Add lightweight tests and CI automation

.github/workflows/ci.yml

@@ -0,0 +1,35 @@
+name: CI
+
+on:
+  push:
+    branches: [ main ]
+  pull_request:
+
+jobs:
+  tests:
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v3
+
+      - name: Set up Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: '3.10'
+
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install torch --index-url https://download.pytorch.org/whl/cpu
+          pip install numpy pandas scikit-learn
+          pip install -r requirements-dev.txt
+
+      - name: Ruff (tests only)
+        run: ruff check tests
+
+      - name: Black (tests only)
+        run: black --check tests
+
+      - name: Run tests
+        run: pytest

README.md

@@ -245,3 +245,21 @@ Repository, çeşitli transformer varyantları için performans karşılaştırm
 - Model optimizasyon teknikleri analizi
 
 Detaylı sonuçlar için `Architecture/` dizinini ve generate edilen PNG dosyalarını kontrol edin. 
+
+## ✅ Test ve Kod Kalitesi
+
+Hafif testleri ve kod kalite kontrollerini çalıştırmak için isteğe bağlı geliştirme bağımlılıklarını yükleyin:
+
+```bash
+pip install -r requirements-dev.txt
+```
+
+Ardından aşağıdaki komutları çalıştırabilirsiniz:
+
+```bash
+pytest
+ruff check tests
+black --check tests
+```
+
+Sürekli entegrasyon iş akışı bu kontrolleri otomatik olarak yürütür.

pytest.ini

@@ -0,0 +1,2 @@
+[pytest]
+testpaths = tests

requirements-dev.txt

@@ -1,5 +1,4 @@
-# Development and Training Utilities
-jupyter>=1.0.0
-notebook>=6.5.0
-fairscale>=0.4.0
-deepspeed>=0.9.0
+# Optional development dependencies
+pytest>=7.4.0
+black>=23.9.0
+ruff>=0.1.9

tests/test_rope_modules.py

@@ -0,0 +1,53 @@
+import importlib.util
+from pathlib import Path
+
+import pytest
+
+torch = pytest.importorskip("torch")
+
+
+def load_rope_module(unique_name: str):
+    base_dir = Path(__file__).resolve().parents[1]
+    module_path = base_dir / "Architecture" / "partial-rope-full-rope.py"
+    spec = importlib.util.spec_from_file_location(unique_name, module_path)
+    module = importlib.util.module_from_spec(spec)
+    assert spec.loader is not None
+    spec.loader.exec_module(module)
+    return module
+
+
+def test_partial_rope_preserves_non_rotary_dimensions():
+    module = load_rope_module("rope_partial_module")
+    dim = 12
+    partial_factor = 0.5
+    rope = module.PartialRoPE(dim=dim, partial_rotary_factor=partial_factor)
+
+    torch.manual_seed(0)
+    q = torch.randn(1, 2, 3, dim)
+    k = torch.randn(1, 2, 3, dim)
+
+    q_embed, k_embed = rope(q.clone(), k.clone())
+
+    assert q_embed.shape == q.shape
+    assert k_embed.shape == k.shape
+
+    rotary_dim = int(dim * partial_factor)
+    assert torch.allclose(q_embed[..., rotary_dim:], q[..., rotary_dim:])
+    assert torch.allclose(k_embed[..., rotary_dim:], k[..., rotary_dim:])
+
+
+def test_attention_with_rope_output_shape():
+    module = load_rope_module("rope_attention_module")
+    dim = 16
+    num_heads = 4
+    head_dim = dim // num_heads
+    rope = module.FullRoPE(dim=head_dim)
+    attention = module.AttentionWithRoPE(dim=dim, num_heads=num_heads, rope_module=rope)
+
+    torch.manual_seed(1)
+    dummy_input = torch.randn(2, 5, dim)
+
+    output = attention(dummy_input)
+
+    assert output.shape == dummy_input.shape
+    assert torch.isfinite(output).all()

tests/test_time_series_transformer.py

@@ -0,0 +1,66 @@
+import importlib.util
+import sys
+from pathlib import Path
+
+import numpy as np
+import pandas as pd
+import pytest
+
+torch = pytest.importorskip("torch")
+
+
+def load_time_series_module(unique_name: str):
+    base_dir = Path(__file__).resolve().parents[1]
+    module_path = base_dir / "Time series - Transformers" / "train.py"
+    spec = importlib.util.spec_from_file_location(unique_name, module_path)
+    module = importlib.util.module_from_spec(spec)
+    assert spec.loader is not None
+    spec.loader.exec_module(module)
+    return module
+
+
+def test_time_series_transformer_forward_output_shape():
+    module = load_time_series_module("time_series_train_forward")
+    model = module.TimeSeriesTransformer(
+        input_dim=3, d_model=12, n_heads=3, num_layers=1
+    )
+
+    dummy_batch = torch.randn(2, 5, 3)
+    output = model(dummy_batch)
+
+    assert output.shape == (2, 1)
+    assert torch.isfinite(output).all()
+
+
+def test_train_creates_model_artifact(tmp_path, monkeypatch):
+    module = load_time_series_module("time_series_train_exec")
+
+    data = pd.DataFrame({"value": np.linspace(0, 1, 40)})
+    data_path = tmp_path / "synthetic.csv"
+    data.to_csv(data_path, index=False)
+
+    model_path = tmp_path / "model.pth"
+
+    monkeypatch.setattr(
+        sys,
+        "argv",
+        [
+            "train.py",
+            "--data",
+            str(data_path),
+            "--seq_length",
+            "5",
+            "--batch_size",
+            "4",
+            "--epochs",
+            "1",
+            "--lr",
+            "0.01",
+            "--model_path",
+            str(model_path),
+        ],
+    )
+
+    module.train()
+
+    assert model_path.exists()