Tensorflow Dependency Switching¶

Overview¶

This log demonstrates Complex Dependency Tree Management.

While the previous demo showed swapping a single binary, this one shows swapping TensorFlow (a massive package) along with its sub-dependencies (typing_extensions, keras) in a nested context. It proves OmniPkg handles the “Butterfly Effect” of dependencies—changing one package updates the entire graph instantly.

TensorFlow Test¶

This demo validates OmniPkg’s ability to handle massive, complex dependency graphs. Unlike simple packages, TensorFlow brings along a huge tree of dependencies (Keras, TensorBoard, Protobuf, etc.). OmniPkg must ensure all of them are swapped correctly to prevent ABI mismatches.

Usage¶

# Run Demo #4: Complex Dependency Graph Switching
omnipkg demo 4

What You’ll Learn¶

1. Deep Graph Swapping: How activating tensorflow==2.13.0 automatically pulls in the correct keras, typing_extensions, and numpy versions.
2. Nested Contexts: How to use omnipkgLoader inside another omnipkgLoader block (Inception-style environment management).
3. Strict Mode Verification: Verifying that no “cloaked” modules (old versions) leak into the new context.

The Code¶

This script performs a nested activation: it first activates a specific version of typing_extensions, and then inside that block, activates an old version of tensorflow that requires a different typing_extensions. OmniPkg handles this conflict resolution in real-time.

from omnipkg.loader import omnipkgLoader

# 1. Outer Context: Activate older typing_extensions
with omnipkgLoader("typing_extensions==4.5.0"):
    import typing_extensions as te
    print(f"Outer Version: {te.__version__}")
    # Output: 4.5.0

    # 2. Inner Context: Activate TensorFlow 2.13.0
    # TensorFlow 2.13.0 might require a different typing_extensions.
    # OmniPkg creates a NEW nested bubble for this block.
    with omnipkgLoader("tensorflow==2.13.0"):
        import tensorflow as tf
        import typing_extensions as te_inner

        print(f"Inner TF Version: {tf.__version__}")
        # Output: 2.13.0

        print(f"Inner TE Version: {te_inner.__version__}")
        # Output: Matches TF's requirement (automatically resolved)

        # Verify the complex graph works by creating a model
        model = tf.keras.Sequential([tf.keras.layers.Dense(1)])
        print("Model created successfully")

    # 3. Back to Outer Context
    # TensorFlow is unloaded; typing_extensions reverts to 4.5.0
    import typing_extensions as te_back
    print(f"Restored Version: {te_back.__version__}")
    # Output: 4.5.0

Live Execution Log¶

Notice the Nested activation (depth=2) line in the logs. This confirms that OmniPkg is maintaining a stack of environments and correctly popping them off as the code exits the with blocks.

(evocoder_env) minds3t@aiminingrig:~/omnipkg$ omnipkg demo 4

================================================================================
  🚀 🚨 OMNIPKG TENSORFLOW DEPENDENCY SWITCHING TEST 🚨
================================================================================

--- Nested Loader Test ---
🌀 Testing nested loader usage...
   ✅ Outer context - Typing Extensions: 4.5.0

🚀 Fast-activating tensorflow==2.13.0 ...
   📂 Searching for bubble: .../tensorflow-2.13.0
   📊 Bubble: 40 packages, 0 conflicts
   🧹 Purging 40 module(s) from memory...
   - 🔒 STRICT mode
   ⚡ HEALED in 25,743.3 μs
   ✅ Inner context - TensorFlow: 2.13.0
   ✅ Inner context - Typing Extensions: 4.5.0
   ✅ Nested loader test: Model created successfully

🌀 omnipkg loader: Deactivating tensorflow==2.13.0 [depth=2]...
   ✅ Environment restored.
   ⏱️  Swap Time: 26,032.537 μs

🌀 omnipkg loader: Deactivating typing_extensions==4.5.0...
   ✅ Environment restored.
   ⏱️  Swap Time: 125,331.240 μs

================================================================================
  🚀 STEP 3: Test Results Summary
================================================================================
Test 1 (TensorFlow 2.13.0 Bubble): ✅ PASSED
Test 2 (Dependency Switching): ✅ PASSED
Test 3 (Nested Loaders): ✅ PASSED

Overall: 3/3 tests passed
🎉 DEMO PASSED! 🎉

How It Works¶

1. Graph Resolution: When tensorflow==2.13.0 is requested, the LibResolver walks its dependency tree. It identifies 40+ sub-dependencies.
2. Memory Purge: It unloads all 40 modules from sys.modules to ensure no stale code remains.
3. Stack Management: The omnipkgLoader tracks the “depth” of the context. When the inner block exits, it doesn’t just clear sys.path—it restores the sys.path of the outer block (restoring typing_extensions==4.5.0).