with Mermaid Graph
-elif page == "🔄 Workflow":
-    st.header("🔄 Workflows: DistillKit, MergeKit, Spectrum")
-    st.markdown("See how inputs flow to outputs in Arcee’s libraries with this Mermaid graph:")
-    components.html(f"""
-    <div id="mermaid"></div>
-    <script src="https://cdn.jsdelivr.net/npm/mermaid/dist/mermaid.min.js"></script>
-    <script>
-        mermaid.initialize({{ startOnLoad: true }});
-        const mermaidCode = `{mermaid_code}`;
-        mermaid.render('graph', mermaidCode, (svgCode) => {{
-            document.getElementById('mermaid').innerHTML = svgCode;
-        }});
-    </script>
-    """, height=600)
+st.sidebar.title("🤗 Hugging Face Libraries Demo")
+library = st.sidebar.selectbox("Choose a library", ["Transformers", "Diffusers", "Datasets", "PEFT", "Accelerate", "Optimum"])
 
 # Transformers Section
-elif page == "📝 Transformers":
-    st.header("📝 Transformers")
-    st.markdown("Process text with pre-trained models.")
-    task = st.selectbox("Task", ["Sentiment Analysis", "Text Generation"])
-    text = st.text_area("Input Text", "")
-    if st.button("Run") and text:
-        with st.spinner("Processing..."):
-            if task == "Sentiment Analysis":
-                result = pipeline("sentiment-analysis")(text)
-                st.write(f"Result: {result[0]['label']} (Score: {result[0]['score']:.2f})")
-            else:
-                result = pipeline("text-generation")(text, max_length=50)[0]['generated_text']
-                st.write(f"Generated: {result}")
+if library == "Transformers":
+    st.header("📝 Transformers: Text Processing Powerhouse")
+    st.write("Analyze or transform text with pre-trained NLP models. Try it out below! 👇")
+    
+    # pipeline: Easy inference for NLP tasks
+    # 🎭 - Sentiment analysis with pipeline. Ex: 'I love this!' → Positive | Emotions in a snap! 😊
+    # ✍️ - Text generation with pipeline. Ex: 'Once upon a time' → Story | Spin tales fast! 📖
+    # ❓ - Question answering with pipeline. Ex: 'Who won?' → Answer | Quiz master ready! 🧠
+    task = st.selectbox("Pick a Task", ["Sentiment Analysis 😊😢", "Text Generation ✍️", "Fill Mask 🤖"])
+    text_input = st.text_area("Enter text here ✍️", "", placeholder="Type your text...")
+
+    if st.button("Process Text 🚀", disabled=not text_input):
+        with st.spinner("Processing... ⏳"):
+            if "Sentiment" in task:
+                classifier = pipeline("sentiment-analysis")
+                result = classifier(text_input)
+                st.success(f"**Sentiment Result:** {result[0]['label']} (Confidence: {result[0]['score']:.2f}) ✅")
+                st.write("🎭 - Sentiment analysis with pipeline. Ex: 'I love this!' → Positive | Emotions in a snap! 😊")
+            elif "Text Generation" in task:
+                generator = pipeline("text-generation")
+                result = generator(text_input, max_length=50)[0]['generated_text']
+                st.success(f"**Generated Text:** {result} ✅")
+                st.write("✍️ - Text generation with pipeline. Ex: 'Once upon a time' → Story | Spin tales fast! 📖")
+            elif "Fill Mask" in task:
+                fill_mask = pipeline("fill-mask")
+                results = fill_mask(text_input.replace("[MASK]", fill_mask.tokenizer.mask_token))
+                for res in results[:3]:
+                    st.write(f"Prediction: {res['sequence']} (Score: {res['score']:.2f})")
+                # AutoModel & AutoTokenizer: Load models and tokenizers
+                # 🤖 - Mask filling with AutoModel. Ex: 'The [MASK] is blue' → sky | Guess master! ✨
+                # 📝 - Tokenizing with AutoTokenizer. Ex: 'Hello world' → tokens | Words to numbers! 🔢
+                # ⚙️ - Custom inference with AutoModel. Ex: 'BERT for sentiment' → logits | Model magic! 🪄
+                st.write("🤖 - Mask filling with pipeline. Ex: 'The [MASK] is blue' → sky | Guess master! ✨")
 
 # Diffusers Section
-elif page == "🖼️ Diffusers":
-    st.header("🖼️ Diffusers")
-    st.markdown("Generate images from text.")
