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 +1,80 @@
+import XCTest
+@testable import MarkBase
+
+class AllModelsFinalTest: XCTestCase {
+    
+    func testAllModelsTextForwardFinal() throws {
+        print("\n═══════════════════════════════════════════════════════════════════")
+        print("  All Models TEXT Forward Final Test (Zero NaN Verification)")
+        print("═══════════════════════════════════════════════════════════════════\n")
+        
+        let engine = try MarkBaseEngine(autoCompile: true)
+        
+        let models = [
+            ("E2B", "/Users/accusys/MarkBaseEngine/models/gemma-4-e2b-it-4bit"),
+            ("26B-Standard", "/Users/accusys/MarkBaseEngine/models/gemma-4-26b-standard"),
+            ("31B", "/Users/accusys/MarkBaseEngine/models/gemma-4-31b-it-4bit"),
+            ("26B-A4B", "/Users/accusys/MarkBaseEngine/models/gemma-4-26b-a4b-it-4bit")
+        ]
+        
+        var successCount = 0
+        var failedModels: [(String, String)] = []
+        
+        for (modelName, modelPath) in models {
+            print("═══════════════════════════════════════════════════════════════════")
+            print("Testing: \(modelName)")
+            print("═══════════════════════════════════════════════════════════════════")
+            
+            guard FileManager.default.fileExists(atPath: modelPath) else {
+                print("  ⚠ Model not found, skipping")
+                failedModels.append((modelName, "Model not found"))
+                continue
+            }
+            
+            do {
+                print("  Loading model...")
+                let model = try E4BModel(modelDir: modelPath, engine: engine, maxContextLength: 128)
+                print("  ✓ Loaded")
+                print("    Layers: \(model.numHiddenLayers)")
+                print("    Hidden: \(model.hiddenSize)")
+                
+                print("  Testing forward pass...")
+                let result = try model.forwardOptimized(tokenId: 2, position: 0)
+                
+                let nanCount = result.filter { $0.isNaN }.count
+                print("  Forward result: NaN=\(nanCount)/\(result.count)")
+                
+                if nanCount == 0 {
+                    print("  ✓✓✓ Zero NaN - Success!")
+                    successCount += 1
+                } else {
+                    print("  ✗ NaN detected")
+                    failedModels.append((modelName, "NaN detected"))
+                }
+                
+            } catch {
+                print("  ✗ Failed: \(error)")
+                failedModels.append((modelName, error.localizedDescription))
+            }
+            
+            print()
+        }
+        
+        print("═══════════════════════════════════════════════════════════════════")
+        print("  Summary")
+        print("═══════════════════════════════════════════════════════════════════")
+        print("  Success: \(successCount)/\(models.count)")
+        
+        if failedModels.isEmpty {
+            print("  ✓✓✓✓✓✓ All models zero NaN!")
+        } else {
+            print("  Failed models:")
+            for (name, reason) in failedModels {
+                print("    - \(name): \(reason)")
+            }
+        }
+        print("═══════════════════════════════════════════════════════════════════\n")
+        
+        XCTAssertGreaterThanOrEqual(successCount, 1, "At least one model should succeed")
+    }
+}