roll back

Dockerfile

@@ -1,4 +1,3 @@
-# syntax=docker/dockerfile:1
 # Read the doc: https://huggingface.co/docs/hub/spaces-sdks-docker
 # you will also find guides on how best to write your Dockerfile
 
@@ -7,16 +6,13 @@ FROM python:3.9
 RUN useradd -m -u 1000 user && mkdir -p /app/models && chown -R user:user /app
 USER user
 ENV PATH="/home/user/.local/bin:$PATH"
-ENV HF_MODEL_REPO_ID="piyazon/uyghur_custom_300"
-ENV HF_MODEL_FILENAME="uyghur_custom_300.bin"
-ENV FASTTEXT_MODEL_PATH="/app/models/uyghur_custom_300.bin"
+ENV FASTTEXT_MODEL_PATH="/app/models/cc.ug.300.bin"
 
 WORKDIR /app
 
 COPY --chown=user ./requirements.txt requirements.txt
 RUN pip install --no-cache-dir --upgrade -r requirements.txt
-RUN --mount=type=secret,id=HF_TOKEN,mode=0444,required=true \
-    python -c "import os; from pathlib import Path; from huggingface_hub import hf_hub_download; token = Path('/run/secrets/HF_TOKEN').read_text().strip(); print('Downloading Hugging Face model:', os.environ['HF_MODEL_REPO_ID'] + '/' + os.environ['HF_MODEL_FILENAME'], flush=True); hf_hub_download(repo_id=os.environ['HF_MODEL_REPO_ID'], filename=os.environ['HF_MODEL_FILENAME'], local_dir='/app/models', token=token)"
+RUN python -c "import gzip, os, shutil, urllib.request; url='https://dl.fbaipublicfiles.com/fasttext/vectors-crawl/cc.ug.300.bin.gz'; gz='/app/models/cc.ug.300.bin.gz'; out='/app/models/cc.ug.300.bin'; print('Downloading fastText model:', url, flush=True); urllib.request.urlretrieve(url, gz); print('Extracting fastText model to:', out, flush=True); f_in=gzip.open(gz, 'rb'); f_out=open(out, 'wb'); shutil.copyfileobj(f_in, f_out); f_in.close(); f_out.close(); os.remove(gz)"
 
 COPY --chown=user . /app
 CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]

README.md

@@ -11,11 +11,9 @@ Check out the configuration reference at https://huggingface.co/docs/hub/spaces-
 
 ## API
 
-The Docker build downloads the custom fastText Uyghur binary model from `piyazon/uyghur_custom_300` on Hugging Face, saves it to `/app/models/uyghur_custom_300.bin`, and loads it once when the FastAPI app starts.
+The Docker build downloads the official fastText Uyghur Common Crawl binary model, extracts it to `/app/models/cc.ug.300.bin`, and loads it once when the FastAPI app starts.
 
-Model source: https://huggingface.co/piyazon/uyghur_custom_300
-
-Because the model repository is private, add a Space secret named `HF_TOKEN` with read access to `piyazon/uyghur_custom_300`. The Dockerfile exposes that secret only during the model download build step.
+Model source: https://fasttext.cc/docs/en/crawl-vectors
 
 ### Query request
 

app.py

@@ -9,7 +9,7 @@ from fastapi import FastAPI, HTTPException, Query
 from pydantic import BaseModel, Field
 
 
-MODEL_PATH = Path(os.getenv("FASTTEXT_MODEL_PATH", "/app/models/uyghur_custom_300.bin"))
+MODEL_PATH = Path(os.getenv("FASTTEXT_MODEL_PATH", "/app/models/cc.ug.300.bin"))
 
 
 class SimilarityRequest(BaseModel):
@@ -21,7 +21,7 @@ def load_fasttext_model() -> Any:
     if not MODEL_PATH.exists():
         raise RuntimeError(
             f"fastText model not found at {MODEL_PATH}. "
-            "Set FASTTEXT_MODEL_PATH or download uyghur_custom_300.bin during the Docker build."
+            "Set FASTTEXT_MODEL_PATH or download cc.ug.300.bin during the Docker build."
         )
 
     return fasttext.load_model(str(MODEL_PATH))

requirements.txt

@@ -1,5 +1,4 @@
 fastapi
 uvicorn[standard]
 fasttext
-huggingface_hub[hf_xet]
 numpy