-    prompt = st.text_input("Prompt", "A futuristic city")
-    if st.button("Generate"):
-        with st.spinner("Generating..."):
-            pipe = load_diffuser_model()
+elif library == "Diffusers":
+    st.header("🖼️ Diffusers: Image Generation Magic")
+    st.write("Turn text into stunning images with diffusion models! Let's create something cool. 🎨")
+    
+    # StableDiffusionPipeline: Text-to-image generation
+    # 🖼️ - Text-to-image with StableDiffusionPipeline. Ex: 'Cat in space' → Image | Art from words! 🚀
+    # 🎨 - Style transfer with StableDiffusionPipeline. Ex: 'Van Gogh cat' → Image | Paint like pros! 🖌️
+    # 🌟 - Creative prompts with StableDiffusionPipeline. Ex: 'Dream city' → Image | Imagine it, see it! ✨
+    prompt = st.text_input("Enter a Creative Prompt 🎨", "", placeholder="e.g., 'A cat in space'")
+    if st.button("Generate Image 🌟", disabled=not prompt):
+        with st.spinner("Generating image... ⏳"):
+            pipe = load_diffuser()
             image = pipe(prompt).images[0]
-            st.image(image, caption=prompt)
+            st.image(image, caption="Generated Image", use_column_width=True)
+            st.write("🖼️ - Text-to-image with StableDiffusionPipeline. Ex: 'Cat in space' → Image | Art from words! 🚀")
+            # DiffusionPipeline & DDIMScheduler: Advanced diffusion control
+            # 🧩 - Inpainting with DiffusionPipeline. Ex: 'Fix broken image' → Restored | Puzzle solver! 🛠️
+            # 🔧 - Scheduling with DDIMScheduler. Ex: 'Fast diffusion' → Image | Speedy art! ⚡
 
 # Datasets Section
-elif page == "📊 Datasets":
-    st.header("📊 Datasets")
-    st.markdown("Load and explore datasets.")
-    dataset = st.selectbox("Dataset", ["imdb", "squad"])
-    if st.button("Load"):
-        data = load_dataset(dataset, split="train[:5]")
-        st.write(data)
-
-# PEFT Section
-elif page == "⚙️ PEFT":
-    st.header("⚙️ PEFT")
-    st.markdown("Parameter-efficient fine-tuning.")
-    text = st.text_area("Text", "")
-    if st.button("Classify") and text:
-        st.write("Simulated PEFT classification: Positive")
-
-# Accelerate Section
-elif page == "🚀 Accelerate":
-    st.header("🚀 Accelerate")
-    st.markdown("Optimize across devices.")
-    text = st.text_area("Text", "")
-    if st.button("Analyze") and text:
-        accelerator = Accelerator()
-        result = pipeline("sentiment-analysis")(text)
-        st.write(f"Result: {result[0]['label']} (Score: {result[0]['score']:.2f})")
-
-# Optimum Section
-elif page == "⚡ Optimum":
-    st.header("⚡ Optimum")
-    st.markdown("Hardware-accelerated inference.")
