- 支持从 PostgreSQL 数据库拉取结构化数据作为训练来源 - 新增 DPO (Direct Preference Optimization) 输出类型 - 支持 chosen/rejected 字段的编辑、校验和发布 - 完善数据预处理切分逻辑和元数据管理 - 移除 OCR 扫描 PDF 功能,保持基础文本解析能力 Co-Authored-By: Claude <noreply@anthropic.com>
26 lines
882 B
Docker
26 lines
882 B
Docker
FROM python:3.12-slim
|
|
|
|
ENV PYTHONDONTWRITEBYTECODE=1 \
|
|
PYTHONUNBUFFERED=1 \
|
|
PIP_NO_CACHE_DIR=1 \
|
|
TIKTOKEN_CACHE_DIR=/opt/tiktoken_cache
|
|
|
|
WORKDIR /app
|
|
|
|
COPY backend/requirements.txt /tmp/requirements.txt
|
|
RUN pip install --upgrade pip -i https://pypi.tuna.tsinghua.edu.cn/simple \
|
|
&& pip install -r /tmp/requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple \
|
|
&& rm -f /tmp/requirements.txt
|
|
|
|
RUN python -c "import fastapi, uvicorn, psycopg, psycopg_pool, sqlalchemy, redis, jwt, passlib, httpx, alembic; print('backend dependency check ok')"
|
|
|
|
RUN mkdir -p /opt/yg-ft/logs/backend /data/yg-ft \
|
|
&& chmod -R 0775 /opt/yg-ft /data/yg-ft
|
|
|
|
# 离线打包 tiktoken cl100k_base 词表,避免无网环境下运行时联网下载
|
|
COPY docker/app/tiktoken /opt/tiktoken_cache
|
|
|
|
EXPOSE 8000
|
|
|
|
CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "8000"]
|