keep llama and llama2 in tests

Author: jiqing-feng

tests/generation/test_modeling.py

@@ -30,7 +30,8 @@
     "gpt2": "hf-internal-testing/tiny-random-gpt2",
     "gpt_neo": "hf-internal-testing/tiny-random-GPTNeoModel",
     "mistral": "echarlaix/tiny-random-mistral",
-    "llama": "Jiqing/tiny_random_llama2",
+    "llama": "fxmarty/tiny-llama-fast-tokenizer",
+    "llama2": "Jiqing/tiny_random_llama2",
     "gpt_bigcode": "hf-internal-testing/tiny-random-GPTBigCodeModel",
 }
 
@@ -54,6 +55,7 @@ class ModelingIntegrationTest(unittest.TestCase):
         "gpt_neo",
         "mistral",
         "llama",
+        "llama2",
         # "gpt_bigcode",
     )
 

tests/ipex/test_inference.py

@@ -41,7 +41,8 @@
     "gpt_neo": "hf-internal-testing/tiny-random-GPTNeoModel",
     "gpt_neox": "hf-internal-testing/tiny-random-GPTNeoXForCausalLM",
     "gpt_bigcode": "hf-internal-testing/tiny-random-GPTBigCodeModel",
-    "llama": "Jiqing/tiny_random_llama2",
+    "llama": "fxmarty/tiny-llama-fast-tokenizer",
+    "llama2": "Jiqing/tiny_random_llama2",
     "opt": "hf-internal-testing/tiny-random-OPTModel",
     "mpt": "hf-internal-testing/tiny-random-MptForCausalLM",
 }
@@ -66,6 +67,7 @@ class IPEXIntegrationTest(unittest.TestCase):
         "gpt_neo",
         # "gpt_bigcode",
         "llama",
+        "llama2",
         "opt",
         "mpt",
     )

tests/ipex/test_modeling.py

@@ -66,7 +66,8 @@
     "gpt_neox": "hf-internal-testing/tiny-random-GPTNeoXForCausalLM",
     "gptj": "hf-internal-testing/tiny-random-GPTJModel",
     "levit": "hf-internal-testing/tiny-random-LevitModel",
-    "llama": "Jiqing/tiny_random_llama2",
+    "llama": "fxmarty/tiny-llama-fast-tokenizer",
+    "llama2": "Jiqing/tiny_random_llama2",
     "marian": "sshleifer/tiny-marian-en-de",
     "mbart": "hf-internal-testing/tiny-random-mbart",
     "mistral": "echarlaix/tiny-random-mistral",
@@ -209,6 +210,7 @@ class IPEXModelForCausalLMTest(unittest.TestCase):
         "gpt_neo",
         "gpt_neox",
         "llama",
+        "llama2",
         "mistral",
         # "phi",
         "mpt",
@@ -226,7 +228,7 @@ def test_compare_to_transformers(self, model_arch):
         self.assertTrue(ipex_model.use_cache)
         tokenizer = AutoTokenizer.from_pretrained(model_id)
         tokens = tokenizer(
-            "This is a sample", return_tensors="pt", return_token_type_ids=False if model_arch == "llama" else None
+            "This is a sample", return_tensors="pt", return_token_type_ids=False if model_arch in ("llama", "llama2") else None
         )
         position_ids = None
         if model_arch.replace("_", "-") in MODEL_TYPES_REQUIRING_POSITION_IDS: