Subindo API Docker com modelo de embeddings

README.md

@@ -1,7 +1,7 @@
 ---
 title: Bible Embeddings Api
-emoji: 🐠
-colorFrom: indigo
+emoji: 📚
+colorFrom: red
 colorTo: purple
 sdk: docker
 pinned: false

app.py

@@ -0,0 +1,31 @@
+from fastapi import FastAPI, Request
+from transformers import AutoTokenizer, AutoModel
+import torch
+
+app = FastAPI()
+
+tokenizer = AutoTokenizer.from_pretrained("thiagocmach/paraphrase-pt-bible")
+model = AutoModel.from_pretrained("thiagocmach/paraphrase-pt-bible").eval()
+
+@app.get("/")
+def read_root():
+    return {"message": "API ativa 🙌"}
+
+@app.post("/embed")
+async def embed_text(request: Request):
+    data = await request.json()
+    texto = data.get('texto')
+    if not texto:
+        return {"error": "Campo 'texto' obrigatório"}
+
+    inputs = tokenizer(texto, return_tensors='pt', truncation=True, padding=True)
+    with torch.no_grad():
+        outputs = model(**inputs)
+        embeddings = outputs.last_hidden_state
+        mask = inputs['attention_mask'].unsqueeze(-1).expand(embeddings.size())
+        masked_embeddings = embeddings * mask
+        summed = torch.sum(masked_embeddings, dim=1)
+        counted = torch.clamp(mask.sum(1), min=1e-9)
+        mean_pooled = (summed / counted).squeeze().tolist()
+
+    return {"embedding": mean_pooled}

dockerfile

@@ -0,0 +1,9 @@
+FROM python:3.10-slim
+
+WORKDIR /app
+COPY . /app
+
+RUN pip install --upgrade pip && \
+    pip install --no-cache-dir -r requirements.txt
+
+CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]