Docker: muhitni ham qulflash

diagramma(
    {
        "Sizning mashinangiz\nPython 3.10, sklearn 1.7": (0, 0, 3.0, 1.1, "manba"),
        "Hamkasb mashinasi\nPython 3.12, sklearn 1.3": (0, 2.2, 3.0, 1.1, "manba"),
        "Server\nPython 3.9, sklearn 1.1": (0, 4.4, 3.0, 1.1, "manba"),
        "DOCKER IMAGE\n(aynan bitta muhit)": (4.2, 2.2, 3.0, 1.1, "qayta"),
        "Bir xil natija\nhamma joyda": (8.0, 2.2, 2.8, 1.1, "saqlash"),
    },
    [
        ("Sizning mashinangiz\nPython 3.10, sklearn 1.7", "DOCKER IMAGE\n(aynan bitta muhit)"),
        ("Hamkasb mashinasi\nPython 3.12, sklearn 1.3", "DOCKER IMAGE\n(aynan bitta muhit)"),
        ("Server\nPython 3.9, sklearn 1.1", "DOCKER IMAGE\n(aynan bitta muhit)"),
        ("DOCKER IMAGE\n(aynan bitta muhit)", "Bir xil natija\nhamma joyda"),
    ],
    nom="docker_nima_uchun.png", kenglik=13,
    izoh="Docker muhitni ham artefaktga aylantiradi: Python, kutubxonalar va tizim fayllari birga qadoqlanadi.",
)

docker --version
docker compose version

# ENG SODDA variant — ataylab optimallashtirilmagan.
FROM python:3.11

WORKDIR /app
COPY . .
RUN pip install scikit-learn pandas fastapi uvicorn

CMD ["python", "-c", "print('salom')"]

# Image ga TUSHMASLIGI kerak bo'lgan narsalar
.venv/
.ombor/
__pycache__/
*.pyc
.pytest_cache/
.mypy_cache/
.git/
*.csv
*.parquet
*.json
modellar/
namuna/

docker build -q -f Dockerfile.sodda -t mlops-kitob:sodda . 2>&1 | tail -2
docker images mlops-kitob:sodda --format "sodda variant: {{.Size}}"

# 1-QADAM: quruvchi (builder) — bu yerda kutubxonalar o'rnatiladi
FROM python:3.11-slim AS quruvchi

WORKDIR /app

# Avval FAQAT bog'liqliklar ro'yxati nusxalanadi.
# Sabab: kod o'zgarganda bu qatlam KESHDAN olinadi, pip qayta ishlamaydi.
COPY talablar.txt .
RUN pip install --no-cache-dir --user -r talablar.txt

# 2-QADAM: yakuniy image — faqat kerakli narsalar
FROM python:3.11-slim

# xavfsizlik: root emas, oddiy foydalanuvchi
RUN useradd --create-home --uid 10001 ishchi
USER ishchi
WORKDIR /app

# quruvchidan faqat o'rnatilgan paketlar ko'chiriladi
COPY --from=quruvchi --chown=ishchi:ishchi /root/.local /home/ishchi/.local
ENV PATH=/home/ishchi/.local/bin:$PATH \
    PYTHONUNBUFFERED=1 \
    PYTHONDONTWRITEBYTECODE=1

COPY --chown=ishchi:ishchi src/ ./src/

CMD ["python", "-c", "import sklearn, pandas; print('tayyor:', sklearn.__version__)"]

scikit-learn==1.7.2
pandas==2.3.3
numpy==2.2.6

docker build -q -t mlops-kitob:yaxshi . 2>&1 | tail -2
docker images --format "{{.Repository}}:{{.Tag}}  {{.Size}}" | grep mlops-kitob

docker images --format "{{.Tag}}|{{.Size}}" | grep -E "^(sodda|yaxshi)" | while IFS='|' read tag size; do
  printf "  %-8s %s\n" "$tag" "$size"
done

import pandas as pd
show(pd.DataFrame([
    {"o'zgarish": "python:3.11 -> python:3.11-slim",
     "sabab": "Asosiy image dan kompilyator va keraksiz tizim paketlari olib tashlangan"},
    {"o'zgarish": "Ko'p bosqichli qurish (multi-stage)",
     "sabab": "pip ning keshi va qurish artefaktlari yakuniy image ga tushmaydi"},
    {"o'zgarish": "--no-cache-dir",
     "sabab": "pip yuklab olgan .whl fayllarini saqlamaydi"},
    {"o'zgarish": ".dockerignore",
     "sabab": "Ma'lumot, .venv va __pycache__ umuman nusxalanmaydi"},
    {"o'zgarish": "COPY src/ (hammasi emas)",
     "sabab": "Faqat kerakli kod; test va hujjatlar tushmaydi"},
]))

echo "# kichik o'zgarish $(date +%s)" >> src/quvur/takrorlanish.py
t0=$SECONDS
docker build -q -t mlops-kitob:yaxshi . > /dev/null 2>&1
echo "kod o'zgardi, qayta qurish: $((SECONDS - t0)) s   (pip KESHDAN olindi)"

