Create app.py

app.py

@@ -0,0 +1,82 @@
+import streamlit as st
+from datasets import load_dataset
+from transformers import AutoTokenizer, AutoModelForSequenceClassification, pipeline
+import torch
+import pandas as pd
+
+def load_orca_dataset():
+    st.info("Loading dataset... This may take a while.")
+    return load_dataset("microsoft/orca-agentinstruct-1M-v1")
+
+@st.cache_data
+def load_model_and_tokenizer(model_name):
+    tokenizer = AutoTokenizer.from_pretrained(model_name)
+    model = AutoModelForSequenceClassification.from_pretrained(model_name)
+    return tokenizer, model
+
+def evaluate_model(ds, tokenizer, model, max_samples, text_field):
+    st.info("Evaluating the model...")
+    classifier = pipeline("text-classification", model=model, tokenizer=tokenizer, device=0 if torch.cuda.is_available() else -1)
+
+    results = []
+    for i, example in enumerate(ds):
+        if i >= max_samples:
+            break
+        input_text = example[text_field]
+        result = classifier(input_text)[0]
+        results.append({"input": input_text, "label": result["label"], "score": result["score"]})
+    return results
+
+def main():
+    st.title("Orca Dataset Browser and Model Evaluator")
+
+    st.sidebar.header("Configuration")
+    load_dataset_btn = st.sidebar.button("Load Dataset")
+
+    if load_dataset_btn:
+        dataset = load_orca_dataset()
+        st.session_state["dataset"] = dataset
+
+    if "dataset" in st.session_state:
+        dataset = st.session_state["dataset"]
+
+        # List available splits
+        available_splits = list(dataset.keys())
+        st.sidebar.subheader("Available Dataset Splits")
+        selected_split = st.sidebar.selectbox("Select Split", available_splits)
+
+        st.subheader("Dataset Explorer")
+        st.write(f"Displaying information for split: `{selected_split}`")
+        st.write(dataset[selected_split].info)
+
+        # Determine available fields
+        sample_entry = dataset[selected_split][0]
+        st.sidebar.subheader("Available Fields in Dataset")
+        available_fields = list(sample_entry.keys())
+        st.sidebar.write(available_fields)
+        text_field = st.sidebar.selectbox("Select Text Field", available_fields)
+
+        sample_size = st.slider("Number of Samples to Display", min_value=1, max_value=20, value=5)
+        st.write(dataset[selected_split].shuffle(seed=42).select(range(sample_size)))
+
+        st.subheader("Model Evaluator")
+        model_name = st.text_input("Enter Hugging Face Model Name", value="distilbert-base-uncased-finetuned-sst-2-english")
+        max_samples = st.number_input("Number of Samples to Evaluate", min_value=1, max_value=100, value=10)
+
+        if st.button("Load Model and Evaluate"):
+            tokenizer, model = load_model_and_tokenizer(model_name)
+
+            results = evaluate_model(dataset[selected_split].shuffle(seed=42).select(range(max_samples)), tokenizer, model, max_samples, text_field)
+
+            st.subheader("Evaluation Results")
+            st.write(results)
+
+            st.download_button(
+                label="Download Results as CSV",
+                data=pd.DataFrame(results).to_csv(index=False),
+                file_name="evaluation_results.csv",
+                mime="text/csv",
+            )
+
+if __name__ == "__main__":
+    main()