fix tests

tests/generate/test_integration_llamacpp.py

@@ -250,16 +250,22 @@ def test_llamacpp_cfg(model):
 
 
 @pytest.mark.parametrize(
-    "repo,model_path",
+    "repo,model_path,hf_tokenizer_uri",
     [
-        ("Qwen/Qwen1.5-0.5B-Chat-GGUF", "*q2*.gguf"),
-        ("QuantFactory/Meta-Llama-3-8B-Instruct-GGUF", "*Q2*.gguf"),
-        ("TheBloke/phi-2-GGUF", "*Q2*.gguf"),
-        ("QuantFactory/Phi-3-mini-4k-instruct-GGUF-v2", "*Q2*.gguf"),
+        ("Qwen/Qwen1.5-0.5B-Chat-GGUF", "*q2*.gguf", "Qwen/Qwen1.5-0.5B-Chat"),
+        ("TheBloke/phi-2-GGUF", "*Q2*.gguf", "microsoft/phi-2"),
     ],
 )
-def test_byte_tokenizer_regression(repo, model_path):
+def test_byte_tokenizer_regression(repo, model_path, hf_tokenizer_uri):
     """Reproduce https://github.com/outlines-dev/outlines/issues/820"""
-    model = models.llamacpp(repo, model_path)
+    import llama_cpp
+
+    model = models.llamacpp(
+        repo,
+        model_path,
+        tokenizer=llama_cpp.llama_tokenizer.LlamaHFTokenizer.from_pretrained(
+            hf_tokenizer_uri
+        ),
+    )
     generator = generate.choice(model, ["skirt", "dress", "pen", "jacket"])
     generator("Pick the odd word out: skirt, dress, pen, jacket")