# talablar.txt ga o'zgarish kiritamiz -> COPY talablar.txt qatlami eskiradi
echo "# qayta qurish belgisi $(date +%s%N)" >> talablar.txt
t0=$SECONDS
docker build -q -t mlops-kitob:yaxshi . > /dev/null 2>&1
echo "talablar.txt o'zgardi: $((SECONDS - t0)) s   (pip QAYTA ishladi)"

docker run --rm mlops-kitob:yaxshi

docker run --rm mlops-kitob:yaxshi python -c "
import sys, sklearn, pandas, numpy, getpass, os
print('Python   :', sys.version.split()[0])
print('sklearn  :', sklearn.__version__)
print('pandas   :', pandas.__version__)
print('numpy    :', numpy.__version__)
print('foydalanuvchi:', os.getuid(), '(0 = root, xavfli)')
"

"""Modelni o'qitib, faylga saqlaydi."""
import sys
from pathlib import Path

import joblib
import pandas as pd
from sklearn.ensemble import RandomForestClassifier
from sklearn.model_selection import train_test_split
from sklearn.metrics import f1_score

XUS = ["xonalar", "maydon", "qavat", "jami_qavat", "narx_mln", "tavsif_uzunligi",
       "rasmlar_soni", "telefon_takrorlanishi", "kunlik_elonlar", "egasi_tasdiqlangan"]


def oqit(malumot: str, chiqish: str) -> float:
    d = pd.read_csv(malumot)
    Xtr, Xte, ytr, yte = train_test_split(d[XUS], d["broker"], test_size=0.25,
                                          stratify=d["broker"], random_state=42)
    m = RandomForestClassifier(n_estimators=150, min_samples_leaf=3,
                               random_state=42, n_jobs=1).fit(Xtr, ytr)
    f1 = float(f1_score(yte, m.predict(Xte)))
    Path(chiqish).parent.mkdir(parents=True, exist_ok=True)
    joblib.dump({"model": m, "xususiyatlar": XUS, "f1": round(f1, 4)}, chiqish)
    return f1


if __name__ == "__main__":
    malumot = sys.argv[1] if len(sys.argv) > 1 else "datasets/ijara.csv"
    chiqish = sys.argv[2] if len(sys.argv) > 2 else "modellar/model.joblib"
    print(f"F1 = {oqit(malumot, chiqish):.4f}  ->  {chiqish}")

sed -i 's/^joblib.*$//' talablar.txt 2>/dev/null
grep -q joblib talablar.txt || echo "joblib==1.5.3" >> talablar.txt
docker build -q -t mlops-kitob:yaxshi . > /dev/null 2>&1
mkdir -p modellar
docker run --rm \
  --mount type=bind,src="$(pwd)/../datasets",dst=/app/datasets,ro \
  --mount type=bind,src="$(pwd)/modellar",dst=/app/modellar \
  -w /app mlops-kitob:yaxshi \
  python src/oqitish/oddiy.py datasets/ijara.csv modellar/model.joblib
ls -la modellar/

python3 -c "
import joblib
m = joblib.load('modellar/model.joblib')
print('model turi   :', type(m['model']).__name__)
print('daraxtlar    :', m['model'].n_estimators)
print('xususiyatlar :', len(m['xususiyatlar']))
print('F1 (o\'qitishda):', m['f1'])
"

docker rmi -f mlops-kitob:sodda > /dev/null 2>&1
docker images --format "{{.Repository}}:{{.Tag}}" | grep -c "^mlops-kitob" || true
echo "qolgan mlops-kitob image lari yuqorida"

import pandas as pd
show(pd.DataFrame([
    {"muammo": "Kutubxona versiyalari", "Docker hal qiladimi": "HA", "izoh": "Image ichida qulflangan"},
    {"muammo": "Tizim kutubxonalari (libgomp)", "Docker hal qiladimi": "HA", "izoh": "Image ga kiradi"},
    {"muammo": "Python versiyasi", "Docker hal qiladimi": "HA", "izoh": "FROM da belgilanadi"},
    {"muammo": "Ma'lumot versiyasi", "Docker hal qiladimi": "YO'Q", "izoh": "5-bob: manifest va ombor"},
    {"muammo": "Tasodifiy seed", "Docker hal qiladimi": "YO'Q", "izoh": "2-bob: kodda qulflanadi"},
    {"muammo": "CPU arxitekturasi (arm64/amd64)", "Docker hal qiladimi": "QISMAN", "izoh": "Har platformaga alohida image"},
    {"muammo": "GPU drayverlari", "Docker hal qiladimi": "QISMAN", "izoh": "nvidia-container-toolkit kerak"},
]))