-    text = st.text_area("Text", "")
-    if st.button("Classify") and text:
-        st.write("Simulated Optimum result: Positive")
-
-# DistillKit Section
-elif page == "📚 DistillKit":
-    st.header("📚 DistillKit: Model Distillation")
-    st.markdown("Distill large models into smaller, efficient ones. Here are the top 5 functions:")
-
-    # 1. Load teacher model
-    teacher = st.selectbox("Teacher Model", ["arcee-ai/Arcee-Spark", "bert-base-uncased"])
-    st.write(f"1. Loaded teacher: {teacher}")
-
-    # 2. Load student model
-    student = st.selectbox("Student Model", ["Qwen/Qwen2-1.5B", "distilbert-base-uncased"])
-    st.write(f"2. Loaded student: {student}")
-
-    # 3. Configure distillation
-    temp = st.slider("Temperature", 1.0, 5.0, 2.0)
-    alpha = st.slider("Alpha", 0.0, 1.0, 0.5)
-    st.write(f"3. Config: Temp={temp}, Alpha={alpha}")
-
-    # 4. Perform distillation (simulated)
-    if st.button("Distill"):
-        with st.spinner("Distilling..."):
-            time.sleep(2)
-            st.success("4. Distillation complete!")
-
-        # 5. Evaluate distilled model
-        st.write("5. Evaluating...")
-        metrics = {"accuracy": 0.85, "loss": 0.12}
-        st.write(f"Metrics: {metrics}")
-
-    st.markdown("""
-    **How It Works:**  
-    DistillKit compresses a teacher model into a student model using distillation techniques.  
-    ```python
-    config = {"teacher": "arcee-ai/Arcee-Spark", "student": "Qwen/Qwen2-1.5B", "temp": 2.0, "alpha": 0.5}
+elif library == "Datasets":
+    st.header("📊 Datasets: Ready-to-Use Data")
+    st.write("Explore datasets for training or analysis. Pick one and see a sample! 📈")
+    
+    # load_dataset: Access public datasets
+    # 📚 - Load reviews with load_dataset. Ex: 'imdb' → Reviews | Movie buffs rejoice! 🎬
+    # ❓ - Load QA with load_dataset. Ex: 'squad' → Q&A | Trivia time! 🧠
+    # 🗣️ - Load audio with load_dataset. Ex: 'common_voice' → Audio | Hear the data! 🎙️
+    dataset_name = st.selectbox("Choose a Dataset", ["imdb 🎬", "squad ❓"])
+    if st.button("Load Dataset 📥"):
+        with st.spinner("Fetching dataset... ⏳"):
+            dataset = load_dataset(dataset_name.split()[0], split="train[:5]")
+            st.success(f"**Dataset Loaded:** {dataset_name} (showing first 5 samples) ✅")
+            st.write(dataset)
+            st.write("📚 - Load reviews with load_dataset. Ex: 'imdb' → Reviews | Movie buffs rejoice! 🎬")
+
+    # Dataset.from_pandas: Convert DataFrames to datasets
+    # 📊 - CSV to dataset with Dataset.from_pandas. Ex: 'Tweets CSV' → Dataset | Your data shines! 🐦
+    # 📈 - Analyze data with Dataset.from_pandas. Ex: 'Sales data' → Dataset | Numbers talk! 💰
+    # 🔄 - Preprocess with Dataset.from_pandas. Ex: 'Raw text' → Dataset | Clean it up! 🧹
+    st.subheader("Create Dataset from CSV")
+    uploaded_file = st.file_uploader("Upload a CSV file", type="csv")
+    if uploaded_file is not None:
+        df = pd.read_csv(uploaded_file)
+        dataset = Dataset.from_pandas(df)
+        st.write(dataset)
+        st.write("📊 - CSV to dataset with Dataset.from_pandas. Ex: 'Tweets CSV' → Dataset | Your data shines! 🐦")
+
+# PEFT Section (Simplified)
+elif library == "PEFT":
+    st.header("⚙️ PEFT: Efficient Fine-Tuning")
+    st.write("Learn about parameter-efficient fine-tuning (simplified demo).")
