feat(): Initial app commit

app.py

@@ -0,0 +1,88 @@
+import requests
+import json
+from transformers import AutoTokenizer, AutoModelForSequenceClassification
+import plotly.graph_objects as go
+import plotly.express as px
+import pandas as pd
+from sklearn.metrics import confusion_matrix
+from datasets import load_dataset
+
+
+def load_model(endpoint: str):
+  tokenizer = AutoTokenizer.from_pretrained(endpoint)
+  model = AutoModelForSequenceClassification.from_pretrained(endpoint)
+  return tokenizer, model
+
+
+def test_model(tokenizer, model, test_data: list, label_map: dict):
+  results = []
+  for text, true_label in test_data:
+      inputs = tokenizer(text, return_tensors="pt",
+                        truncation=True, padding=True)
+      outputs = model(**inputs)
+      pred_label = label_map[int(outputs.logits.argmax(dim=-1))]
+      results.append((text, true_label, pred_label))
+  return results
+
+
+def generate_report_card(results, label_map):
+  true_labels = [r[1] for r in results]
+  pred_labels = [r[2] for r in results]
+
+  cm = confusion_matrix(true_labels, pred_labels,
+                        labels=list(label_map.values()))
+
+  fig = go.Figure(
+      data=go.Heatmap(
+          z=cm,
+          x=list(label_map.values()),
+          y=list(label_map.values()),
+          colorscale='Viridis',
+          colorbar=dict(title='Number of Samples')
+      ),
+      layout=go.Layout(
+          title='Confusion Matrix',
+          xaxis=dict(title='Predicted Labels'),
+          yaxis=dict(title='True Labels', autorange='reversed')
+      )
+  )
+
+  fig.show()
+
+
+def load_sst2_data(split="test"):
+  dataset = load_dataset("glue", "sst2", split=split)
+  data = [(item["sentence"], "positive" if item["label"] == 1 else "negative")
+          for item in dataset]
+  return data
+
+
+# Define your model endpoint and label map
+# model_endpoint = "your-model-endpoint"
+
+# Modify this according to your model's labels
+# label_map = {0: "label0", 1: "label1"}
+
+model_endpoint = "distilbert-base-uncased-finetuned-sst-2-english"
+label_map = {0: "negative", 1: "positive"}
+
+# Load the model and tokenizer
+tokenizer, model = load_model(model_endpoint)
+
+# Prepare your test data (list of tuples containing text and true label)
+#test_data = [
+#    ("Sample text 1", "label0"),
+#    ("Sample text 2", "label1"),
+#   # Add more test samples here
+#]
+
+# Load the test data from the SST-2 dataset
+test_data = load_sst2_data()
+# Use a smaller subset of test_data for a quicker demonstration (optional)
+test_data = test_data[:100]
+
+# Test the model and generate results
+results = test_model(tokenizer, model, test_data, label_map)
+
+# Generate the visual report card
+generate_report_card(results, label_map)