translate model facebook/seamless-m4t-v2-large

.gitignore

@@ -0,0 +1,22 @@
+# Build Artifacts
+build/
+
+# Core Dumps
+core
+
+# Byte-Compiled Modules
+__pycache__/
+
+# Extension Modules
+*.so
+
+# Packaging Artifacts
+*.egg-info
+*.whl
+
+# IDEs and Tools
+.idea/
+.gdb_history
+.vscode/
+# Other
+.DS_Store

Dockerfile

@@ -0,0 +1,45 @@
+FROM nvidia/cuda:12.1.0-cudnn8-devel-ubuntu20.04
+LABEL maintainer="Hugging Face"
+
+ARG DEBIAN_FRONTEND=noninteractive
+
+RUN apt update
+RUN apt install -y git libsndfile1-dev tesseract-ocr espeak-ng python3 python3-pip ffmpeg
+RUN python3 -m pip install --no-cache-dir --upgrade pip
+# RUN apt-get install -y git libsndfile1-dev tesseract-ocr espeak-ng ffmpeg
+
+# Prevents Python from writing pyc files.
+ENV PYTHONDONTWRITEBYTECODE=1
+# Open MP threads. It may need to change in production env.
+ENV OMP_NUM_THREADS=1
+
+# Keeps Python from buffering stdout and stderr to avoid situations where
+# the application crashes without emitting any logs due to buffering.
+ENV PYTHONUNBUFFERED=1
+
+ENV HF_HUB_CACHE="/hub"
+
+WORKDIR /app
+
+# Create a non-privileged user that the app will run under.
+# See https://docs.docker.com/go/dockerfile-user-best-practices/
+RUN useradd -m -u 1000 user
+# Download dependencies as a separate step to take advantage of Docker's caching.
+# Leverage a cache mount to /root/.cache/pip to speed up subsequent builds.
+# Leverage a bind mount to requirements.txt to avoid having to copy them into
+# into this layer.
+RUN --mount=type=cache,target=/root/.cache/pip \
+  --mount=type=bind,source=requirements.txt,target=requirements.txt \
+  python3 -m pip install -r requirements.txt
+
+# Switch to the non-privileged user to run the application.
+USER user
+
+# Copy the source code into the container.
+COPY . .
+
+# Expose the port that the application listens on.
+EXPOSE 8088
+
+# Run the application.
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]

README.md

@@ -9,3 +9,19 @@ license: cc-by-4.0
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+
+# Translate
+
+## Usage
+
+```
+conda activate hf
+```
+
+VS Code Python select interpreter hf
+
+## Composa
+
+```
+docker compose up --build
+```

compose.yaml

@@ -0,0 +1,54 @@
+# Comments are provided throughout this file to help you get started.
+# If you need more help, visit the Docker compose reference guide at
+# https://docs.docker.com/go/compose-spec-reference/
+
+# Here the instructions define your application as a service called "server".
+# This service is built from the Dockerfile in the current directory.
+# You can add other services your application may depend on here, such as a
+# database or a cache. For examples, see the Awesome Compose repository:
+# https://github.com/docker/awesome-compose
+services:
+  server:
+    build:
+      context: .
+    ports:
+      - 8089:7860
+    volumes:
+#      - ${PWD}:/app:rw
+      - /Users/hugorodrigues/.cache/huggingface/hub:/hub:rw
+    environment:
+      - HF_HUB_CACHE=/hub
+
+# The commented out section below is an example of how to define a PostgreSQL
+# database that your application can use. `depends_on` tells Docker Compose to
+# start the database before your application. The `db-data` volume persists the
+# database data between container restarts. The `db-password` secret is used
+# to set the database password. You must create `db/password.txt` and add
+# a password of your choosing to it before running `docker compose up`.
+#     depends_on:
+#       db:
+#         condition: service_healthy
+#   db:
+#     image: postgres
+#     restart: always
+#     user: postgres
+#     secrets:
+#       - db-password
+#     volumes:
+#       - db-data:/var/lib/postgresql/data
+#     environment:
+#       - POSTGRES_DB=example
+#       - POSTGRES_PASSWORD_FILE=/run/secrets/db-password
+#     expose:
+#       - 5432
+#     healthcheck:
+#       test: [ "CMD", "pg_isready" ]
+#       interval: 10s
+#       timeout: 5s
+#       retries: 5
+# volumes:
+#   db-data:
+# secrets:
+#   db-password:
+#     file: db/password.txt
+

main.py

@@ -0,0 +1,65 @@
+import time
+
+# from typing import Union
+# from pydantic import BaseModel
+from fastapi import FastAPI
+from fastapi.middleware.cors import CORSMiddleware
+
+# from fastapi.staticfiles import StaticFiles
+# from fastapi.responses import FileResponse
+
+import torch
+# from transformers import pipeline
+
+from transformers import SeamlessM4Tv2Model
+from transformers import AutoProcessor
+
+processor = AutoProcessor.from_pretrained("facebook/seamless-m4t-v2-large")
+model = SeamlessM4Tv2Model.from_pretrained("facebook/seamless-m4t-v2-large")
+
+
+device = "cuda:0" if torch.cuda.is_available() else "cpu"
+# torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
+
+model.to(device)
+
+app = FastAPI(docs_url="/api/docs")
+
+app.add_middleware(
+    CORSMiddleware,
+    allow_origins=["*"],
+    allow_methods=["*"],
+    allow_headers=["*"],
+    allow_credentials=True,
+)
+
+BATCH_SIZE = 8
+
+
+@app.get("/device")
+def getDevice():
+    start_time = time.time()
+    print("Time took to process the request and return response is {} sec".format(
+        time.time() - start_time))
+    return device
+
+
+@app.get("/translate")
+def transcribe(inputs, src_lang="eng", tgt_lang="por"):
+    start_time = time.time()
+
+    if inputs is None:
+        raise "No audio file submitted! Please upload or record an audio file before submitting your request."
+
+    text_inputs = processor(text=inputs,
+                            src_lang=src_lang, return_tensors="pt").to(device)
+
+    output_tokens = model.generate(
+        **text_inputs, tgt_lang=tgt_lang, generate_speech=False)
+
+    translated_text_from_text = processor.decode(
+        output_tokens[0].tolist()[0], skip_special_tokens=True)
+
+    print("Time took to process the request and return response is {} sec".format(
+        time.time() - start_time))
+    return translated_text_from_text

packages.txt

@@ -0,0 +1 @@
+ffmpeg

requirements.txt

@@ -0,0 +1,10 @@
+fastapi
+pydantic
+typing
+transformers
+python-multipart
+sentencepiece
+protobuf
+torch
+uvicorn[standard]
+ffmpeg