Switch test to microsoft tokenizer, verify in Python

Author: pcuenca

Tests/TokenizersTests/AddedTokensTests.swift

@@ -11,12 +11,12 @@ import Hub
 
 class AddedTokensTests: XCTestCase {
     func testPhiAddedTokens() async throws {
-        let tokenizer = try await AutoTokenizer.from(pretrained: "mlx-community/Phi-3-mini-128k-instruct-4bit")
+        let tokenizer = try await AutoTokenizer.from(pretrained: "microsoft/Phi-3-mini-128k-instruct")
         let inputIds = tokenizer("This is the <|end|>. My only friend, the <|end|>")
-        XCTAssertEqual(inputIds, [1, 910, 338, 278, 29871, 32007, 29889, 1619, 871, 5121, 29892, 278, 29871, 32007])
+        XCTAssertEqual(inputIds, [910, 338, 278, 29871, 32007, 29889, 1619, 871, 5121, 29892, 278, 29871, 32007])
 
         let decoded = tokenizer.decode(tokens: inputIds)
-        XCTAssertEqual(decoded, "<s> This is the <|end|>. My only friend, the <|end|>")
+        XCTAssertEqual(decoded, "This is the <|end|>. My only friend, the <|end|>")
     }
 
     func testGemmaAddedTokens() async throws {