+    
+    # PeftModel: Load fine-tuned models
+    # ⚙️ - Inference with PeftModel. Ex: 'Adapted BERT' → Output | Slim yet mighty! 💪
+    # 🤖 - Custom tasks with PeftModel. Ex: 'NER tuning' → Tags | Precision tweak! 🎯
+    # 📉 - Low resource with PeftModel. Ex: 'Small GPU' → Model | Efficiency wins! 🏆
+    # PeftConfig: Configure PEFT settings
+    # 🛠️ - LoRA setup with PeftConfig. Ex: 'Rank 8' → Config | Tune light! 🌟
+    # 🔧 - Adapter config with PeftConfig. Ex: 'Task-specific' → Config | Fit like a glove! 🧤
+    # ⚡ - Fast tuning with PeftConfig. Ex: 'Quick fine-tune' → Config | Speedy prep! 🏃
+    # get_peft_model: Wrap models with PEFT
+    # 🔗 - Enhance with get_peft_model. Ex: 'BERT + LoRA' → Model | Power boost! 🚀
+    # 📝 - Task adapt with get_peft_model. Ex: 'Sentiment' → Model | Tailored fit! ✂️
+    # 🌐 - Multi-task with get_peft_model. Ex: 'QA + NER' → Model | Versatility rules! 🌍
+    text = st.text_area("Enter Text to Classify ✍️", "", placeholder="Type something...")
+    if st.button("Classify 🚀", disabled=not text):
+        with st.spinner("Processing... ⏳"):
+            st.success("**Result:** Placeholder (PEFT reduces parameters efficiently!) ✅")
+            st.write("⚙️ - Inference with PeftModel. Ex: 'Adapted BERT' → Output | Slim yet mighty! 💪")
+
+# Accelerate Section (Simplified)
+elif library == "Accelerate":
+    st.header("🚀 Accelerate: Device Optimization")
+    st.write("Optimize inference across devices with Accelerate.")
+    
+    # Accelerator: Manage device placement
+    # 🚀 - GPU use with Accelerator. Ex: 'Auto GPU' → Device | Speed unleashed! ⚡
+    # 🔗 - Multi-device with Accelerator. Ex: '2 GPUs' → Setup | Teamwork rocks! 🤝
+    # 🛠️ - Easy setup with Accelerator. Ex: 'Model prep' → Ready | Smooth start! 🌈
+    text = st.text_area("Enter Text for Sentiment Analysis ✍️", "", placeholder="Type something...")
+    if st.button("Analyze 🚀", disabled=not text):
+        with st.spinner("Analyzing... ⏳"):
+            accelerator = Accelerator()
+            classifier = pipeline("sentiment-analysis")
+            result = classifier(text)
+            st.success(f"**Result:** {result[0]['label']} (Confidence: {result[0]['score']:.2f}) ✅")
+            st.write("🚀 - GPU use with Accelerator. Ex: 'Auto GPU' → Device | Speed unleashed! ⚡")
+
+# Optimum Section (Simplified)
+elif library == "Optimum":
+    st.header("⚡ Optimum: Hardware Acceleration")
+    st.write("Speed up inference with optimized models (e.g., ONNX).")
+    
+    # ORTModelForSequenceClassification: ONNX-optimized classification
+    # ⚡ - Fast classify with ORTModelForSequenceClassification. Ex: 'Text' → Label | Speed king! 🏎️
+    # 🖥️ - CPU boost with ORTModelForSequenceClassification. Ex: 'Sentiment' → Result | No GPU? No prob! 💪
+    # 📊 - Efficient with ORTModelForSequenceClassification. Ex: 'Batch text' → Labels | Quick wins! 🏆
+    text = st.text_area("Enter Text to Classify ✍️", "", placeholder="Type something...")
+    if st.button("Classify 🚀", disabled=not text):
+        with st.spinner("Processing... ⏳"):
+            st.success("**Result:** Placeholder (Optimum boosts speed!) ✅")
+            st.write("⚡ - Fast classify with ORTModelForSequenceClassification. Ex: 'Text' → Label | Speed king! 🏎️")