Initial commit: E4B-MarkBase model integration with passing tests

- E4B-MarkBase model (42 layers, 4.4GB) loaded successfully - All Phase 1-6 tests passed (model loading, forward pass, vision/audio towers, token generation, performance) - All stress tests passed (5/5 in 127.6s) - Concurrent inference - Memory stress (67.5 tok/s, 0 NaN) - Continuous generation - Batch processing - Long-running stability - Swift Metal inference engine with multimodal support
2026-06-23 18:12:35 +08:00
commit ac75faa0cc
301 changed files with 63426 additions and 0 deletions
@@ -0,0 +1,54 @@
+import XCTest
+@testable import MarkBase
+
+final class MoETests: XCTestCase {
+    
+    func test26BA4BLoading() async throws {
+        let modelPath = "/Users/accusys/MarkBaseEngine/models/gemma-4-26b-a4b-it-4bit"
+        
+        print("\n=====================================")
+        print("Testing 26B-A4B MoE Model Loading")
+        print("=====================================")
+        
+        let start = Date()
+        let engine = try MarkBaseEngine(autoCompile: true)
+        let model = try E4BModel(modelDir: modelPath, engine: engine, maxContextLength: 512)
+        let loadTime = Date().timeIntervalSince(start)
+        
+        print("✓ Model loaded in \(String(format: "%.3f", loadTime))s")
+        print("  Layers: \(model.numHiddenLayers)")
+        print("  Hidden: \(model.hiddenSize)")
+        print("  Vocab: \(model.vocabSize)")
+        
+        XCTAssertGreaterThan(model.numHiddenLayers, 0)
+        XCTAssertEqual(model.hiddenSize, 2816)
+        XCTAssertEqual(model.vocabSize, 262144)
+    }
+    
+    func test26BA4BGeneration() async throws {
+        let modelPath = "/Users/accusys/MarkBaseEngine/models/gemma-4-26b-a4b-it-4bit"
+        
+        print("\n=====================================")
+        print("Testing 26B-A4B MoE Token Generation")
+        print("=====================================")
+        
+        let engine = try MarkBaseEngine(autoCompile: true)
+        let model = try E4BModel(modelDir: modelPath, engine: engine, maxContextLength: 512)
+        let tokenizer = try TokenizerFactory.load(modelDir: modelPath)
+        let generator = StreamingGenerator(model: model, tokenizer: tokenizer, engine: engine)
+        
+        let prompt = "Hello"
+        let config = GenerationConfig(maxTokens: 10, temperature: Float(0.7))
+        
+        print("Testing generation...")
+        let start = Date()
+        let response = try generator.generateComplete(prompt: prompt, config: config)
+        let genTime = Date().timeIntervalSince(start)
+        
+        print("✓ Generated in \(String(format: "%.3f", genTime))s")
+        print("  Response: \"\(response)\"")
+        print("  Speed: \(String(format: "%.2f", 10.0 / genTime)) tok/s")
+        
+        XCTAssertGreaterThan(response.count, 0)
+    }
+}