Upload 6 files

Dockerfile

@@ -0,0 +1,22 @@
+# Base image
+FROM python:3.9
+
+# Set working directory
+WORKDIR /app
+
+# Copy the application files
+COPY app.py .
+COPY MachineTranslation.py .
+COPY requirements.txt .
+COPY utils.py .
+COPY download_model.py .
+
+# Install dependencies
+RUN pip3 install -r requirements.txt
+RUN python3 download_model.py
+
+# Expose the port
+EXPOSE 8000
+
+# Run the application
+CMD ["python3", "app.py"]

MachineTranslation.py

@@ -0,0 +1,28 @@
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
+import time
+
+
+class MachineTranslation:
+    def __init__(self, name_model="facebook/nllb-200-distilled-600M"):
+        self.name_model = name_model
+
+        self.start_time_tokenizer = time.perf_counter()
+        self.tokenizer = AutoTokenizer.from_pretrained(
+            self.name_model)
+        self.end_time_tokenizer = time.perf_counter()
+        # load time tokenizer
+        self.time_tokenizer = self.end_time_tokenizer - self.start_time_tokenizer
+        print("load time tokenizer :", self.time_tokenizer)
+        self.start_time_nllb = time.perf_counter()
+        self.model = AutoModelForSeq2SeqLM.from_pretrained(
+            self.name_model)
+        self.end_time_nllb = time.perf_counter()
+        # load time nllb
+        self.time_nllb = self.end_time_nllb - self.start_time_nllb
+        print("load time nllb :", self.time_nllb)
+        
+
+    def predict(self, text, from_lang, to_lang):
+        classifier = pipeline("translation", model=self.model, tokenizer=self.tokenizer, 
+                                src_lang=from_lang, tgt_lang=to_lang, max_length=len(text))
+        return classifier(text)[0]["translation_text"]

app.py

@@ -0,0 +1,58 @@
+import uvicorn
+from pydantic import BaseModel
+from fastapi import FastAPI, status
+from starlette.responses import JSONResponse
+from MachineTranslation import MachineTranslation
+from utils import PROVIDED_LANGUAGES
+from fastapi.middleware.cors import CORSMiddleware
+import time
+
+class RequestBody(BaseModel):
+    text: str
+    from_lang: str
+    to_lang: str
+
+app = FastAPI(docs_url=None, redoc_url=None)
+machine_translation = MachineTranslation("facebook/nllb-200-distilled-600M")
+
+origins = ['*']
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=origins,
+    allow_credentials=True,
+    allow_methods=["*"],
+    allow_headers=["*"],
+)
+
+@app.post("/predict")
+async def predict(data_request: RequestBody):
+    if not data_request.text:
+        return JSONResponse({
+            "errors": "Please fill text!"
+        }, status_code=status.HTTP_400_BAD_REQUEST)
+    if len(data_request.text) > 5000:
+        return JSONResponse({
+            "errors": "The Number of Characters Exceeds The Limit"
+        }, status_code=status.HTTP_400_BAD_REQUEST)
+    if data_request.from_lang not in PROVIDED_LANGUAGES or data_request.to_lang not in PROVIDED_LANGUAGES:
+        return JSONResponse({
+            "errors": "Language not found!"
+        }, status_code=status.HTTP_400_BAD_REQUEST)
+    try:
+        # waktu prediksi
+        time_before = time.perf_counter()
+        result = machine_translation.predict(data_request.text, data_request.from_lang, data_request.to_lang)
+        time_after = time.perf_counter()
+        return JSONResponse({
+            "result": result,
+            "inference_time": time_after - time_before
+        }, status_code=status.HTTP_200_OK)
+    except Exception:
+        return JSONResponse({
+            "errors": "Please contact your administrator"
+        }, status_code=status.HTTP_500_INTERNAL_SERVER_ERROR)
+
+
+if __name__ == "__main__":
+    uvicorn.run(app, host="localhost", port=8000)

download_model.py

@@ -0,0 +1,7 @@
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
+
+tokenizer = AutoTokenizer.from_pretrained(
+            "facebook/nllb-200-distilled-600M")
+
+model = AutoModelForSeq2SeqLM.from_pretrained(
+            "facebook/nllb-200-distilled-600M")

requirements.txt

@@ -0,0 +1,7 @@
+pydantic==1.8.2
+starlette==0.14.2
+transformers==4.30.0
+uvicorn[standard]
+fastapi
+torch>=1.13.1
+torchvision>=0.14.1

utils.py

@@ -0,0 +1 @@
+PROVIDED_LANGUAGES = ["bjn_Latn", "ind_Latn",]