Initial ModernFinBERT space setup

README.md

@@ -0,0 +1,19 @@
+# ModernFinBERT Space
+
+A lightweight Hugging Face Space that hosts [`tabularisai/ModernFinBERT`](https://huggingface.co/tabularisai/ModernFinBERT) for financial sentiment analysis.
+
+- **Model size**: 0.1B parameters (fits on CPU)
+- **Hardware**: CPU Basic (no GPU required)
+- **SDK**: Gradio
+
+## API usage
+
+The Gradio API is exposed automatically. You can call it from your backend with a simple HTTP `POST` to `/run/predict`.
+
+Example payload:
+```json
+{
+  "fn_index": 0,
+  "data": ["Revenue grew 15% YoY\nFed raises rates by 0.75bps"]
+}
+```

app.py

@@ -0,0 +1,61 @@
+import gradio as gr
+from transformers import pipeline
+
+print("Loading tabularisai/ModernFinBERT on CPU...")
+classifier = pipeline(
+    "text-classification",
+    model="tabularisai/ModernFinBERT",
+    device=-1,  # CPU only — 0.1B params fits comfortably
+)
+print("Model ready.")
+
+
+def predict_sentiment(text_block):
+    """
+    Accepts multiple lines of text, classifies each one.
+    Returns a JSON list of {label, score} dicts.
+    """
+    if not text_block:
+        return []
+
+    # Split by newline, strip, drop empties
+    texts = [t.strip() for t in text_block.splitlines() if t.strip()]
+    if not texts:
+        return []
+
+    # Batch inference
+    raw_results = classifier(texts, batch_size=32)
+
+    # Normalise output
+    results = [
+        {"label": r["label"], "score": float(r["score"])}
+        for r in raw_results
+    ]
+    return results
+
+
+with gr.Blocks(title="ModernFinBERT") as demo:
+    gr.Markdown("""
+    # ModernFinBERT Sentiment Analysis
+    Financial sentiment classifier powered by
+    [`tabularisai/ModernFinBERT`](https://huggingface.co/tabularisai/ModernFinBERT).
+    Runs on CPU — no GPU required.
+    """)
+
+    with gr.Row():
+        with gr.Column(scale=1):
+            input_box = gr.Textbox(
+                lines=10,
+                label="Input texts",
+                placeholder="Paste one headline / sentence per line...",
+            )
+            submit_btn = gr.Button("Analyze", variant="primary")
+        with gr.Column(scale=1):
+            output_json = gr.JSON(label="Results")
+
+    submit_btn.click(fn=predict_sentiment, inputs=input_box, outputs=output_json)
+
+    # Also expose a direct API at /run/predict (fn_index 0)
+
+if __name__ == "__main__":
+    demo.launch(server_name="0.0.0.0", server_port=7860)

requirements.txt

@@ -0,0 +1,4 @@
+gradio>=5.0,<6.0
+transformers>=4.36
+torch>=2.0
+numpy